<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Startup Intelligence blog</title>
  <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/"/>
  <link rel="self" type="application/atom+xml" href="https://www.startupintelligence.si/feed.xml"/>
  <id>https://www.startupintelligence.si/blog/</id>
  <updated>2026-10-05T00:00:00Z</updated>
  <entry>
    <title>Assumptions that look safe in a business plan, and are not</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/assumptions-that-look-safe-and-are-not/"/>
    <id>https://www.startupintelligence.si/blog/assumptions-that-look-safe-and-are-not/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>The assumptions that sink plans are often the ones nobody doubts: market share, channel cost, cost to serve, retention. How to find them by working backwards.</summary>
    <content type="html">&lt;h2 id="a-plan-that-passed-every-test-it-was-given"&gt;A plan that passed every test it was given&lt;/h2&gt;
&lt;p&gt;Picture a founder planning lunch subscriptions for office workers in one city. This is a made-up example, and every detail and number in it is invented for illustration. They have done the work most guides recommend. They spoke to forty office workers, and most said the lunch options near work are poor and expensive. They priced a few rival services. Their spreadsheet reaches break-even in the second year. Every assumption they worried about, they checked.&lt;/p&gt;
&lt;p&gt;The trouble is the assumptions they did not worry about. How much does it cost to win each new subscriber once friends and colleagues have signed up? How many subscribers cancel in the third month, when the novelty wears off? What does a missed delivery cost in refunds and in time? None of these felt risky, so none was tested. Yet any one of them can decide whether the plan works.&lt;/p&gt;
&lt;p&gt;This is a common shape. Founders test the assumptions they can see. The ones that sink plans are often the ones that look too safe to question.&lt;/p&gt;
&lt;h2 id="why-some-assumptions-feel-safe"&gt;Why some assumptions feel safe&lt;/h2&gt;
&lt;p&gt;Assumptions slip past review for a few recurring reasons.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;They describe an outcome, not a cause.&lt;/strong&gt; &amp;quot;We will win a small share of a large market&amp;quot; sounds modest, so it feels safe. But it says nothing about who adopts first, why they adopt, or how one customer leads to the next.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;They live in the parts of the plan nobody enjoys.&lt;/strong&gt; Founders think hard about the product and the customer&amp;#x27;s pain. They think less about the channel, the cost to serve and what keeps customers paying.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;They are borrowed.&lt;/strong&gt; A retention rate or a conversion rate copied from a different business, country or price point looks like evidence. It is really an assumption about how similar the two situations are.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;They are demanded, not supported.&lt;/strong&gt; The target needs a figure to be true, so the figure goes into the sheet. Nobody asks whether anyone has ever achieved it.&lt;/li&gt;&lt;/ul&gt;
&lt;h2 id="five-assumptions-that-often-look-safe"&gt;Five assumptions that often look safe&lt;/h2&gt;
&lt;h3 id="we-only-need-a-small-share-of-the-market"&gt;&amp;quot;We only need a small share of the market&amp;quot;&lt;/h3&gt;
&lt;p&gt;This is a familiar one. It turns a hard question (why would these people switch, and in what order?) into easy arithmetic. A better form names a mechanism: the first customers are this group, they adopt because of this, and each one brings the next through this channel at this cost.&lt;/p&gt;
&lt;p&gt;Startup Intelligence takes this seriously enough to refuse the shortcut. When the engine builds the model of your outcome, any part of it that is a bare share of a total market, such as &amp;quot;we capture 2% of the market&amp;quot;, is rejected. Adoption is modelled instead as a curve with two rates: people who adopt on their own, and people who adopt because others already have.&lt;/p&gt;
&lt;h3 id="customers-will-find-us"&gt;&amp;quot;Customers will find us&amp;quot;&lt;/h3&gt;
&lt;p&gt;The path to the customer is one of the nine claim links a complete idea states. It is also an easy link to leave as a hope. The assumption to write down is specific: which channel, what it costs to reach one more paying customer through it, and whether that cost holds after the easiest customers are gone.&lt;/p&gt;
&lt;h3 id="serving-one-more-customer-costs-almost-nothing"&gt;&amp;quot;Serving one more customer costs almost nothing&amp;quot;&lt;/h3&gt;
&lt;p&gt;It is tempting, especially for software, to assume the cost to serve is close to zero. In practice onboarding, support, refunds, payment fees and manual work can grow with every customer. Write the cost to serve one more customer as a number with a range, and test it by doing the work yourself for a handful of customers.&lt;/p&gt;
&lt;h3 id="once-they-start-they-will-stay"&gt;&amp;quot;Once they start, they will stay&amp;quot;&lt;/h3&gt;
&lt;p&gt;Many plans quietly assume customers keep paying. Retention is often one of the figures a revenue target leans on most, and it is hard to know before customers have had time to leave. Ask what builds up over time that makes staying easier than leaving. If the honest answer is &amp;quot;nothing yet&amp;quot;, the plan should not assume loyalty.&lt;/p&gt;
&lt;h3 id="competitors-will-not-react"&gt;&amp;quot;Competitors will not react&amp;quot;&lt;/h3&gt;
&lt;p&gt;A plan that works only if nobody responds rests on an assumption about other people&amp;#x27;s behaviour. Two questions help: how late would an established player notice, and how hard could it respond? A slow, crushing response and a fast, weak one are very different futures.&lt;/p&gt;
&lt;h2 id="work-backwards-from-your-own-target"&gt;Work backwards from your own target&lt;/h2&gt;
&lt;p&gt;The most reliable way to expose a safe-looking assumption is to start from the result you need and work backwards.&lt;/p&gt;
&lt;p&gt;Take the made-up lunch plan again; these figures are invented for illustration. Say the founder&amp;#x27;s bar is ₹2 crore of annual recurring revenue by month 24, and the price is ₹2,000 a month. That is ₹24,000 a year per subscriber, so the bar needs about 834 paying subscribers at month 24. If some cancel every month, the founder has to win more than that, and keep winning, just to stand still. Each of those figures is now something the target demands. The question becomes: has any comparable service reached it?&lt;/p&gt;
&lt;p&gt;Startup Intelligence builds this step into every run. The report page &amp;quot;The model behind the number&amp;quot; works backwards from the threshold you set to the few figures that would have to hold, such as how many customers, at what price and at what retention. Where it can, it places each one against a class of comparable ventures. When a figure sits beyond what comparable ventures have achieved, the page says so plainly: &amp;quot;This plan requires outlier performance&amp;quot;. That is not a prediction that the plan fails. It is a statement that this figure needs evidence of its own instead of being assumed.&lt;/p&gt;
&lt;p&gt;The same page names what runs out first and how late and how hard an incumbent might respond, so the competitor assumption is written down too, as the model&amp;#x27;s own judgement.&lt;/p&gt;
&lt;h2 id="what-the-report-shows-and-its-limits"&gt;What the report shows, and its limits&lt;/h2&gt;
&lt;p&gt;The report&amp;#x27;s page &amp;quot;Everything it rests on&amp;quot; lists every assumption the engine extracted from your description, not just the five to eight that carry the most weight. Each is placed by how much the outcome leans on it and how well it is evidenced, and the heavy, weakly evidenced ones are marked as the corner to test first. That makes the quiet assumptions visible, including the ones you did not think to write down.&lt;/p&gt;
&lt;p&gt;Be clear about one limit. Today the engine consults no outside source during a run, so its comparisons with other ventures come from the AI model&amp;#x27;s own knowledge and are filed on the lowest rungs of the evidence ladder. Read an &amp;quot;outlier&amp;quot; flag as a prompt to go and check, not as a finding about the market.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Before you test the assumptions you are worried about, list the ones you are not: market share, channel cost, cost to serve, retention and competitor response. Then work backwards from your target and ask what it quietly demands. The figures that make you uncomfortable are the ones to test first. For how to rank them, read &lt;a href="/blog/load-bearing-assumptions-in-a-startup-idea/"&gt;load-bearing assumptions&lt;/a&gt;, and for the full process see &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;. The &lt;a href="/features/load-bearing-assumptions/"&gt;load-bearing assumptions feature&lt;/a&gt; shows how a report lays them out.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>The cheapest experiments to test a startup idea first</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/cheapest-experiments-to-test-a-startup-idea/"/>
    <id>https://www.startupintelligence.si/blog/cheapest-experiments-to-test-a-startup-idea/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>How to match a cheap experiment to your riskiest assumption, set a pass mark before you start, and act on the result instead of explaining it away.</summary>
    <content type="html">&lt;h2 id="why-start-cheap"&gt;Why start cheap&lt;/h2&gt;
&lt;p&gt;The first experiments on a new idea have one job: to find out quickly whether the assumptions holding up the plan are true. At that stage most of what you believe will change. Spending heavily before you know which beliefs survive means paying full price for answers you could have had for little.&lt;/p&gt;
&lt;p&gt;Cheap does not mean careless. A good cheap experiment is small, specific and decisive. It costs little, gives an answer in days or weeks, and the answer actually changes what you do next.&lt;/p&gt;
&lt;h2 id="match-the-experiment-to-the-assumption"&gt;Match the experiment to the assumption&lt;/h2&gt;
&lt;p&gt;Start from your list of load-bearing assumptions, not from a list of favourite tactics. For each one, ask: what is the smallest thing I could do that would show this is wrong? Some common pairings:&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;Is the problem real and frequent?&lt;/strong&gt; Structured conversations with a defined group, asking about the last time it happened and what it cost them, not about whether they like your idea.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Will they switch from what they use today?&lt;/strong&gt; A manual version of the service for a few customers, run by hand, at the price you intend to charge.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Will they pay this price?&lt;/strong&gt; A pre-order or deposit at a real price. Interest at zero cost tells you little about willingness to pay.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Can the channel reach them at a sensible cost?&lt;/strong&gt; A small, capped campaign in the channel you plan to rely on, measuring cost per qualified conversation.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Can you serve them at the cost you assumed?&lt;/strong&gt; Do the work yourself for a handful of customers and time it.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;Each of these tests one assumption. That is the point. An experiment that tests five assumptions at once usually cannot tell you which one failed.&lt;/p&gt;
&lt;h2 id="set-the-pass-mark-before-you-start"&gt;Set the pass mark before you start&lt;/h2&gt;
&lt;p&gt;Before running anything, write two results down: the result that would let the idea continue, and the result that would end it, or at least end this version of it. Tie the second to the kill threshold of the assumption you are testing. If you would keep going whatever happened, the experiment is not worth running.&lt;/p&gt;
&lt;p&gt;Take a made-up example. A founder plans weekly warehouse space for small online brands and believes many of them would move stock out of their current warehouse. A cheap test is to interview a defined group of brands that have left a warehouse provider in the last year and reconstruct why. The founder writes down, before the first call, how many of them would need to cite a reason the new service fixes for the assumption to survive.&lt;/p&gt;
&lt;h2 id="order-matters"&gt;Order matters&lt;/h2&gt;
&lt;p&gt;If you have several experiments, start with the one that is cheapest and most decisive: the one that could end the idea for the least money and time. There is no point paying for a channel test if conversations would show the problem is rare. Each result also sharpens the next experiment, so running them in order is cheaper than running them at once.&lt;/p&gt;
&lt;h2 id="what-to-do-with-the-result"&gt;What to do with the result&lt;/h2&gt;
&lt;p&gt;Read the result against the marks you wrote, not against your hopes. If it passes, move to the next assumption. If it fails, act on the decision you made in advance. If it lands in between, the experiment was probably too small; run a slightly larger version rather than reinterpreting the first.&lt;/p&gt;
&lt;h2 id="how-startup-intelligence-helps"&gt;How Startup Intelligence helps&lt;/h2&gt;
&lt;p&gt;A Startup Intelligence report lists three experiments aimed at the assumptions that carry your plan, cheapest and most decisive first. Each one describes the experiment, an estimated cost and number of days, the result that would be decisive and the result that would end the idea. The costs and durations are the engine&amp;#x27;s estimates, not quotes, so check them against your own situation.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Aim each experiment at one assumption that carries the plan, decide the pass mark and the kill result first, and start with the test that could end the idea for the least. To set those kill results, read &lt;a href="/blog/kill-thresholds-decide-when-to-stop/"&gt;kill thresholds&lt;/a&gt;, and for conversations that tell you more than compliments, read &lt;a href="/blog/how-to-get-honest-feedback-on-a-startup-idea/"&gt;how to get honest feedback on a startup idea&lt;/a&gt;. For founders, the &lt;a href="/use-cases/founders/"&gt;founders&amp;#x27; page&lt;/a&gt; shows how the report fits this work.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>How to check a startup forecast after the fact</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/how-to-check-a-startup-forecast-later/"/>
    <id>https://www.startupintelligence.si/blog/how-to-check-a-startup-forecast-later/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Most forecasts can never be checked because nobody wrote down a range, a date and the reasons. How to record yours, and what a report records for you.</summary>
    <content type="html">&lt;h2 id="the-deck-from-two-years-ago"&gt;The deck from two years ago&lt;/h2&gt;
&lt;p&gt;Picture a founder clearing out an old folder and finding the pitch deck from two years ago. This is a made-up example, but a familiar one. The deck said the business would have a certain number of paying customers by now. It has fewer. Was the forecast badly wrong, or roughly right with bad luck? Nobody can say.&lt;/p&gt;
&lt;p&gt;The deck gave one number and no range. It did not say which assumptions the number rested on, or how sure anyone was about them. And in the meantime the story has been retold many times: the market was slower, a competitor appeared, the price was changed. Every explanation is plausible. None can be checked, because the forecast was never written in a form that could be checked.&lt;/p&gt;
&lt;h2 id="why-most-forecasts-escape-being-checked"&gt;Why most forecasts escape being checked&lt;/h2&gt;
&lt;p&gt;Three habits make a forecast uncheckable.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;A single number.&lt;/strong&gt; &amp;quot;We will have 2,000 customers&amp;quot; is almost certainly wrong in detail, so it cannot tell you whether the thinking behind it was good.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;No date for the check.&lt;/strong&gt; Without a stated month, a forecast drifts. A year later it becomes &amp;quot;eventually&amp;quot;.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;No recorded reasons.&lt;/strong&gt; If the assumptions were not written down, you cannot tell which one failed, so you cannot learn from the miss.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;Each of these lets a forecast be reinterpreted after the fact instead of scored.&lt;/p&gt;
&lt;h2 id="write-it-down-before-you-know"&gt;Write it down before you know&lt;/h2&gt;
&lt;p&gt;The fix is simple and slightly uncomfortable. Before the outcome is known, write down:&lt;/p&gt;
&lt;ol&gt;&lt;li&gt;&lt;strong&gt;A range, not a point.&lt;/strong&gt; &amp;quot;Between 400 and 1,600 customers by month 24&amp;quot;, with the range as wide as your honest uncertainty.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The month from which it can be checked.&lt;/strong&gt;&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The assumptions it rests on&lt;/strong&gt;, each with its own range, and where each figure came from.&lt;/li&gt;&lt;/ol&gt;
&lt;p&gt;Then, when the month arrives, ask one question of each range: did the truth land inside it? Over many forecasts this tells you something valuable. If you say &amp;quot;I am fairly sure&amp;quot; and the truth keeps landing outside your ranges, your ranges are too narrow. If it always lands inside, they may be wider than they need to be. This is called calibration, and it can only be measured from forecasts recorded in advance.&lt;/p&gt;
&lt;h2 id="what-a-startup-intelligence-run-records"&gt;What a Startup Intelligence run records&lt;/h2&gt;
&lt;p&gt;Startup Intelligence takes the same view of its own output. The product&amp;#x27;s documentation puts its position bluntly: the verdict is never the product; the ordered queue of things that could kill the idea is. A run also keeps two records that make the engine itself checkable.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The ranges, recorded in advance.&lt;/strong&gt; At the time of the run, the engine writes down the range it stated for each assumption that carries the outcome, with its evidence rung, before anything is known. The report&amp;#x27;s last page lists these recorded claims with the month from which they can be scored, and says how they will be scored: by whether the truth lands inside the range that was stated.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The settings, recorded on the run.&lt;/strong&gt; The same page lists the seed, the number of simulated paths and the model settings the result came from, so the simulation can be re-run from the same inputs and give the same distribution. Someone who distrusts a result can check it instead of arguing with it.&lt;/p&gt;
&lt;h2 id="what-it-does-not-claim-yet"&gt;What it does not claim yet&lt;/h2&gt;
&lt;p&gt;This is the part we want to be most careful about. No run has yet reached the month when its ranges can be checked. So there is no record of how well the engine&amp;#x27;s ranges have matched reality, and the report does not pretend there is. Its own words on that page are that there is no calibration curve yet, and that inventing one would be a lie.&lt;/p&gt;
&lt;p&gt;Two more limits belong here. A simulation is a calculator: it computes the consequences of the assumptions it was given and does not check whether they are true. And today the engine consults no outside source during a run, so the recorded ranges rest on the AI model&amp;#x27;s own knowledge and are filed on the lowest rungs of the evidence ladder. A recorded range makes a forecast checkable; it does not make it right.&lt;/p&gt;
&lt;h2 id="start-your-own-ledger"&gt;Start your own ledger&lt;/h2&gt;
&lt;p&gt;You do not need any tool to begin. Keep one page per idea with a few columns: the claim, the range you believe today, where the figure came from, and the month you will check it. Seal it in the sense that matters: do not edit old rows. When a result arrives, add a new row instead. As results come in, you learn whether your sense of certainty can be trusted, which is worth more than any single forecast.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;A forecast earns trust only when it can be checked. Write down a range, a date and the reasons before the outcome is known, then score the range honestly when the date arrives. To decide in advance what result would make you stop, read &lt;a href="/blog/kill-thresholds-decide-when-to-stop/"&gt;kill thresholds&lt;/a&gt;, and for why the report shows ranges instead of a single figure, read &lt;a href="/blog/why-the-report-gives-no-score/"&gt;why the report gives no single score&lt;/a&gt;. The &lt;a href="/features/evidence-grades/"&gt;evidence grades feature&lt;/a&gt; explains the ladder behind every figure, and &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt; puts the steps in order.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>How to get honest feedback on a startup idea</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/how-to-get-honest-feedback-on-a-startup-idea/"/>
    <id>https://www.startupintelligence.si/blog/how-to-get-honest-feedback-on-a-startup-idea/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Friends say they love it; strangers say they would use it. Why kind answers carry little information, and how tests designed to break the idea tell you more.</summary>
    <content type="html">&lt;h2 id="everyone-liked-it"&gt;Everyone liked it&lt;/h2&gt;
&lt;p&gt;Picture a founder with an idea for a booking tool for small physiotherapy clinics. This is a made-up example. Over a month they describe it to everyone they can: friends, former colleagues, two clinic owners they know, people at a meetup. The response is warm. Most people say it is a great idea. Several say they would definitely use it.&lt;/p&gt;
&lt;p&gt;Three months after launch, almost none of them have signed up. Nobody lied. Each answer was sincere at the moment it was given. But the founder had collected a great deal of encouragement and almost no evidence.&lt;/p&gt;
&lt;h2 id="why-kind-answers-tell-you-so-little"&gt;Why kind answers tell you so little&lt;/h2&gt;
&lt;p&gt;There are a few reasons, and none of them is about people being dishonest.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;Saying yes is free.&lt;/strong&gt; Telling a friend their idea sounds good costs nothing and keeps the conversation pleasant.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Hypothetical questions get hypothetical answers.&lt;/strong&gt; &amp;quot;Would you use this?&amp;quot; asks people to predict their own future behaviour, which is hard to do accurately.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;You are asking about your solution, not their problem.&lt;/strong&gt; Once they have heard your pitch, people react to it rather than telling you how they deal with the problem today.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;Opinions still have a use: they reveal gaps in the plan and words people use for the problem. They are weak evidence that anyone will change what they do or pay you.&lt;/p&gt;
&lt;h2 id="try-to-break-the-idea-not-to-confirm-it"&gt;Try to break the idea, not to confirm it&lt;/h2&gt;
&lt;p&gt;The more useful stance is to stop collecting reasons the idea might work and start looking for the result that would show it does not. Confirmation never ends: there is always one more person who might say yes. A test aimed at breaking an assumption can end, because you decide in advance what failure looks like.&lt;/p&gt;
&lt;p&gt;That decision is the whole discipline. Before you run anything, write down the result that would end the idea, and put a number in it. &amp;quot;Fewer than 5 of 50 clinic owners we approach agree to a paid pilot&amp;quot; is a commitment you can be held to. &amp;quot;Clinic owners are not interested&amp;quot; is not, because once the data arrives you can always argue about what &amp;quot;interested&amp;quot; meant. The numbers here are made up; what matters is the form.&lt;/p&gt;
&lt;h2 id="questions-that-produce-honest-answers"&gt;Questions that produce honest answers&lt;/h2&gt;
&lt;p&gt;When you do talk to people, a few habits help.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;Ask about the last time.&lt;/strong&gt; &amp;quot;When did this last happen? What did you do? What did it cost you?&amp;quot; Past behaviour is far more reliable than predictions.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Ask what they use today&lt;/strong&gt;, including the unglamorous answers: a notebook, a spreadsheet, a receptionist&amp;#x27;s memory.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Ask for something small that costs them.&lt;/strong&gt; Time for a pilot, a deposit, a signed letter of intent, an introduction to the person who signs the cheques. A polite yes that turns into a no at this point is the most useful answer you will get.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Talk less than they do.&lt;/strong&gt; Describe the problem, not your product, and let the silence do some of the work.&lt;/li&gt;&lt;/ul&gt;
&lt;h2 id="how-startup-intelligence-designs-the-tests"&gt;How Startup Intelligence designs the tests&lt;/h2&gt;
&lt;p&gt;The report page &amp;quot;What to do on Monday&amp;quot; in a Startup Intelligence report is built on the same stance. The engine designs experiments meant to kill the assumptions that carry your plan, not to confirm them, and it prefers tests that need no product, facility or capital: conversations, public filings, a landing page, a spreadsheet.&lt;/p&gt;
&lt;p&gt;Every experiment must state the result that would end the idea, and that result must contain a number. A proposed test without one is rejected and never shown to you. Each experiment on the page lists its method, an estimated cost and number of days, the result that would be decisive, and the result that kills the idea.&lt;/p&gt;
&lt;p&gt;The order is computed, not guessed. The engine weighs how much each assumption carries, how little is known about it and how hard a wrong bet would be to reverse, against what the test costs and how many days it takes to answer. Only the top three are shown. The costs and durations are the engine&amp;#x27;s estimates, and each cost is filed on the lowest rung of the evidence ladder, so check them against your own situation.&lt;/p&gt;
&lt;p&gt;The page&amp;#x27;s note on what changes next is a plain instruction: run the first one and re-run the report. In its own words, the value is not in that document but in the difference between it and the next one.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Treat kind words as a map of what people noticed, not as evidence. Ask about past behaviour, ask for small commitments, and write the result that would end the idea, with a number, before you run anything. For how to choose and order those tests, read &lt;a href="/blog/cheapest-experiments-to-test-a-startup-idea/"&gt;the cheapest experiments to test a startup idea&lt;/a&gt;, and to fix your stop points first, read &lt;a href="/blog/kill-thresholds-decide-when-to-stop/"&gt;kill thresholds&lt;/a&gt;. The whole sequence is in &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;, and the &lt;a href="/features/cheapest-experiments/"&gt;cheapest experiments feature&lt;/a&gt; shows how a report lays the tests out.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>How to prepare your idea for an incubator interview</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/how-to-prepare-for-an-incubator-interview/"/>
    <id>https://www.startupintelligence.si/blog/how-to-prepare-for-an-incubator-interview/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>A short slot with a review panel goes further with five things: your claims, the assumptions behind them, your bar, your stop points and your tests.</summary>
    <content type="html">&lt;h2 id="fifteen-minutes-and-a-deck"&gt;Fifteen minutes and a deck&lt;/h2&gt;
&lt;p&gt;Picture a founder invited to interview for an incubation programme. This is a made-up example. They have a short slot with a panel, and their instinct is to spend the week before polishing the deck: a sharper logo, a bigger market figure, a slide of features.&lt;/p&gt;
&lt;p&gt;Every programme has its own process and criteria, and you should follow whatever it asks for. But a polished deck mostly shows what the idea does. In a short conversation, what tends to make an idea credible is showing that you know what it depends on, and how you would find out if you are wrong. That takes a different kind of preparation.&lt;/p&gt;
&lt;h2 id="bring-five-things"&gt;Bring five things&lt;/h2&gt;
&lt;h3 id="1-the-idea-as-claims-with-the-gaps-marked"&gt;1. The idea as claims, with the gaps marked&lt;/h3&gt;
&lt;p&gt;Write the idea as nine claims about the world: who the customer is, their job and pain, what they do today, the mechanism that makes your product better, the value created, the value captured, the path to the customer, the cost to serve and what persists over time.&lt;/p&gt;
&lt;p&gt;Some of these will be weak or empty. Say so. &amp;quot;We have not yet worked out how we reach the second hundred customers&amp;quot; is a more credible sentence than a channel plan nobody has tested. A panel can help with a gap you name; it cannot help with one you hide.&lt;/p&gt;
&lt;h3 id="2-the-few-assumptions-that-carry-the-plan"&gt;2. The few assumptions that carry the plan&lt;/h3&gt;
&lt;p&gt;From those claims, pick the five to eight assumptions the outcome depends on most. For each one, be ready to say where the figure came from: measured, borrowed from a comparable business, worked out from other figures, or your own belief. Being honest that most of them are still beliefs at this stage is normal. Pretending otherwise is what erodes trust.&lt;/p&gt;
&lt;h3 id="3-the-bar-you-are-aiming-for-and-what-counts"&gt;3. The bar you are aiming for, and what counts&lt;/h3&gt;
&lt;p&gt;State the outcome that would make the idea worth your time: a measure, an amount and a month, and exactly what counts towards it. &amp;quot;Contracted recurring revenue, excluding one-off setup fees&amp;quot; is clearer than &amp;quot;revenue&amp;quot;. Be ready to say why that bar matters to you.&lt;/p&gt;
&lt;h3 id="4-where-you-would-stop"&gt;4. Where you would stop&lt;/h3&gt;
&lt;p&gt;For the heaviest assumptions, know the value at which the plan stops working. If a panel member asks &amp;quot;what would make you give this up?&amp;quot;, a specific answer shows discipline. Our guide to &lt;a href="/blog/kill-thresholds-decide-when-to-stop/"&gt;kill thresholds&lt;/a&gt; explains how to work them out.&lt;/p&gt;
&lt;h3 id="5-the-experiments-with-the-failing-result-written-first"&gt;5. The experiments, with the failing result written first&lt;/h3&gt;
&lt;p&gt;List the experiments you have run and the ones you will run next. For each, say in advance the result that would end the idea, with a number in it. &amp;quot;Fewer than 5 of 50 shop owners we approach agree to a paid pilot&amp;quot; is a commitment; &amp;quot;see if people like it&amp;quot; is not. We explain why in &lt;a href="/blog/how-to-get-honest-feedback-on-a-startup-idea/"&gt;how to get honest feedback on a startup idea&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="questions-worth-being-ready-for"&gt;Questions worth being ready for&lt;/h2&gt;
&lt;p&gt;Whatever the panel asks, a few questions are worth preparing because the five items above answer them directly.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;Who exactly is the customer?&lt;/strong&gt; Your first claim.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Why now?&lt;/strong&gt; Name and date the change that makes the idea possible today. If you cannot, say so, and say what you will do to find out why nobody has done it. See &lt;a href="/blog/why-now-the-question-a-simulation-cannot-answer/"&gt;why now: the question a simulation cannot answer&lt;/a&gt;.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What has to be true?&lt;/strong&gt; Your load-bearing assumptions.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What would make you stop?&lt;/strong&gt; Your kill lines.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What will you do next, and what will it cost?&lt;/strong&gt; Your experiments.&lt;/li&gt;&lt;/ul&gt;
&lt;h2 id="if-you-bring-a-report"&gt;If you bring a report&lt;/h2&gt;
&lt;p&gt;Some founders prepare with a Startup Intelligence report. If you do, a few points help it land well.&lt;/p&gt;
&lt;p&gt;The report is built in layers, and its own &amp;quot;How to read this report&amp;quot; page suggests where to start: page one if a reader has two minutes; add &amp;quot;What actually moves the outcome&amp;quot; and &amp;quot;What to do on Monday&amp;quot; if they are deciding whether to commit money or months; the whole document for an analyst or an investor; and the last two pages for anyone who wants to check the engine rather than the idea. Page one states which next step the evidence supports, chosen from a fixed set, with a conviction level that describes the evidence rather than the idea; treat it as a summary of the pages behind it, not a verdict. The PDF is the same document as the web report, so you can bring it on paper.&lt;/p&gt;
&lt;p&gt;Present it for what it is. The report gives no single score, and its share of futures that clear your bar is a reading against a threshold you chose, not a rating of the idea. It is not investment advice and gives no valuation. Today its base figures come from the AI model&amp;#x27;s own knowledge and are filed on the lowest rungs of its evidence ladder, so be ready to say which of them you have since checked yourself.&lt;/p&gt;
&lt;h2 id="what-no-preparation-replaces"&gt;What no preparation replaces&lt;/h2&gt;
&lt;p&gt;A panel is also judging things no document shows: your commitment, how fast you learn, and your access to customers. Our post for programme teams, &lt;a href="/blog/incubators-reviewing-ideas-on-one-frame/"&gt;reviewing many ideas on one frame&lt;/a&gt;, makes the same point from their side. The best preparation shows those qualities through the work above: a founder who has named the gaps and already designed the tests to close them.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Walk in with the idea written as claims, the assumptions that carry it, your bar, your stop points and your next experiments with their failing results written down. It is less polished than a feature slide and far more convincing. To build these pieces step by step, start with &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;, or see how the report supports this on the &lt;a href="/use-cases/founders/"&gt;founders page&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>How to test a startup idea before you build it</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/how-to-test-a-startup-idea-before-you-build/"/>
    <id>https://www.startupintelligence.si/blog/how-to-test-a-startup-idea-before-you-build/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Start with assumptions, not features: find the few your idea rests on, grade the evidence, set a stop point and run the cheapest experiment first.</summary>
    <content type="html">&lt;h2 id="start-with-claims-not-features"&gt;Start with claims, not features&lt;/h2&gt;
&lt;p&gt;Picture a founder with an idea for weekly warehouse space rented by the pallet, aimed at small online brands. The first instinct is to design the booking screen. That feels like progress, and it is the part of the plan the founder controls. But the booking screen is not what decides whether the business works. Whether small brands really have seasonal peaks, whether they would move stock out of the warehouse they use today, whether the price covers the cost of the space: those decide it, and none of them is tested by building a screen.&lt;/p&gt;
&lt;p&gt;So the first step is to rewrite the idea as claims about the world. A useful frame has nine parts:&lt;/p&gt;
&lt;ol&gt;&lt;li&gt;&lt;strong&gt;The customer&lt;/strong&gt;, defined narrowly enough that you could count them and reach them.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Their job and pain&lt;/strong&gt;: what they are trying to get done and what it costs them today.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What they do now&lt;/strong&gt;, including spreadsheets, a cousin who helps, or nothing at all.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The mechanism&lt;/strong&gt;: why your product would cause a better outcome. Not the feature, the cause.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The value created&lt;/strong&gt;, in the customer&amp;#x27;s own units.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The value captured&lt;/strong&gt;: the price, and what share of the value it takes.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The path to the customer&lt;/strong&gt;: the channel that reaches them again and again, and its cost.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The cost to serve&lt;/strong&gt; one more customer, on top of the fixed base.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What persists&lt;/strong&gt;: what builds up that makes the business harder to copy next year.&lt;/li&gt;&lt;/ol&gt;
&lt;p&gt;If you cannot write two of these, especially the customer or the mechanism, that is your first finding. It is cheaper to learn it on paper than after a launch.&lt;/p&gt;
&lt;h2 id="find-the-assumptions-that-carry-the-weight"&gt;Find the assumptions that carry the weight&lt;/h2&gt;
&lt;p&gt;Every claim rests on assumptions, and an honest list will be long. They are not equally important. A few of them decide the outcome; the rest barely move it. The skill is to find those few.&lt;/p&gt;
&lt;p&gt;A simple way is to ask, for each assumption: if this were badly wrong, would the plan still clear the bar I care about? If the answer is no, the assumption is load-bearing. Then ask a second question: what do I actually know about it? An assumption that carries the plan and rests on nothing yet is where your attention belongs. We go deeper on this in &lt;a href="/blog/load-bearing-assumptions-in-a-startup-idea/"&gt;load-bearing assumptions&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="grade-the-evidence-honestly"&gt;Grade the evidence honestly&lt;/h2&gt;
&lt;p&gt;For each load-bearing assumption, write down where your number comes from. Did you measure it? Did you borrow it from a comparable business? Did you work it out from other figures? Or is it your own belief? These are very different kinds of evidence, and a plan built mostly on beliefs should be held more loosely than one built on measurements. A belief also deserves a wider range: if you are guessing a conversion rate, say &amp;quot;somewhere between one and six percent&amp;quot;, not &amp;quot;three percent&amp;quot;.&lt;/p&gt;
&lt;h2 id="decide-where-you-would-stop"&gt;Decide where you would stop&lt;/h2&gt;
&lt;p&gt;Before you test anything, write down the outcome the idea has to reach to be worth your time: a measure, an amount and a date, and a sentence on why that bar matters to you. Then, for each load-bearing assumption, work out the value at which the plan would no longer clear it. That is your stop point. Writing it down first is what keeps you honest when a result disappoints. Our guide to &lt;a href="/blog/kill-thresholds-decide-when-to-stop/"&gt;kill thresholds&lt;/a&gt; covers how.&lt;/p&gt;
&lt;h2 id="run-the-cheapest-experiment-first"&gt;Run the cheapest experiment first&lt;/h2&gt;
&lt;p&gt;Now choose experiments. The best first experiment is the cheapest one that could prove your riskiest assumption wrong. Often that is a handful of structured conversations, a small paid pilot run by hand, or a pre-order page with a real price. Each experiment should say in advance what result would end the idea and what result would let it continue. See &lt;a href="/blog/cheapest-experiments-to-test-a-startup-idea/"&gt;the cheapest experiments to run first&lt;/a&gt; for patterns.&lt;/p&gt;
&lt;h2 id="where-a-report-fits"&gt;Where a report fits&lt;/h2&gt;
&lt;p&gt;Startup Intelligence does this work in one pass. You describe the idea in plain words and set your bar. The engine breaks the idea into the nine claim links and refuses if too few are stated, naming the sentences that are missing. It then grades the key assumptions, simulates the range of outcomes against your threshold, and gives you stop points and the three cheapest experiments.&lt;/p&gt;
&lt;p&gt;Two limits are worth knowing. Today the base figures come from the AI model&amp;#x27;s own knowledge, and the report files them on the lowest rungs of its evidence ladder, so read them as assumptions to test, not as research. And the report gives no single score: it shows ranges and conditions, because nobody can rate an idea. Our About page lists the people who build it: Divya Prakash Jha (CEO), Vaibhav Jain (COO) and Zeeshan Khan (CMO).&lt;/p&gt;
&lt;h2 id="the-short-version"&gt;The short version&lt;/h2&gt;
&lt;p&gt;Write the idea as claims. Find the few that carry the weight. Be honest about what you know. Decide where you would stop. Then run the smallest experiment that could tell you that you are wrong. If you would like a report that does the first four steps with you, call the team to &lt;a href="/contact/"&gt;ask for access&lt;/a&gt;; see &lt;a href="/product/"&gt;what a report contains&lt;/a&gt; first.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>For incubators: reviewing many ideas on one frame</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/incubators-reviewing-ideas-on-one-frame/"/>
    <id>https://www.startupintelligence.si/blog/incubators-reviewing-ideas-on-one-frame/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Why a common frame helps incubators review startup ideas fairly, what to compare across ideas, and what no report can decide for a review panel.</summary>
    <content type="html">&lt;h2 id="the-review-day-problem"&gt;The review-day problem&lt;/h2&gt;
&lt;p&gt;An incubator&amp;#x27;s selection or review day follows a familiar pattern. A small panel sees many ideas in a short time. Each founder presents what they are proudest of. Each reviewer asks the questions they always ask. By the end, ideas have been compared on presentation skills, on the topics that happened to come up, and on a scoring sheet whose numbers are hard to explain a week later.&lt;/p&gt;
&lt;p&gt;None of this is anyone&amp;#x27;s fault. It is what happens when each idea arrives in its own shape.&lt;/p&gt;
&lt;h2 id="why-one-frame-helps"&gt;Why one frame helps&lt;/h2&gt;
&lt;p&gt;A common frame does not decide anything. It makes sure the same questions are asked of every idea, so the discussion is about the ideas rather than about the pitches. A good frame for early ideas has three properties.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;It covers the whole chain.&lt;/strong&gt; Who the customer is, their problem, what they do today, why the product would work, the value, the price, the path to customers, the cost to serve and what persists. Gaps become visible: an idea with a brilliant product and no stated path to customers looks different from one with both.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;It separates importance from evidence.&lt;/strong&gt; For each idea, which assumptions carry the plan, and how much is actually known about them? This is where reviewers can add the most value, by pointing at a heavy assumption the founder has not tested.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;It respects the founder&amp;#x27;s own goal.&lt;/strong&gt; A lifestyle business and a venture-scale company need different outcomes. Judging both against one house score is unfair to both. Judging each against the bar its founder set is more useful.&lt;/p&gt;
&lt;h2 id="what-to-compare-across-ideas"&gt;What to compare across ideas&lt;/h2&gt;
&lt;p&gt;With a common frame, a panel can compare ideas on things that matter:&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;how many of the nine claim links each idea states, and which are missing;&lt;/li&gt;&lt;li&gt;how many load-bearing assumptions are weakly evidenced;&lt;/li&gt;&lt;li&gt;how close each idea is to its own kill lines;&lt;/li&gt;&lt;li&gt;how cheap and decisive its next experiments are.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;These do not add up to a rank, and they should not. They are the material for a good conversation and for a programme&amp;#x27;s support: the incubator&amp;#x27;s help can be aimed at the weakest heavy assumption of each idea.&lt;/p&gt;
&lt;h2 id="what-a-report-cannot-decide"&gt;What a report cannot decide&lt;/h2&gt;
&lt;p&gt;A report reads only what the founder wrote. Commitment, learning speed and the team&amp;#x27;s access to customers are not in that text, and they are exactly what a review panel exists to judge. Nor can a report test a part of the plan the founder left out: those links show as empty.&lt;/p&gt;
&lt;h2 id="how-startup-intelligence-fits"&gt;How Startup Intelligence fits&lt;/h2&gt;
&lt;p&gt;Startup Intelligence produces the same report structure for every idea: the nine claim links, the load-bearing assumptions with their evidence rungs, the range of outcomes against the founder&amp;#x27;s own threshold, the kill lines and the three cheapest experiments. It gives no single score, so it will not rank a cohort for you.&lt;/p&gt;
&lt;p&gt;Some practical points. New accounts are by invitation during early access. There is no batch upload: each idea is its own run, and runs are handled one at a time. Founders&amp;#x27; ideas are confidential, so agree with your founders before you run them; runs are private to the account that created them and are stored for now on our server in the USA, with hosting in India planned before a full launch. Today the base figures come from the AI model&amp;#x27;s own knowledge and are labelled as the weakest kind of evidence.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Ask the same questions of every idea, judge each against its own founder&amp;#x27;s bar, and finish every review with cheap, specific experiments. Our &lt;a href="/use-cases/incubators/"&gt;incubators page&lt;/a&gt; describes how the report supports this, and &lt;a href="/blog/why-the-report-gives-no-score/"&gt;why the report gives no single score&lt;/a&gt; explains why it will not rank a cohort. For founders on the other side of the table, we wrote &lt;a href="/blog/how-to-prepare-for-an-incubator-interview/"&gt;how to prepare your idea for an incubator interview&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Kill thresholds: how to decide in advance when to stop</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/kill-thresholds-decide-when-to-stop/"/>
    <id>https://www.startupintelligence.si/blog/kill-thresholds-decide-when-to-stop/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>A kill threshold is the result that would make you stop. How to set one before you test, read results against it, and tell stopping apart from pivoting.</summary>
    <content type="html">&lt;h2 id="the-problem-with-deciding-later"&gt;The problem with deciding later&lt;/h2&gt;
&lt;p&gt;Most founders never decide what result would make them stop. They run an experiment, get a disappointing number, and then decide what it means. At that point the decision is made by the person who most wants the idea to work. The number is too small, so the sample must have been wrong; the timing was bad; next month will be different. Months pass.&lt;/p&gt;
&lt;p&gt;A kill threshold fixes this by moving the decision to before the data arrives, when you can still think clearly.&lt;/p&gt;
&lt;h2 id="step-1-set-the-bar-the-idea-has-to-clear"&gt;Step 1: set the bar the idea has to clear&lt;/h2&gt;
&lt;p&gt;Start with the outcome that would make the idea worth your time. State it in four parts:&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;a measure&lt;/strong&gt;, such as annual recurring revenue or paying customers;&lt;/li&gt;&lt;li&gt;&lt;strong&gt;an amount&lt;/strong&gt;;&lt;/li&gt;&lt;li&gt;&lt;strong&gt;a month&lt;/strong&gt; by which it has to be reached;&lt;/li&gt;&lt;li&gt;&lt;strong&gt;a sentence on why&lt;/strong&gt; that bar matters, for example that it is the point at which the business pays two founders.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;The sentence matters more than it looks. It turns an arbitrary target into a commitment you understand, and it is the thing you will come back to when you are tempted to move the line.&lt;/p&gt;
&lt;h2 id="step-2-work-out-a-stop-value-for-each-assumption"&gt;Step 2: work out a stop value for each assumption&lt;/h2&gt;
&lt;p&gt;Now go through your load-bearing assumptions one at a time. For each, ask: how bad would this value have to be before the plan no longer clears the bar? That value is the assumption&amp;#x27;s kill threshold.&lt;/p&gt;
&lt;p&gt;Take a made-up example. A weekly warehouse rental plan needs a certain share of small online brands to have a seasonal peak. If the share falls below some level, there are too few customers in the city to reach the bar by the chosen month, however well everything else goes. That level is the kill line for that assumption. Every load-bearing assumption has one.&lt;/p&gt;
&lt;h2 id="step-3-ask-how-likely-it-is-you-are-already-past-it"&gt;Step 3: ask how likely it is you are already past it&lt;/h2&gt;
&lt;p&gt;A kill line on its own is half the picture. The other half is how likely it is that the truth already sits on the wrong side of it, given what you know. As a rule of thumb, an assumption whose kill line sits well inside the range you think plausible deserves attention before one whose kill line sits at the far edge. That is a good way to decide which line to test first.&lt;/p&gt;
&lt;h2 id="step-4-decide-what-you-will-do-when-a-result-crosses-the-line"&gt;Step 4: decide what you will do when a result crosses the line&lt;/h2&gt;
&lt;p&gt;Write it down now. If the first experiment lands past the kill line, will you stop? Change the customer? Change the price? The point is not to forbid pivots. It is to make sure a pivot is a decision about a new plan, with its own threshold, and not a way of avoiding the result of the old one.&lt;/p&gt;
&lt;h2 id="reading-results-against-the-line"&gt;Reading results against the line&lt;/h2&gt;
&lt;p&gt;When a result comes in, compare it with the line you wrote, not with how you feel. If it is clearly on the safe side, move on to the next assumption. If it is clearly past the line, act on what you decided. If it is close, that usually means the experiment was too small to settle it, which is a reason for a slightly bigger test, not for a new story.&lt;/p&gt;
&lt;h2 id="how-startup-intelligence-shows-it"&gt;How Startup Intelligence shows it&lt;/h2&gt;
&lt;p&gt;When you start a Startup Intelligence run, you state the bar in the same spirit: a measure, an amount in rupees, the month by which it has to be reached, and a sentence on what counts, exactly, which is printed in the report word for word so nobody can read the metric two ways. Keep your sentence on why the bar matters in your own notes. The engine simulates the outcome across thousands of possible futures, shows the spread, and tells you how many of those futures clear your bar. Then, for each load-bearing assumption, it finds the value at which the plan stops clearing it and shows how likely the truth is to be on the wrong side already.&lt;/p&gt;
&lt;p&gt;Those figures are only as good as the assumptions behind them, and today the base figures come from the AI model&amp;#x27;s own knowledge, labelled as the weakest kind of evidence. The share of futures that clears your bar is a reading of your assumptions against the bar you chose, not a rating of the idea, not a chance that the business succeeds and not advice.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Decide where you would stop before you look. Turn your goal into a stop value for each assumption that carries the plan, and write down what you will do if a result crosses it. Then run the &lt;a href="/blog/cheapest-experiments-to-test-a-startup-idea/"&gt;cheapest experiment&lt;/a&gt; that could take you past a line. To see how a report lays this out, read about &lt;a href="/features/outcome-range-and-kill-thresholds/"&gt;outcome ranges and kill thresholds&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Load-bearing assumptions: the 5 to 8 your idea rests on</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/load-bearing-assumptions-in-a-startup-idea/"/>
    <id>https://www.startupintelligence.si/blog/load-bearing-assumptions-in-a-startup-idea/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>What a load-bearing assumption is, how to find the few that decide your startup's outcome, and how to write each one so it can be tested.</summary>
    <content type="html">&lt;h2 id="an-arch-has-two-stones-that-matter-more-than-the-rest"&gt;An arch has two stones that matter more than the rest&lt;/h2&gt;
&lt;p&gt;Take any stone out of the side of an arch and it may still stand. Take out the keystone and it falls. A business plan is built the same way. It rests on dozens of assumptions, but only a few of them hold the rest up. Those are the load-bearing assumptions, and finding them early is the most useful thing a founder can do before spending money.&lt;/p&gt;
&lt;h2 id="what-makes-an-assumption-load-bearing"&gt;What makes an assumption load-bearing&lt;/h2&gt;
&lt;p&gt;An assumption is load-bearing when the plan cannot reach its goal if the assumption turns out badly wrong. That makes it a question about your goal as much as about the assumption. If your bar is a certain amount of revenue by a certain month, then an assumption is load-bearing when a realistic bad value for it pushes the plan below that bar.&lt;/p&gt;
&lt;p&gt;This is why a load-bearing assumption is not the same as a risky one. Many assumptions are uncertain but barely matter: whether customers prefer a blue or a green button rarely sinks a business. And some assumptions matter enormously but are already well known: the rent on a warehouse you have a quote for. The ones to watch are both: heavy and uncertain.&lt;/p&gt;
&lt;h2 id="start-from-a-chain-of-claims"&gt;Start from a chain of claims&lt;/h2&gt;
&lt;p&gt;It is hard to find assumptions in a pitch deck, because a deck is organised around features. It is easier to start from a chain of claims about the world. A complete idea states who the customer is, the job they need done, what they do today, the mechanism that makes your product better, the value it creates, the price, how you reach the customer, what it costs to serve them and what builds up over time.&lt;/p&gt;
&lt;p&gt;Each link rests on assumptions. Take a made-up example: weekly warehouse space rented by the pallet for small online brands. The customer link assumes these brands have seasonal peaks. The current-alternative link assumes they are unhappy enough with their existing warehouse to move stock. The value-captured link assumes a weekly price covers the cost of the space. Written out, the list is long. That is fine. The next step shortens it.&lt;/p&gt;
&lt;h2 id="rank-by-weight-then-by-evidence"&gt;Rank by weight, then by evidence&lt;/h2&gt;
&lt;p&gt;For each assumption, ask two separate questions.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;How much does it carry?&lt;/strong&gt; If it moved from its expected value to a pessimistic one, how far would the outcome move? You can reason this through on paper, or you can simulate it: vary one assumption at a time and watch the result.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;How well is it evidenced?&lt;/strong&gt; Is your figure measured, borrowed from a comparable business, worked out from other figures, or simply believed? Be strict here. &amp;quot;Everyone says so&amp;quot; is a belief.&lt;/p&gt;
&lt;p&gt;Plot the answers. The assumptions that carry a lot and rest on little are the ones to test first. Those that carry a lot and are well evidenced you can monitor. Those that carry little you can set aside, while keeping a note that you considered them.&lt;/p&gt;
&lt;h2 id="write-each-one-so-it-can-be-tested"&gt;Write each one so it can be tested&lt;/h2&gt;
&lt;p&gt;&amp;quot;Brands will like it&amp;quot; cannot be tested. &amp;quot;At least 40 percent of the brands we speak to have a peak month that needs twice their usual space&amp;quot; can be. A testable assumption names a group, a measure and a value, and ideally a range rather than a single number. That made-up 40 percent is only an illustration; what matters is the form.&lt;/p&gt;
&lt;h2 id="how-startup-intelligence-does-it"&gt;How Startup Intelligence does it&lt;/h2&gt;
&lt;p&gt;A Startup Intelligence report follows the same path. It breaks your idea into nine claim links, refuses if fewer than four are stated or if the customer or the mechanism is missing, and writes down the assumptions behind each link. It then simulates your outcome and marks the five to eight assumptions that carry most of the uncertainty as load-bearing. Each one carries a grade for its evidence, and a chart places every assumption by how much it carries against how well it is evidenced, so the corner to test first is easy to see.&lt;/p&gt;
&lt;p&gt;Today the engine&amp;#x27;s base figures come from the AI model&amp;#x27;s own knowledge, and the report files them as the weakest kind of evidence. That is deliberate: the report shows you which assumptions matter, and is plain that most of them still need real evidence. That is where your experiments come in.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Do not try to test everything. Find the keystones: the few assumptions your goal cannot survive without, and rank them by how little you know. Then test those first. For the full process, read &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;; for the ones that are easy to miss, read &lt;a href="/blog/assumptions-that-look-safe-and-are-not/"&gt;assumptions that look safe and are not&lt;/a&gt;; or see &lt;a href="/product/"&gt;what a report contains&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Premortem, canvas or de-risking report: which to use when</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/premortem-canvas-or-de-risking-report/"/>
    <id>https://www.startupintelligence.si/blog/premortem-canvas-or-de-risking-report/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Five common ways to stress a startup idea compared: asking people, a spreadsheet, a premortem, a one-page canvas and a de-risking report. What each catches.</summary>
    <content type="html">&lt;h2 id="one-idea-one-free-weekend"&gt;One idea, one free weekend&lt;/h2&gt;
&lt;p&gt;Picture a founder with a new idea and a free weekend to pressure-test it before telling their employer they are leaving. This is a made-up situation, but the choice is real. There are several familiar ways to stress an idea, and they are not interchangeable. Each one is good at catching a certain kind of problem and blind to others.&lt;/p&gt;
&lt;p&gt;This guide compares five of them, including our own product, so you can choose the right one for the question in front of you. Where we describe where a method comes from, we cite a source. Where we say what a method catches or misses, that is our editorial reading, and we try to be fair to each.&lt;/p&gt;
&lt;h2 id="the-five-methods-at-a-glance"&gt;The five methods at a glance&lt;/h2&gt;
&lt;div class="table-wrap" role="region" tabindex="0" aria-label="Table: Method, What it is, What it is good at"&gt;&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th scope="col"&gt;Method&lt;/th&gt;&lt;th scope="col"&gt;What it is&lt;/th&gt;&lt;th scope="col"&gt;What it is good at&lt;/th&gt;&lt;th scope="col"&gt;What it tends to leave out&lt;/th&gt;&lt;th scope="col"&gt;Fits when&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;Asking people you trust&lt;/td&gt;&lt;td&gt;Describing the idea to friends, mentors or likely customers and listening&lt;/td&gt;&lt;td&gt;Fast, free, catches obvious gaps and missing context&lt;/td&gt;&lt;td&gt;Kind answers, opinions rather than commitments&lt;/td&gt;&lt;td&gt;You need a first sense check in an afternoon&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;A spreadsheet forecast&lt;/td&gt;&lt;td&gt;Revenue and cost lines over time&lt;/td&gt;&lt;td&gt;Shows the arithmetic consequences of your inputs&lt;/td&gt;&lt;td&gt;Where the inputs came from; single numbers instead of ranges&lt;/td&gt;&lt;td&gt;You already trust most of the inputs&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;A premortem&lt;/td&gt;&lt;td&gt;The team imagines the project has failed and lists why&lt;/td&gt;&lt;td&gt;Brings out risks people hesitate to say aloud&lt;/td&gt;&lt;td&gt;Quantities, stop points and an order of tests&lt;/td&gt;&lt;td&gt;A team is about to commit and needs dissent heard&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;A one-page canvas&lt;/td&gt;&lt;td&gt;The business model laid out in boxes on one page&lt;/td&gt;&lt;td&gt;Shows the whole model and its empty boxes at once&lt;/td&gt;&lt;td&gt;Which box carries the outcome; how good the evidence is&lt;/td&gt;&lt;td&gt;You are shaping or explaining the model&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;A de-risking report&lt;/td&gt;&lt;td&gt;Assumptions extracted, graded and simulated against your bar&lt;/td&gt;&lt;td&gt;Weighs assumptions, grades evidence, sets stop points, orders tests&lt;/td&gt;&lt;td&gt;Anything not in your description; today, outside data&lt;/td&gt;&lt;td&gt;You need to decide what to test first&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="asking-people-you-trust"&gt;Asking people you trust&lt;/h2&gt;
&lt;p&gt;This is where many founders start, and it is worth doing. A good mentor will spot a missing piece of the plan in minutes. The weakness is that most people are kind. Asked &amp;quot;would you use this?&amp;quot;, many say yes, because saying yes costs them nothing. Opinions are useful for finding gaps; they are weak evidence that anyone will pay. We cover how to get more honest answers in &lt;a href="/blog/how-to-get-honest-feedback-on-a-startup-idea/"&gt;how to get honest feedback on a startup idea&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="a-spreadsheet-forecast"&gt;A spreadsheet forecast&lt;/h2&gt;
&lt;p&gt;A spreadsheet is a calculator. It tells you what follows from your inputs, and it is very good at that. It does not tell you whether the inputs are true. A conversion rate typed into a cell looks exactly as solid as one you measured. Spreadsheets also invite single numbers, which hide how unsure you are. If you use one, write a range for every input and a note on where each number came from.&lt;/p&gt;
&lt;h2 id="a-premortem"&gt;A premortem&lt;/h2&gt;
&lt;p&gt;The premortem was described by Gary Klein in the Harvard Business Review in September 2007. A team imagines that the project has already failed and each person writes down the reasons why. Because the failure is assumed, people who have doubts find it easier to voice them. It is quick, cheap and good at surfacing risks a confident team would otherwise talk past.&lt;/p&gt;
&lt;p&gt;Our reading of its limits: a premortem produces a list of possible failure reasons, but not how much each one matters, how likely it is, or at what point you should stop. It also draws only on what the people in the room know.&lt;/p&gt;
&lt;h2 id="a-one-page-canvas"&gt;A one-page canvas&lt;/h2&gt;
&lt;p&gt;Lean Canvas, created by Ash Maurya as a startup-focused adaptation of the Business Model Canvas, lays a business model out in boxes on a single page. Seeing the whole model at once makes empty boxes hard to ignore, and it is an excellent way to shape an idea and explain it to someone else.&lt;/p&gt;
&lt;p&gt;Our reading of its limits: every box looks equally important on the page. A canvas does not tell you which box the outcome depends on most, or how strong the evidence is behind what you wrote in it.&lt;/p&gt;
&lt;h2 id="a-de-risking-report"&gt;A de-risking report&lt;/h2&gt;
&lt;p&gt;This is what Startup Intelligence makes. You describe the idea in plain words and set the bar it has to clear. The engine breaks the idea into nine claim links and refuses if too few are stated. It extracts the assumptions behind the links, grades the evidence for each one on a seven-rung ladder, simulates thousands of possible futures against your bar, and marks the five to eight assumptions that move the outcome most. For each of those it shows the value at which the plan stops working, and it ends with three experiments, ordered by how much uncertainty each one settles for its cost and time.&lt;/p&gt;
&lt;p&gt;Its limits, stated plainly. The engine reads only what you wrote, so it cannot judge your team or test a part of the plan you left out. Today it consults no outside source during a run, so its base figures come from the AI model&amp;#x27;s own knowledge and are filed on the lowest rungs of the ladder. It is not market research, not a five-year model and not investment advice. New accounts are by invitation during early access, and no price is set yet.&lt;/p&gt;
&lt;h2 id="they-work-better-together"&gt;They work better together&lt;/h2&gt;
&lt;p&gt;These methods are not rivals. They are good at different stages of the same work.&lt;/p&gt;
&lt;p&gt;A canvas is a natural first step: it gets the model onto one page. A premortem with your co-founders then produces the failure reasons nobody had written down. Those reasons are raw material: each one is really an assumption that might be false. A de-risking report then weighs them, tells you which few carry the outcome and orders the experiments. And the conversations with people you trust become far more useful once you know which one assumption you are trying to break.&lt;/p&gt;
&lt;h2 id="how-to-choose"&gt;How to choose&lt;/h2&gt;
&lt;ul&gt;&lt;li&gt;If you have an afternoon and no plan on paper yet, &lt;strong&gt;talk to people and draw a canvas&lt;/strong&gt;.&lt;/li&gt;&lt;li&gt;If your team is about to commit money or months, &lt;strong&gt;run a premortem&lt;/strong&gt; first, so doubts are heard while they are cheap.&lt;/li&gt;&lt;li&gt;If you have a written plan and want to know &lt;strong&gt;what to test first and when to stop&lt;/strong&gt;, a de-risking report is built for that question.&lt;/li&gt;&lt;li&gt;Whatever you choose, keep a range and a source note beside every number.&lt;/li&gt;&lt;/ul&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;No single method catches everything. Use the fast ones to find gaps, the premortem to hear doubts, and a de-risking report to weigh the assumptions and decide where to start. For the step-by-step version, read &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;, or see &lt;a href="/product/"&gt;what a Startup Intelligence report contains&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Startup idea validator tools: six things to check first</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/startup-idea-validator-tools-what-to-check/"/>
    <id>https://www.startupintelligence.si/blog/startup-idea-validator-tools-what-to-check/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Six questions to ask any startup idea validator before you trust its report, and our own honest answers for Startup Intelligence, limits included.</summary>
    <content type="html">&lt;h2 id="a-crowded-category"&gt;A crowded category&lt;/h2&gt;
&lt;p&gt;Search for a way to check a startup idea and you will find many tools that promise to do it in a few clicks. Some are thoughtful; some turn a sentence into a polished report full of numbers. From the outside they look alike. This guide gives six questions that separate a report you can act on from one that only looks finished. We answer each one for our own product too, including where the answer is not yet what we want.&lt;/p&gt;
&lt;p&gt;We do not name or rate other tools here. The questions work on any of them.&lt;/p&gt;
&lt;h2 id="1-does-it-show-where-every-number-comes-from"&gt;1. Does it show where every number comes from?&lt;/h2&gt;
&lt;p&gt;A report about a future business is full of estimates. The question is whether each one says where it came from: measured, borrowed from a comparable business, worked out from other figures, or simply believed. If a report cannot tell you, you cannot tell which conclusions are solid.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; Every figure in a Startup Intelligence report carries a rung on a seven-rung evidence ladder and a note on its source, and weak evidence widens the range around it. Today no outside source is consulted during a run, so the base figures come from the AI model&amp;#x27;s own knowledge and sit on the lowest rungs. The report says so. Linking assumptions to published data you can open and check is planned.&lt;/p&gt;
&lt;h2 id="2-does-it-give-a-single-score"&gt;2. Does it give a single score?&lt;/h2&gt;
&lt;p&gt;A score is convenient and almost always misleading. It hides the reasons, the evidence and your own goals behind one number.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; No single score. The report shows ranges, evidence rungs, stop points and experiments. We explain why in &lt;a href="/blog/why-the-report-gives-no-score/"&gt;why our report gives no single score&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="3-what-does-it-do-when-your-idea-is-vague"&gt;3. What does it do when your idea is vague?&lt;/h2&gt;
&lt;p&gt;Give the tool one vague line. If it still produces a full report, it filled the gaps itself.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; If fewer than four of the nine claim links are stated, or the customer or the mechanism is missing, the engine refuses and names the sentences that would fill each gap. A refused run is not counted against you.&lt;/p&gt;
&lt;h2 id="4-does-it-tell-you-what-to-do-next"&gt;4. Does it tell you what to do next?&lt;/h2&gt;
&lt;p&gt;A report that ends with a verdict leaves you where you started. A useful one ends with specific, cheap actions aimed at the riskiest assumptions.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; Every report lists the three cheapest experiments for the riskiest assumptions, with cost, time and the result that would settle each one. The costs are estimates, not quotes.&lt;/p&gt;
&lt;h2 id="5-what-happens-to-your-idea-after-you-submit-it"&gt;5. What happens to your idea after you submit it?&lt;/h2&gt;
&lt;p&gt;Your idea is confidential. Ask where it is stored, who can see it, and whether you can delete it.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; Your idea goes only into the signed-in app. Runs are private to your account. For now they are stored on our server in the USA, and we plan to host the product in India (Navi Mumbai) before a full launch. There is no self-service deletion or export yet; contact us if you need either. More in &lt;a href="/blog/where-your-startup-idea-is-stored/"&gt;where your startup idea is stored&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="6-is-it-clear-about-what-it-cannot-do"&gt;6. Is it clear about what it cannot do?&lt;/h2&gt;
&lt;p&gt;Every tool has limits. A trustworthy one states them.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our answer.&lt;/strong&gt; The report is not market research and not a five-year financial model today; whether to add those is still being decided. It is not investment advice and gives no valuation. Report delivery by email is coded, but no mail service is connected on our server yet. New accounts are by invitation, and no price is set.&lt;/p&gt;
&lt;h2 id="how-to-use-these-questions"&gt;How to use these questions&lt;/h2&gt;
&lt;p&gt;Run the same idea through any tool you are considering and read the report with these six questions beside it. The answers will tell you more than a feature list.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Trust a report that shows its evidence, refuses to invent, tells you what to do next and is honest about its limits. To see how ours handles each point, read &lt;a href="/product/"&gt;what a Startup Intelligence report contains&lt;/a&gt;, check the &lt;a href="/faq/"&gt;FAQ&lt;/a&gt;, or see &lt;a href="/pricing/"&gt;pricing&lt;/a&gt;. To compare a report with methods that need no tool at all, read &lt;a href="/blog/premortem-canvas-or-de-risking-report/"&gt;premortem, canvas or de-risking report&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>What happens in the few minutes a report is being built</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/what-happens-during-a-run/"/>
    <id>https://www.startupintelligence.si/blog/what-happens-during-a-run/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>The nine stages of a Startup Intelligence run, from your idea to the finished report: what each stage does, why runs go one at a time, and what you can watch.</summary>
    <content type="html">&lt;h2 id="from-a-description-to-a-report"&gt;From a description to a report&lt;/h2&gt;
&lt;p&gt;You write your idea in plain words and set the outcome it has to reach. A few minutes later you have a report. In between, the engine works through nine stages, and the run&amp;#x27;s own page shows each one as it completes. Here is what happens, in order.&lt;/p&gt;
&lt;h2 id="the-nine-stages"&gt;The nine stages&lt;/h2&gt;
&lt;h3 id="1-decompose"&gt;1. Decompose&lt;/h3&gt;
&lt;p&gt;The idea is broken into nine claim links: the customer, their job and pain, what they do today, the mechanism, the value created, the value captured, the path to the customer, the cost to serve and what persists. If fewer than four links are stated, or the customer or the mechanism is missing, the run stops here and says which sentences would fill the gaps. Stopping at this stage means the expensive work never starts, and the refused run is not counted against you.&lt;/p&gt;
&lt;h3 id="2-extract"&gt;2. Extract&lt;/h3&gt;
&lt;p&gt;For each link, the assumptions it depends on are written down one by one.&lt;/p&gt;
&lt;h3 id="3-type"&gt;3. Type&lt;/h3&gt;
&lt;p&gt;Each assumption is placed by two things: how much the outcome leans on it and how well it is evidenced. The heavy, weakly evidenced ones form the &amp;quot;kill zone&amp;quot; that gets attention first. The placement is provisional: after the simulation has measured what actually moves the outcome, every assumption is placed again.&lt;/p&gt;
&lt;h3 id="4-price"&gt;4. Price&lt;/h3&gt;
&lt;p&gt;Each assumption gets a range and a rung on the seven-rung evidence ladder, from a measured figure to an unsourced belief. A weakly evidenced figure gets a wider range. Today no outside source is consulted during a run, so the base figures come from the AI model&amp;#x27;s own knowledge and are filed on the lowest rungs. Where no figure can be given, the gap is marked for you to fill rather than invented.&lt;/p&gt;
&lt;h3 id="5-assemble"&gt;5. Assemble&lt;/h3&gt;
&lt;p&gt;The assumptions are joined into one model of the outcome you asked about, working back from your threshold to the quantities that would have to hold.&lt;/p&gt;
&lt;h3 id="6-propagate"&gt;6. Propagate&lt;/h3&gt;
&lt;p&gt;The model is simulated across thousands of possible futures, each drawing its assumptions from their ranges. This produces the spread of outcomes and the share of futures that clear your bar.&lt;/p&gt;
&lt;h3 id="7-kill-lines"&gt;7. Kill lines&lt;/h3&gt;
&lt;p&gt;For each assumption, the engine finds the value at which the plan stops clearing your threshold, and how likely it is that the truth is already on the wrong side.&lt;/p&gt;
&lt;h3 id="8-structural-reading"&gt;8. Structural reading&lt;/h3&gt;
&lt;p&gt;A second reading asks four questions that a simulation cannot answer: why now, what compounds, what is melting, and whether the downside is bounded while the upside is open. It is asked to judge the structure on its own terms, without regard to whether the numbers clear your bar, so it is allowed to disagree with the arithmetic. When the two disagree, the report says so and treats the disagreement as a finding. We explain this stage in &lt;a href="/blog/why-now-the-question-a-simulation-cannot-answer/"&gt;why now: the question a simulation cannot answer&lt;/a&gt;.&lt;/p&gt;
&lt;h3 id="9-emit"&gt;9. Emit&lt;/h3&gt;
&lt;p&gt;The last stage designs the experiments that could end the idea fastest and keeps the top three, each with the result that would kill it stated as a number. It also writes a few adjacent versions of the idea, each aimed at a specific weak assumption, and records the run&amp;#x27;s stated ranges so they can be checked later. The report is then put together page by page, with every figure carrying its rung, on the web and as a PDF.&lt;/p&gt;
&lt;h2 id="how-long-it-takes-and-why-one-at-a-time"&gt;How long it takes, and why one at a time&lt;/h2&gt;
&lt;p&gt;A full run usually takes a few minutes. Runs are handled one at a time on our server, so if several people start runs together, a queue can make yours take longer. The run page shows progress while you wait, and the same address opens the finished report.&lt;/p&gt;
&lt;h2 id="what-you-can-watch"&gt;What you can watch&lt;/h2&gt;
&lt;p&gt;While a run works, its page shows the stages completing and fills in the findings established so far. When it finishes, the report opens at the same address. If the run was refused, the page shows the refusal with the missing links instead. Email delivery is coded, but no mail service is connected on our server yet, so the page and your runs list are the reliable places to find a report today.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;A run is a sequence of small, checkable steps: decompose, extract, type, price, assemble, propagate, find the kill lines, read the structure, write the report. The first step can refuse; every later step keeps its evidence visible. See the &lt;a href="/how-it-works/#walkthrough"&gt;walkthrough on the how-it-works page&lt;/a&gt; or read &lt;a href="/blog/when-an-idea-is-too-thin-to-test/"&gt;when the engine refuses&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>When the engine refuses: what makes an idea too thin to test</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/when-an-idea-is-too-thin-to-test/"/>
    <id>https://www.startupintelligence.si/blog/when-an-idea-is-too-thin-to-test/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>Why Startup Intelligence refuses ideas with too little detail, the minimum it needs, how to add it, and why a refusal beats a confident guess.</summary>
    <content type="html">&lt;h2 id="the-temptation-to-fill-the-gaps"&gt;The temptation to fill the gaps&lt;/h2&gt;
&lt;p&gt;Imagine typing a single line into an idea tool: &amp;quot;an app that helps restaurants reduce waste&amp;quot;. A tool that is eager to please can turn that into a market size, a revenue forecast and a list of competitors. The output looks thorough. But almost every number in it was supplied by the tool, not by you, and nothing on the page tells you which parts are yours and which were invented.&lt;/p&gt;
&lt;p&gt;Startup Intelligence takes the opposite position. If an idea does not contain enough to test, it refuses, and tells you exactly what is missing.&lt;/p&gt;
&lt;h2 id="what-the-engine-needs"&gt;What the engine needs&lt;/h2&gt;
&lt;p&gt;Every run starts by turning the idea into nine claim links: the customer; their job and pain; what they do today; the mechanism that makes your product better; the value created; the value captured; the path to the customer; the cost to serve; and what persists over time.&lt;/p&gt;
&lt;p&gt;Two rules decide whether the run continues.&lt;/p&gt;
&lt;ol&gt;&lt;li&gt;&lt;strong&gt;At least four of the nine links must be stated&lt;/strong&gt; in what you wrote.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Two links are always required&lt;/strong&gt;: who the customer is, and the mechanism. Without a customer there is nobody to be wrong about. Without a mechanism there is no reason for anything to change. Every other gap can become a finding; these two cannot.&lt;/li&gt;&lt;/ol&gt;
&lt;p&gt;You can try the rule on our &lt;a href="/#chain"&gt;home page&lt;/a&gt; by ticking the links an idea states. Nothing you tick is stored or sent.&lt;/p&gt;
&lt;h2 id="what-a-refusal-looks-like"&gt;What a refusal looks like&lt;/h2&gt;
&lt;p&gt;A refusal is a designed result, not an error. It states which links could not be filled and which were filled only by implication. For each empty link it offers a sentence pattern to complete, such as who exactly would use the product and what they do instead today, or why they would switch and what the current way costs them. It stops before the expensive stages run, and the screen says the run has not been counted against you.&lt;/p&gt;
&lt;h2 id="how-to-add-the-missing-detail"&gt;How to add the missing detail&lt;/h2&gt;
&lt;p&gt;The missing detail is often only a few sentences. Some patterns that help:&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;The customer:&lt;/strong&gt; name a group narrow enough to count. &amp;quot;Owners of single-location restaurants in Pune with more than forty covers&amp;quot; is testable; &amp;quot;restaurants&amp;quot; is not.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The mechanism:&lt;/strong&gt; say why your product causes a better outcome. &amp;quot;It forecasts tomorrow&amp;#x27;s covers from bookings so the kitchen preps less&amp;quot; is a mechanism; &amp;quot;AI-powered waste reduction&amp;quot; is a label.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;What they do today:&lt;/strong&gt; include the unglamorous answers: a notebook, the chef&amp;#x27;s memory, nothing at all.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The price and the path:&lt;/strong&gt; what you would charge, and how you would reach the next hundred customers.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;These examples are made up to show the form, not facts about any market.&lt;/p&gt;
&lt;h2 id="why-a-refusal-is-better-than-a-guess"&gt;Why a refusal is better than a guess&lt;/h2&gt;
&lt;p&gt;If a tool filled in the missing links itself, its report would give you confidence about a plan you have not written yet. A refusal gives you something more useful: a precise list of the parts of your plan that are still blank. Writing those sentences can change the idea before any analysis begins.&lt;/p&gt;
&lt;h2 id="limits"&gt;Limits&lt;/h2&gt;
&lt;p&gt;The check looks at whether a link is stated, not whether it is true. A clearly written but wrong claim passes this stage and is tested by the later ones, where its assumptions are graded and simulated.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Before you ask any tool to judge an idea, make sure the idea says who it is for and why it would work. If you are not sure, the nine links are a good checklist. For what happens after a run is accepted, read &lt;a href="/blog/what-happens-during-a-run/"&gt;what happens during a run&lt;/a&gt;, or see the &lt;a href="/features/honest-refusal/"&gt;honest refusal feature&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Where your startup idea is stored, and who can see it</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/where-your-startup-idea-is-stored/"/>
    <id>https://www.startupintelligence.si/blog/where-your-startup-idea-is-stored/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>What you submit to Startup Intelligence, where it is stored today, who can see it, the planned move to India, and what to ask us about your data.</summary>
    <content type="html">&lt;h2 id="a-fair-question-to-ask-first"&gt;A fair question to ask first&lt;/h2&gt;
&lt;p&gt;A startup idea is often the most valuable thing a founder has in the early months, and it is natural to hesitate before typing it into any tool. The right response is not reassurance but facts: what you submit, where it goes, who can see it and what you can do about it. Here they are for Startup Intelligence, including the parts that are not yet where we want them.&lt;/p&gt;
&lt;h2 id="what-you-submit"&gt;What you submit&lt;/h2&gt;
&lt;p&gt;To start a run you write the idea in plain words and set a threshold: a measure, an amount, a month and a sentence on what counts, exactly. That happens inside the signed-in app. The marketing website you are reading never asks for your idea: its contact page shows a phone number and asks you not to describe your idea in a first message.&lt;/p&gt;
&lt;h2 id="where-it-is-stored-today"&gt;Where it is stored today&lt;/h2&gt;
&lt;p&gt;For now, runs and reports are stored on our server in the USA. We plan to host the product in India (Navi Mumbai) before a full launch. We say &amp;quot;for now&amp;quot; deliberately: the current location is a fact of early access, not the intended end state.&lt;/p&gt;
&lt;h2 id="who-can-see-it"&gt;Who can see it&lt;/h2&gt;
&lt;p&gt;Runs and reports belong to your account. The app shows a run only to the account it belongs to, and there is no public link to share a report. New accounts are by invitation during early access, so the people using the product today are people we have invited.&lt;/p&gt;
&lt;h2 id="what-the-engine-does-with-it"&gt;What the engine does with it&lt;/h2&gt;
&lt;p&gt;The engine uses your description to build the report: it breaks the idea into claim links, extracts and grades the assumptions, simulates the outcome and writes the report. That work uses AI models run by outside providers, called from our server. No outside data source is consulted during a run today, so your idea is not sent to search services as part of the evidence step.&lt;/p&gt;
&lt;h2 id="what-you-cannot-do-yet"&gt;What you cannot do yet&lt;/h2&gt;
&lt;p&gt;There is no self-service way to delete your account or export your runs yet. If you need either, contact us. Email for verification and report delivery is coded, but no mail service is connected on our server yet, so do not expect a copy of your report by email.&lt;/p&gt;
&lt;h2 id="practical-advice"&gt;Practical advice&lt;/h2&gt;
&lt;p&gt;Whatever tool you use, a few habits help:&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;Describe the plan, not secrets. The engine needs the customer, the mechanism, the price and the path to customers; it does not need passwords, customer lists or unpublished partner terms.&lt;/li&gt;&lt;li&gt;If the idea belongs to a client or a founder you are advising, get their agreement before you run it.&lt;/li&gt;&lt;li&gt;Ask any provider where data is stored and how to delete it. If they cannot answer plainly, treat that as an answer.&lt;/li&gt;&lt;/ul&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Your idea goes into the signed-in app only, stays private to your account and, for now, sits on our server in the USA, with hosting in India planned. Deletion and export are not self-service yet. For more, read the &lt;a href="/faq/"&gt;FAQ&lt;/a&gt;, the &lt;a href="/privacy/"&gt;website privacy facts&lt;/a&gt; or &lt;a href="/blog/startup-idea-validator-tools-what-to-check/"&gt;six things to check in any idea tool&lt;/a&gt;. Questions go to the team through the &lt;a href="/contact/"&gt;contact page&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Why now? The question a simulation cannot answer</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/why-now-the-question-a-simulation-cannot-answer/"/>
    <id>https://www.startupintelligence.si/blog/why-now-the-question-a-simulation-cannot-answer/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>If your idea is this good, why has nobody done it? How to answer why now, what compounds, what is melting and whether the downside is bounded.</summary>
    <content type="html">&lt;h2 id="a-model-that-clears-the-bar"&gt;A model that clears the bar&lt;/h2&gt;
&lt;p&gt;Picture a founder with a plan for pay-per-use storage lockers in apartment buildings. This is a made-up example. They have built a careful model: residents per building, share who rent a locker, price per month, cost per installation. Run against their target, most of the futures they can imagine clear it comfortably. The numbers say go.&lt;/p&gt;
&lt;p&gt;Then a friend asks a simple question. Lockers are not new, apartment buildings are not new, and paying monthly for storage is not new. If this works so well, why is nobody doing it already?&lt;/p&gt;
&lt;p&gt;The founder has no answer. Nothing in the model could have produced one, because a model of outcomes takes the idea&amp;#x27;s structure as given. It computes what follows from the assumptions. It cannot ask whether the world is ready.&lt;/p&gt;
&lt;h2 id="what-the-arithmetic-cannot-see"&gt;What the arithmetic cannot see&lt;/h2&gt;
&lt;p&gt;Our own report puts it plainly on its last page: a simulation is a calculator. It computes the consequences of the assumptions it was given and does not check whether they are true. Four questions sit outside any such calculation, and each of them can sink a plan whose numbers look fine.&lt;/p&gt;
&lt;h3 id="1-why-now"&gt;1. Why now?&lt;/h3&gt;
&lt;p&gt;Name the specific change that made this possible recently: a cost that fell, a regulation that changed, a habit that shifted, a new capability. Then date it. &amp;quot;Growing digitisation&amp;quot; is not an answer; it is the absence of one dressed up as an answer.&lt;/p&gt;
&lt;p&gt;If you cannot name and date a change, there is an uncomfortable but useful reading: the idea has been possible for some time, and the market has not taken it up. That silence is evidence. It may have a good explanation, but it has to be explained, not dismissed.&lt;/p&gt;
&lt;h3 id="2-what-compounds"&gt;2. What compounds?&lt;/h3&gt;
&lt;p&gt;Which advantage grows over time if the business works? Common candidates are scale economics, network effects, switching costs, brand, a process others cannot easily copy, a scarce resource, or counter-positioning: a model an established player cannot copy without hurting its own business. If the honest answer is &amp;quot;nothing&amp;quot;, expect competitors to arrive once you prove there is demand, and expect margins to be squeezed toward zero.&lt;/p&gt;
&lt;h3 id="3-what-is-melting"&gt;3. What is melting?&lt;/h3&gt;
&lt;p&gt;Which parts of your offering are turning from bespoke into commodity, and how fast? If the thing you sell will be cheap or free in two years, that is a timing problem, not a strategy problem, and the difference matters for what you build first.&lt;/p&gt;
&lt;h3 id="4-is-the-shape-asymmetric"&gt;4. Is the shape asymmetric?&lt;/h3&gt;
&lt;p&gt;Is the downside bounded and the upside open? An idea that can lose only a known, limited amount but could pay off far more than expected can be worth acting on even when the odds look poor. A bet with a large downside needs far better odds. Ask how much you can lose, and whether the good outcomes are capped.&lt;/p&gt;
&lt;h2 id="the-lockers-again"&gt;The lockers again&lt;/h2&gt;
&lt;p&gt;Back to the made-up lockers. Working through the four questions changes the conversation. Why now: the founder cannot date a change, so the silence becomes the first thing to investigate, perhaps by asking building managers whether they have been offered this before and why they said no. What compounds: perhaps contracts with building managers, if they are exclusive. What is melting: little, since lockers are hardware. The shape: the downside per building is bounded by the installation cost. None of this was in the model, and all of it changes what to test first.&lt;/p&gt;
&lt;h2 id="how-startup-intelligence-asks-these-questions"&gt;How Startup Intelligence asks these questions&lt;/h2&gt;
&lt;p&gt;Every Startup Intelligence run includes a structural reading built around exactly these four questions. The engine asks for it separately from the simulation and tells the model to judge the structure on its own terms, without regard to whether the numbers clear your bar, so that it is free to disagree with them.&lt;/p&gt;
&lt;p&gt;Two rules keep it honest. If no enabling change can be dated, the reading has to say plainly that the idea has been available for some time and that the market&amp;#x27;s non-adoption is itself evidence; the prompt names &amp;quot;the timing is favourable&amp;quot; as the kind of answer that will be rejected, and if the model&amp;#x27;s text leaves the plain sentence out, the engine adds it. And if nothing compounds, the margin is modelled as decaying toward zero rather than holding, and the summary has to say so. The run page shows &amp;quot;No enabling change could be dated&amp;quot; when that is the case.&lt;/p&gt;
&lt;p&gt;When the structural reading and the simulation point different ways, the report does not average them. The page &amp;quot;What this comes to&amp;quot; leads with a callout that the two readings disagree and that this is the finding, and neither is overruled.&lt;/p&gt;
&lt;h2 id="the-limits-of-this-reading"&gt;The limits of this reading&lt;/h2&gt;
&lt;p&gt;The structural reading is the AI model&amp;#x27;s judgement on what you wrote. The engine has no search tool during a run, and the run page marks each of the four answers as having no source. Treat its answers as questions to check with real people and real documents, not as conclusions about your market.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Before trusting a model that clears your bar, answer the four questions it cannot: why now, what compounds, what is melting, and whether the downside is bounded. If you cannot date the change that makes your idea possible now, find out why the market has stayed silent. To see how this fits into a run, read &lt;a href="/blog/what-happens-during-a-run/"&gt;what happens during a run&lt;/a&gt;, and to find the assumptions behind your answers, read &lt;a href="/blog/assumptions-that-look-safe-and-are-not/"&gt;assumptions that look safe and are not&lt;/a&gt;. For the whole process, start with &lt;a href="/blog/how-to-test-a-startup-idea-before-you-build/"&gt;how to test a startup idea before you build it&lt;/a&gt;; the &lt;a href="/how-it-works/"&gt;how it works&lt;/a&gt; page shows all nine stages.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Why our startup idea report gives no single score</title>
    <link rel="alternate" type="text/html" href="https://www.startupintelligence.si/blog/why-the-report-gives-no-score/"/>
    <id>https://www.startupintelligence.si/blog/why-the-report-gives-no-score/</id>
    <published>2026-10-05T00:00:00Z</published>
    <updated>2026-10-05T00:00:00Z</updated>
    <author><name>The Startup Intelligence team</name></author>
    <summary>A single score hides what an idea rests on and invites false comparisons. What Startup Intelligence shows instead: ranges, evidence rungs and stop points.</summary>
    <content type="html">&lt;h2 id="the-appeal-of-a-number"&gt;The appeal of a number&lt;/h2&gt;
&lt;p&gt;It is easy to see why idea tools give scores. A founder wants to know whether the idea is good, and &amp;quot;seven out of ten&amp;quot; looks like an answer. Investors and programme managers like scores too, because they make a pile of ideas sortable.&lt;/p&gt;
&lt;p&gt;But a score is an answer to a question nobody can actually answer: how good is this idea? The honest reply depends on who runs it, which customers it reaches, at what price, and how the world moves over the next two years. Compressing all of that into one number does not remove the uncertainty. It hides it.&lt;/p&gt;
&lt;h2 id="what-a-single-score-hides"&gt;What a single score hides&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;The reasons.&lt;/strong&gt; A score of six could mean a strong idea with one fatal assumption, or a mediocre idea with nothing obviously wrong. The number is the same; what you should do next is completely different.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The evidence.&lt;/strong&gt; A score built on measured figures and a score built on guesses look identical. The reader cannot tell how much of it is known.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The founder&amp;#x27;s own bar.&lt;/strong&gt; A side project and a venture-backed company need very different outcomes to be worth doing. A single scale ignores that the same result can be a success for one and a failure for the other.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The next step.&lt;/strong&gt; A score tells you how you did. It does not tell you what to test on Monday.&lt;/p&gt;
&lt;h2 id="what-we-show-instead"&gt;What we show instead&lt;/h2&gt;
&lt;p&gt;Startup Intelligence was built on a plain position: nobody can rate an idea. So the report gives no single score. Instead it shows the parts a decision actually rests on.&lt;/p&gt;
&lt;ul&gt;&lt;li&gt;&lt;strong&gt;The assumptions the idea depends on.&lt;/strong&gt; The five to eight load-bearing ones are marked, each with a grade for its evidence.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;A range of outcomes, against your own threshold.&lt;/strong&gt; You set the bar. The report simulates thousands of possible futures and shows how many clear it. Because the bar is yours, this is a statement about your assumptions, not a grade of the idea.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Every figure with its evidence rung.&lt;/strong&gt; A seven-rung ladder runs from a measured figure to an unsourced belief, and the rung is printed beside each number. Weak evidence also widens the range, so a guess can never look precise.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;Stop points.&lt;/strong&gt; For each load-bearing assumption, the value at which the plan stops working.&lt;/li&gt;&lt;li&gt;&lt;strong&gt;The next three experiments&lt;/strong&gt;, cheapest and most decisive first.&lt;/li&gt;&lt;/ul&gt;
&lt;p&gt;None of these can be sorted into a league table, and that is the intent. Two ideas can be compared on what they rest on and how close each is to its own kill lines, which is a better conversation than which one scored higher.&lt;/p&gt;
&lt;h2 id="a-note-on-the-reports-first-page"&gt;A note on the report&amp;#x27;s first page&lt;/h2&gt;
&lt;p&gt;The report&amp;#x27;s first page states which next step the evidence supports, chosen from a fixed set (for example, to run cheap, bounded tests before anything bigger), with a conviction level and the rule behind it printed beside it so you can check it and disagree. Conviction there describes the evidence, not the idea. It is not a score, but it is a summary, and like everything in the report it should be read together with the assumptions underneath it.&lt;/p&gt;
&lt;h2 id="what-this-means-for-you"&gt;What this means for you&lt;/h2&gt;
&lt;p&gt;If you want a number to show someone, a Startup Intelligence report will disappoint you. If you want to know what your idea depends on, how much of it is known and what to do first, it is built for that. Today its base figures come from the AI model&amp;#x27;s own knowledge, and the report files them on the lowest rungs of the ladder and says so, which is itself part of not pretending.&lt;/p&gt;
&lt;h2 id="takeaway"&gt;Takeaway&lt;/h2&gt;
&lt;p&gt;Distrust a single number about a future nobody has seen. Ask what it rests on, how well that is evidenced and where the plan would stop. To find those parts yourself, start with &lt;a href="/blog/load-bearing-assumptions-in-a-startup-idea/"&gt;load-bearing assumptions&lt;/a&gt;, or see &lt;a href="/features/evidence-grades/"&gt;how the evidence ladder works&lt;/a&gt;. For how a stated range can be checked later, read &lt;a href="/blog/how-to-check-a-startup-forecast-later/"&gt;how to check a startup forecast after the fact&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Reports support planning. They are not investment advice and not a guarantee of results.&lt;/em&gt;&lt;/p&gt;</content>
  </entry>
</feed>
