A Quince AI × InFiNet Code workshop

Real value vs. hype.
Test your AI idea in one afternoon.

Most AI demos die in the market, not in the editor. In one afternoon you'll learn to put a price tag on value, niche a go-to-market three levels deep, and size a market bottom-up. Then you run your own idea through all of it.

Start reading Open the workshop guide

In a live session or on your own: everything on this page works self-paced. Read the pages, then test your idea in the worksheet below.

60 minlecture: value, GTM, market analysis
1 houryour idea through the worksheet
3 minyour pitch, with numbers and names
01 · Read

The lecture, page by page

By the end you can do three things: put a price tag on an idea, niche it three levels deep, and size its market bottom-up. Each page tells its part of the story on its own, so read at your pace.

← → or Space to turn the page · F read full screen · G all pages · click the progress bar to jump · swipe works too

InFiNet Code × Quince AI · Workshop

REAL VALUE vs. HYPE

Test your AI idea in one afternoon.

What this afternoon is really about

Generative AI made it astonishingly cheap to build something that looks impressive, which is wonderful and also a problem, because the thing that used to separate the winners from everyone else has quietly stopped being the hard part. This workshop is about the part that is still hard, and it unfolds as a single story in three movements. First we learn to tell genuine value apart from a convincing demo. Then we find the narrow slice of the market that you, with no budget and no team, can actually win. Finally we check that the numbers underneath the idea are honest enough to build a business on.

And then you build something real

When the story is told, you get an hour to run your own idea through exactly the same path, using the worksheet waiting below these pages. Nothing here is theory for its own sake; every idea on every page ends in something concrete you can do this coming week.

Move with ← → or swipe, and read at whatever pace suits you.

Why you're here

Your code is rarely why your startup dies.

Everything impressive became cheap, all at once

A single developer can now ship in a weekend what an entire funded team needed a year to produce back in 2018, and that changes what is possible for someone working alone. The catch hides in plain sight: the cost of building did not fall only for you. It fell for everyone, at the very same moment, because the same Lovable, the same Claude Code, the same Codex sit on every other laptop in the room and far beyond it. When a capability is universal, it can no longer be the thing that makes you special.

So what is still expensive, and therefore still valuable

Knowing a particular customer deeply enough to predict their week. Earning the kind of trust that survives a failed deploy. Owning a channel that reliably puts you in front of the right people. Proving, in numbers a buyer believes, that you changed their outcome. None of these got cheaper, and so value quietly migrated from the old question, "can you build it," to a far harder one: do you know exactly whose problem this is, and what it is costing them right now?

Every hype cycle before this one, from dot-com to mobile to blockchain, punished capability-first ideas and rewarded the pain-first ones. The technology was always the enabler, never the reason anyone reached for their wallet.

Part 1 · Value vs. Hype

The hype trap

The first founder starts with the technology

She opens a model playground, watches it summarize a long document in seconds, and feels the obvious pull: this is amazing, let us build a summarizer. Ask who it is for and the answer is everyone; ask what the value is and the answer arrives as an adjective, smarter or seamless or ten times faster. Her pitch leads with the model she built on, and right now the market is crowded with exactly this founder, all of them impressive and most of them interchangeable.

The second founder starts with a person in pain

He begins somewhere completely different, with a sentence like "this particular person loses twelve hours every month doing X by hand, so can we simply remove that." His target is a role, a sector, sometimes an individual he can name. His value is a number he can defend. And the artificial intelligence inside his product is almost invisible in the way he describes it, which is precisely where it belongs, because it was never the point.

Nobody has ever wanted to buy "AI." People buy fewer hours, fewer errors, and fewer fines, and they buy them from whoever understands the hours, errors and fines best.

Part 1 · Value vs. Hype

The 30-second test

Describe your product out loud without naming a single piece of technology.

The sentence either survives or it collapses

Try this with the first founder and the sentence falls apart in real time: "it is, well, an AI that summarizes things," followed by an awkward silence when you ask for whom and why. Try it with the second founder and he barely notices the constraint, because his natural description never needed the technology in the first place: "property managers stop retyping inspection reports, so an evening of work becomes ten minutes." The test works because hype lives in the technology words, and removing them removes the disguise.

Four things a real founder can always name

Who actually pays, as opposed to who merely benefits. What those people do today instead, because there is always a current way of coping, even if it is suffering quietly in a spreadsheet. What that current way costs them, in hours or money or risk. And what measurably changes for them after 30 days of using your product. Hold these four in your head; you will meet them again as the first questions in your worksheet.

Try it right now, silently: describe your own idea without the tech words. Did the sentence survive, or did it need rescuing?

If you cannot answer all four, you do not yet have a product. You have a demo, and the market is unkind to demos.

Part 1 · Value vs. Hype

Where real ideas come from: your daily life

The fruit seller sees problems you never will

Picture someone who sells fruit for a living. Every day they wrestle with buying stock at the right moment, keeping it fresh before it turns, pricing it against the stall across the street, balancing the books at night and keeping the shop itself from falling apart. They live inside these problems so completely that they can feel a new one the instant it appears. What they cannot do is build the software that would make any of it easier.

You can build, but you cannot see from in here

You have the opposite condition. You can solve almost any problem once it is described to you, yet sitting all day inside a development context, you simply never encounter the fruit seller's world, and an opportunity you cannot see is one you can never seize. This is the quiet tragedy of the talented developer with no idea worth building.

The two worlds have to meet inside one person

Almost every product you admire was born when someone was so frustrated by a problem in their own ordinary life that they spent their nights solving it, not because it was a cool use of technology and not because some model could finally do it, but to make a daily ache go away. So go looking in your own life: your hobbies, your side jobs, the family business, the thing that irritates you every single week. Find the burning problem, then ask the question that separates the real opportunities from the rest: why has nobody solved this yet? Sometimes there is a good reason and you should walk away. Very often the only reason was that a solution used to be too expensive to build, and that, as of this year, is no longer true.

A founder I met recently who did exactly this

The cleanest version of this pattern I have come across lately belongs to an ML engineer whose wife works as a doctor in a Swedish hospital. A patient arrived one day who had been beaten by her husband, and she refused a translator because the community was small enough that anyone in the room might know him. So he built an on-prem translation app for hospitals, with no big-tech dependencies, because the human in front of him needed it. The product later took off in the United States for the opposite regulatory reason, where hospitals will not let big tech anywhere near patient conversations. Same pain. Different law. No marketing, an outdated WordPress site, and the right hundred people who already knew him.

You do not need a cleverer idea than everyone else. You need a problem you already live inside, because that is the one you understand better than your competitors ever will.

Part 1 · Value vs. Hype

Value is a number, or it isn't value

VALUE / customer / year = cost of the problem today − cost of your solution
12 h/momanual work today
× €80/hcost of that person
+ €10k/yrerrors · fines · rework
≈ €21.5kproblem cost / year
Why we insist on a number per customer per year

Forcing the value into euros for a single customer over a single year drags you into unit economics long before anyone is tempted to wave at a giant market, and it hands you your pricing anchor for free. A healthy business captures somewhere between 10% and 25% of the value it creates, which leaves the customer an obvious 4-10× return and leaves you a real margin. If the arithmetic cannot support a price that sustains a business, no amount of go-to-market brilliance later will rescue it. And when the objection comes, "these numbers are just guesses," the honest answer is yes, for now, because a written-down guess is the only kind you can test, and every customer conversation trades one guess for one fact.

The question that will be asked of you

A customer asked me this 2 years ago, and you will hear your own version soon enough: "why would I pay you, when ChatGPT can already do this?" Thin software wrapped around a model is dying in front of us, and your future customers will cheerfully try to vibecode their own replacement over a weekend. The only durable answer is to be so deeply nested in their specific problem, their data, their workflow, their compliance obligations, that a generic tool cannot follow you in. Refusing to estimate the value is not humility; it is a way of avoiding the moment your idea has to meet arithmetic.

The numbers above are not abstract. They belong to A11yReview, the worked example you will meet three pages from now.

Part 1 · Value vs. Hype

Three kinds of value, and the one you should hope for

Time is easy to count and hard to sell

Saved hours multiply cleanly into a number, which is why every pitch reaches for them, but saved time is treacherous because it frequently lands in no one's budget. The work disappears and the salary stays exactly the same, so before you build on a time argument, ask the uncomfortable question of whose budget actually gets smaller when you succeed.

Money is the argument a buyer believes instantly

Fewer errors, avoided fines, recovered revenue: these land directly in a spreadsheet the buyer already maintains, and their finance team can approve them without a leap of faith. When you can credibly move money, the sales conversation gets dramatically shorter.

Risk is the quiet champion of business software

Compliance, legal exposure and security carry something the other two can only envy: a deadline and an enforcer. Someone with authority is requiring the customer to act by a certain date, which converts a nice-to-have into a must-buy. This is exactly why the example threaded through this afternoon is built on accessibility law rather than on convenience.

If neither you nor the buyer can measure it, the buyer's CFO cannot approve it. In the end, money talks.

Part 1 · Value vs. Hype

Value vs. Effort: the map you steer by

EFFORT = build + sell + support + compliance
The mistake that hides in the effort axis

Almost everyone scores only the build effort, and since artificial intelligence made building tiny, they cheerfully drop their idea into the easy corner. But effort is the whole journey, not just the code. The moment building is easy for you, it is equally easy for every competitor, and that floods the market and makes the selling brutal: no differentiation, no trust, no way to be heard. This is the precise mechanism by which a flawless weekend prototype ends up stranded in the hype trap.

The map is not only for judging an idea; it is for steering one. Niching, which is the whole of Part 2, pushes your idea up and to the left at the same time, and you are about to watch it happen.

Part 1 · Meet the worked example

A11yReview: the idea we'll test all afternoon

a11y = accessibility  (the letter a, then eleven letters, then y, the same shorthand trick as i18n for internationalization). It means making the web usable for people with disabilities: screen readers, keyboard-only navigation, sufficient colour contrast, captions. The WCAG standard defines how it is done, and for the public sector both EU and Swedish law make it mandatory.
What it is, and who has no choice but to care

A11yReview sits inside a web team's continuous-integration pipeline and checks every single release against accessibility law before it ever ships, proposing the fix directly in the team's own code. Its customers are Swedish public-sector web teams, the municipalities, regions and agencies, together with the roughly 50 digital agencies that build their sites. Crucially, none of them are choosing to care about this as a preference: DIGG audits them, and since June 2025 the EU Accessibility Act has extended the same obligation across much of the private sector too.

What it replaces, and the money that makes it real

Today these teams cope with manual checking, expensive external audits that arrive too late, and scanners that vomit 400 warnings without telling anyone how to fix a single one. Add up the roughly 12 hours a month, the rework and the audit findings and you reach a problem worth about €21,500 per organisation every year. A11yReview charges €400 a month, which is around 22% of that value, leaving the customer comfortably ahead. Roughly 700 organisations can be named, and a law with a deadline is doing the selling.

It is a fictional product, but every number and every regulation behind it is real. And, as the next page admits, it did not begin life this clever.

Part 1 · Live demo

Same technology, a completely different fate

A11yReview started as something far more ordinary, and before you see where it ended up, decide for yourself how you would score that first version on value and on effort, from zero to ten.

Where it started: the hype trap

“AI code review assistant”

For every developer alive, competing against free incumbents, with no way to reach anyone. Value 3 · Effort 8

Where it landed: a quick win

A11yReview

A legal deadline, roughly 700 customers you can name, one integration to build. Value 8 · Effort 4

The underlying technology never changed. The only thing that moved was who it was for, and that single move is the whole subject of Part 2.

Part 2 · Go-to-market

Your idea already exists.
Good.

Go ahead and look. It really is there.

Search your idea on Product Hunt or in the Y Combinator directory right now and you will find it, in some shape, very probably already funded. Two reactions are tempting and both are wrong: the despair that says the space is taken, and the denial that insists "we are different" while being unable to say how. Neither will help you.

A crowded market is proof somebody paid for the answer

The healthy reaction is to recognise a crowded market for what it is, namely millions of euros that other people already spent proving that customers will pay for this. Consider that MailChimp served around 8 million customers, and sit with what that number forces to be true: you cannot possibly serve 8 million different people well with anything other than a deliberately generic product. Genericness at that scale is not a failure, it is structural and unavoidable, which means every giant incumbent is constantly manufacturing niches it cannot serve properly. The larger the incumbent, the more of these gaps exist for you.

So the useful question was never "is someone already building this." It is "which underserved slice of this proven market can I, specifically, own."

Part 2 · Go-to-market

Why the incumbent won't simply crush you

Three reasons, and they reinforce each other

First, your niche is a rounding error in their revenue, so the €168k wedge that could feed you cannot possibly justify a line on their roadmap, and it stays ignored until the day it is too late to ignore. Second, serving your niche properly would mean stripping out the very generality that their other 7.9 million customers depend on, so doing your job well is something they structurally cannot do. Third, by the time you have grown enough to matter to them, you already own the community, the trust and the exact-fit product, and they would be the ones entering your market as the newcomer.

⚠ The one exception you must respect

Foundation-model giants do not play by the rules above. ChatGPT, Gemini and the rest expand their ecosystems week after week and casually swallow thin startups with every update and integration they ship. If your entire product is one prompt away from becoming a built-in feature, it eventually will be. Watch the trap in the wild: a startup ships a charming AI digital-wardrobe app that lets you mix and match clothes from your camera roll; Google Photos ships the same feature natively, and the startup is redundant overnight. The defence is the same medicine prescribed everywhere in this workshop, only stronger: a niche so deep in a customer's data, workflow and regulator that a general-purpose platform cannot reasonably follow you there. The defence is depth, not delight.

This is also your honest answer when, during the pitches later, someone asks why the obvious incumbent will not just add your feature.

Part 2 · Case

ConvertKit: niching three levels deep

The clearest niching story I know walks down four steps, and it is worth following each one closely:

Level 0 · “email marketing”

In 2013 Nathan Barry launches with $5,000 and a public revenue challenge, taking on MailChimp head-on, and the project stalls for roughly two years because it is competing on nothing in particular.

Level 1 · “… for authors”

He turns to the customers he personally knows, literally writing down names like Chris Guillebeau and Joel Runyon, and changes the homepage to speak only to authors. The first real traction appears.

Level 2 · “… for professional bloggers”

He narrows again, reaches out personally to bloggers with 30,000 to 250,000 subscribers, and removes the real obstacle, which was never desire but the pain of switching, by migrating their lists for them by hand. Growth ignites.

Level 3 · “a hundred people you can name”

His own doctrine: a niche of a million is no niche at all. Bloggers becomes food bloggers becomes paleo-recipe bloggers becomes women paleo-recipe bloggers, a community so tight that winning the most respected member converts the rest.

It grew into tens of millions in revenue, and only then expanded back up the ladder to serve all creators, from a position of strength. The niche was the wedge that opened the door, never the ceiling on the ambition.

Part 2 · Go-to-market

Why niching actually works, mechanically

A sharp message converts, and tight communities carry it

"Email for everyone" persuades no one in particular, whereas "email for paleo-recipe bloggers" makes exactly the right person feel the product was built for them, which lifts conversion and drops the cost of finding each customer. Better still, narrow communities share overlapping social circles, so one delighted and well-known customer reaches the entire niche on your behalf, the same word of mouth that simply evaporates in a broad market.

An exact product, aimed at people you can actually find

Because the niche is small, you can make a focused product that is ten times better for those specific people than the incumbent's deliberately general version, and that generality is the one weakness the incumbent can never remove without betraying everyone else. On top of that, a true niche is findable: you can sit down and list a hundred real prospects and start contacting them this week, with no budget at all, which is the difference between a plan and a wish.

But surely niching caps how big I can get?

It feels that way, and it is wrong. Facebook began inside one university, Amazon sold only books, Tesla shipped a single expensive Roadster. Each used the niche as a beachhead and expanded from strength once it had won. The graveyard, by contrast, is full of companies that tried to be everything to everyone on the very first day.

The honest test for "deep enough" is simple: could you write down a hundred specific customers by name in one afternoon? If not, you have another level to go.

Part 2 · Go-to-market

One level down is almost never enough

It feels like niching, and it isn't

Saying "AI tools, but for developers" feels like a decisive narrowing, yet developers are still millions of wildly different people building wildly different things, and you will discover this the hard and expensive way, in your own customer conversations, long after you committed.

The warning sign The Mom Test taught us to read

Rob Fitzpatrick's short book on talking to customers names this trap precisely. Interview twenty people who are really twenty different kinds of customer and their feedback contradicts itself until it feels like meaningless noise, simply because you are quietly researching two dozen separate businesses at once. The rule that follows is clean: if you are not hearing the same problems and the same goals again and again, your segment is not yet specific enough, so go down another level. Expect to descend three or more times before the "hundred names" test finally passes.

Take comfort that niching is cheap to undo, since it is only a headline and an outreach list rather than a rewritten codebase. Staying generic is the expensive mistake. It teaches you nothing you can act on.

Part 2 · Go-to-market

The go-to-market recipe, six blocks

Once the niche is chosen, the rest almost writes itself, provided you actually work through all six blocks in order. The first three are about understanding and saying; the last three are about pricing, reaching and landing.

1 · Market insight

A clear read of the need, the behaviour, the trends and the competitors before you build anything.

2 · Target group

Separate the person who pays from the person who uses. Building for the user while selling to the buyer is a classic, fatal mix-up.

3 · Positioning

One sentence: for a niche who suffer a pain, your product is the category that delivers a benefit, unlike the alternative they use today.

4 · Offer & price

Anchored to the value equation from Part 1, capturing that 10-25% of the value you create.

5 · Channel

The one or two places your 100 already gather, plus the unfair tricks: host the meetup, give the talk, write the definitive guide everyone reads.

6 · First 10 customers

Named, or described so precisely you could find them tomorrow morning.

Notice where companies actually die: blocks 1, 3 and 5, market insight, positioning and channel. They are exactly the ones a developer skips, because none of them are code.

Part 3 · Market analysis

The four sizes of one market

Before any numbers, the vocabulary, because the same market has four nested sizes and confusing them is how pitches lose credibility:

TAM
SAM
SOM
LAM
T
Total addressable market

Everyone who could conceivably buy: for A11yReview, every EU organisation with a website that must be accessible.

S
Serviceable addressable market

The slice your product and channel can really serve, here the roughly 700 Swedish public-sector bodies and their agencies.

S
Serviceable obtainable market

What you can realistically win within about two years, here 5%, or 35 organisations.

L
Launch addressable market

Who you can reach at the very start. Can you begin 5 conversations, or 100? That is 1 customer, or 15, and your runway should be planned around it.

Investors love to hear about TAM. You, this week, actually live inside LAM.

Part 3 · Market analysis

Size it from the bottom up, or not at all

SOM = reachable customers × realistic price × realistic 2-year win rate
≈ 700reachable orgs + agencies
× €4.8kper year (≈ 22% of value)
× 5%win rate in 2 years
≈ €168kper year, and honest
Why a small, honest number beats a huge, hollow one

The top-down fantasy, "it is a €50 billion market and we will take 1%," encodes not a single real decision and cannot be tested, which is exactly why experienced investors discount it to nothing. The bottom-up figure here is small, about €168k a year, and that honesty is the entire point, because a modest number attached to a credible expansion ladder beats an enormous number attached to hand-waving. The ladder runs from Swedish public sector to the Nordic public sector of roughly 2,000 bodies, onward to every EU organisation under the Accessibility Act, and finally to general accessibility tooling: ConvertKit's climb, in reverse.

Two sanity checks worth running out loud

Does the number actually feed you? A solo developer needs perhaps €100-150k a year to live, while a venture-funded team needs a believable path to €10M and beyond. And is the win rate defensible? 5% of a named, reachable niche you can contact by hand is plausible, whereas 5% of "all small businesses everywhere" is pure fantasy.

One last check before you leave today, your LAM: how many launch conversations can you actually start, 5 or 100? That is your first 1, or your first 15, customers.

Part 3 · Market analysis

The twenty-minute competitor scan

Search the way your customer would, then sweep

Begin not with your clever product name but with the pain phrase the customer would actually type into Google, because people search for their problem and never for your solution. From there, sweep the obvious gathering places in turn: Product Hunt, the relevant category on G2 or Capterra, the Y Combinator directory, and GitHub.

The complaints are a free map of your positioning

For the leading three to five competitors, note who they target and what they charge, and then read what their users complain about in the reviews, because those complaints hand you your opening for nothing. A11yReview's entire pitch, finding and fixing problems inside the developer's own code, came straight from scanner users complaining that the existing tools dump four hundred warnings and no idea how to fix any of them.

Never forget the competitor that costs nothing

The most dangerous competitor is not another startup but the status quo: the spreadsheet, the intern, the once-a-year consultant, or simply tolerating the pain. It already has 100% of the market and asks for no money up front, so unless something has made the pain newly intolerable, it usually wins.

Without a credible "why now," a deadline, a new law, a sudden cost, the status quo quietly defeats you.

Part 3 · Market analysis

Prove the demand without writing code

The three rules of The Mom Test

The Mom Test, Rob Fitzpatrick's short and worthwhile book, gives three rules for these conversations: talk about their life rather than your idea, ask about specific things that happened in the past rather than opinions about the future, and talk less while listening more. The worthless questions, the ones that invite a polite lie, are "would you buy this" and "how much would you pay." The valuable ones dig into what already happened: what did you do the last time this came up, what did it cost you, what else have you tried.

A compliment is a polite form of failure

A meeting only truly succeeded if the other person gave up something they value: their time in the form of a scheduled pilot, their reputation in the form of an introduction to their boss, or their money in the form of a signed letter of intent or a prepayment. Warm encouragement that costs them nothing is not progress.

The landing page, done the way it actually worked

Buffer's famous landing-page test is misremembered as a numbers game; what made it work was that Joel emailed every single person who signed up and spoke with them, so the conversations were the real data and the conversion rate was almost incidental. And if you worry someone will steal the idea, remember that you barely mention it in a Mom Test conversation anyway, and that it is already being built by others, so secrecy protects nothing while silence guarantees you learn nothing.

A footnote for anyone selling in Sweden

The polite-form-of-failure problem is worse here than almost anywhere. Swedes are polite by default. "Yeah, it's good" frequently means "I will never pay for this," and you will not find out until the renewal does not arrive. Compare the US market, where a customer will tell you on the first call that this is, in their words, BS, and you know exactly what to fix. Both are useful signals. One requires twice the work to extract. Ask past-focused questions twice. Push for the specific cost of the last incident. Watch what they do, not what they say.

Demand gets the first payment. Stickiness tells you the rest.

One further trap that hides inside the compliment problem. A customer can love the demo, sign up, even pay for the first month, and then quietly let it die at renewal. The product was not bad. The pain was not deep enough to defend a recurring line item. Measure stickiness, not signups. A cancellation is the most honest customer interview you will ever get.

"Sounds great, keep me in the loop" is not encouragement. It is a rejection wearing a smile.

Workshop briefing

Your turn: one hour, one idea, one worksheet

Now the story stops being about A11yReview and starts being about your idea, walked down exactly the path you just read, using the worksheet below these pages.

Part A, twenty minutes, value or hype

You will write the technology-free sentence, build the value equation with real numbers, score the matrix for your idea as you first imagined it, and answer the kill questions that an investor or a wasted year would eventually force on you.

Part B, twenty minutes, niche three levels deep

You will find three competitors that already exist and treat that as the good news it is, ladder down until you reach the hundred names test, write your positioning sentence, choose a channel, and list your first ten customers. Then you return to the matrix in Part A, score the niched version, and watch your own dot move the way A11yReview's did.

Part C, fifteen minutes, market analysis

You will size the market from the bottom up with its expansion ladder, gather the competitor complaints, articulate your why-now, and write an evidence plan for next week. Then you compress the whole thing into a three-minute pitch: my idea was X, my niche is now Y, the value is Z per year, my first ten customers are W, and my biggest risk is V.

Numbers and names, never "many" and "companies." Choose one idea before the hour begins, because the hour is for testing it, not for choosing it.

Keep these

Three things worth carrying out the door

1

Hype describes a technology, while value describes a specific customer's saved hours, recovered money, or reduced risk, and it always carries a number.

2

Someone is already building your idea, and that is good news, because it proves the demand. Your job is to niche three levels deeper than feels comfortable.

3

A niche of a million people is not a niche at all. Find the hundred you can name, win one of them properly, and let that community carry you to the rest.

Take a short break, choose the idea you will test, and open the worksheet. The hour begins the moment you do.

Page 1 of 22
02 · Workshop guide

The worksheet: test your idea in 1 hour

Fill every box. Where it asks for numbers or names, write numbers and names. “Many” and “companies” are not answers. Answers live in this tab only: save your worksheet before closing.

My currency:

How to run this on your own

1
Read the pages above

20-30 minutes. They tell the whole story on their own, no presenter needed.

2
Pick one idea

Your own, from a hobby, a side job, a daily frustration. One. The worksheet tests it, it doesn't choose it.

3
One hour: A → B → back to A4 → C

Score the generic version in A4 first. After niching in Part B, come back and score the niched version. That return trip is the lesson.

4
Score yourself, save the PDF, then say it out loud

No audience? Pitch your phone camera or a friend for 3 minutes. Saying it out loud is the test.

A

Value or hype?

20 MIN

Goal: decide whether your idea creates measurable value, or just looks impressive. You leave this part with a € number and a position on the map.

—problem cost € / customer / year

Can't fill it in? That's normal: it means you need facts from a real customer, not better guessing. The Mom Test is a short book by Rob Fitzpatrick about getting honest answers. Its core rule: ask about facts from their past, never opinions about the future. “Would you pay for this?” invites a polite lie. “What did the last audit cost you?” gets a fact. Write 3 questions of that kind below, ones you could actually ask someone this week.

What is this for?

Before you spend an hour (or a year) on an idea, place it on a simple map: value it creates vs. total effort it takes. The quadrant tells you what to do with it. You don't guess the scores, you calculate them from six questions below.

Why score it twice?

First score your idea as you originally imagined it (the generic version), now. After Part B you come back and score the niched version. Watching your dot move between quadrants is the whole lesson: the same idea, the same tech... only "who it's for" changed.

1 · The generic version (score it now)

Value: how much is solving this worth?

0 = mild annoyance · 10 = business-critical, or required by law
0 = once a year · 10 = every single day
0 = nobody owns this problem · 10 = they already pay to fix it today

Effort: what does it really take?

0 = a weekend with AI tools · 10 = years of deep tech
0 = 100 people you can name · 10 = a crowded market, no channel, no trust
0 = none · 10 = heavy compliance and integrations

2 · The niched version (come back after Part B)

Value: how much is solving this worth?

0 = mild annoyance · 10 = business-critical, or required by law
0 = once a year · 10 = every single day
0 = nobody owns this problem · 10 = they already pay to fix it today

Effort: what does it really take?

0 = a weekend with AI tools · 10 = years of deep tech
0 = 100 people you can name · 10 = a crowded market, no channel, no trust
0 = none · 10 = heavy compliance and integrations

Dot moved? Continue to Part C →

Quick winsStrategic bets FillersHype traps
generic
niched
Worked example: A11yReview, Part A

A1, the tech-free sentence: “Web teams get every release automatically checked against accessibility law, with fixes proposed in their own code, before an auditor sees the problem.” (The first version was “an AI code review assistant”. Notice how that sentence collapses without the word AI.)

A2, who hurts: the pain: front-end leads and web managers in public-sector teams. Who pays: the municipality IT budget, or the digital agency delivering the site. Today instead: manual checks plus a yearly external audit.

A3, the value equation: ~3 h/week (12 h/month) × €80/h × 52 + €10,000/year in audit findings and rework ≈ €21,500/year per organization. Price: €400/month ≈ 22% of the value created.

A4, matrix, calculated: generic “AI code reviewer”: pain 3, frequency 6, budget 1 → value 3.3. Build 2, sell 9 (everyone competes, no channel), support 6 → effort 5.7. Hype trap. A11yReview: pain 8 (a law with a deadline), frequency 8 (every release), budget 8 (audit money is already being spent) → value 8.0. Build 4, sell 3 (700 customers you can name), support 4 → effort 3.7. Quick win.

A5, kill questions: Most likely cause of failure: audit firms bundle “good enough” tooling, or public procurement is too slow for a startup. Must be true: a municipality can buy a €400/month tool without a tender process, and one reference customer agrees to be named. Better off after 1 month: the web lead ships releases with zero critical WCAG findings and saves ~12 hours.

B

GTM: niche 3 levels deep

20 MIN

Goal: find the slice of the market you can actually win, so specific you could list ~100 customers by name. Then go back to A4 and score the niched version.

Example of what to do and write: for A11yReview you would google “WCAG audit failed fix”, open the G2 category “Digital accessibility tools”, and write: “1) Siteimprove / large organizations / ~€500+ per month. 2) axe DevTools / developers / free to ~$40. 3) Accessibility consultancies / public sector / ~€10k per audit.” Three lines, ten minutes, done.

Level 0 · generic
Level 1
Level 2
Level 3 · your wedge

How this maps to the pages: Level 0 is roughly your TAM (everyone who could buy). Level 3 is the slice you can serve, your SAM. In Part C you'll turn it into SOM (what you can win) and LAM (who you can reach at launch).

Worked example: A11yReview, Part B

B1, what already exists: 1) Generic AI code review (CodeRabbit, Copilot) / all developers / $10-30 per seat. 2) Accessibility scanners (axe, Siteimprove) / web teams broadly / free to €500+. 3) Audit consultancies / public sector / ~€10k per audit. Demand: validated.

B2, the ladder: Level 0: all developers, ~30 million. Level 1: web accessibility, millions of sites. Level 2: EU organizations legally required to comply, hundreds of thousands. Level 3: Swedish public-sector web teams + the ~50 agencies building their sites, ~700 organizations. 100-names test: passes, the municipality list is public.

B3, positioning sentence: For Swedish public-sector web teams who must pass DIGG/WCAG audits on every release but check accessibility manually and late, A11yReview is the CI-integrated compliance reviewer that finds and fixes violations before the auditor does, unlike scanners that dump 400 warnings with no fixes.

B4, channels: #1: DIGG's supplier ecosystem and the webbriktlinjer community. #2: Swedish web-agency meetups. Unfair trick: publish “The annual state of municipal web accessibility” ranking, the bottom 50 become the pipeline.

B5, first 10 customers: the 5 largest digital agencies serving municipalities + 5 municipalities flagged in published accessibility audit reports. All findable in one afternoon.

Niche found? Two steps left before Part C

Your idea just changed. Go score the niched version on the A4 matrix and watch the dot move. That move is the whole point of today.

2 · Then continue to Part C ↓
C

Market analysis

15 MIN

Goal: check there's a real business behind the wedge: honest size, real competitors, a reason it must happen now, and a plan to prove demand without writing code.

—SOM € / year
Worked example: A11yReview, Part C

C1, bottom-up sizing: customers in the niche: ~700. Price: €4,800/year. Win rate in 2 years: 5%. SOM ≈ €168,000/year. Small, and honest. Expansion ladder: Nordic public sector (~2,000 orgs) → all EU organizations under the Accessibility Act → general accessibility tooling. LAM at launch: ~15 warm conversations possible through agency contacts this month → realistically 1-2 paying pilots.

C2, competitor reality check: Complaints about the top competitor: “false positives” and “400 warnings, no idea how to fix” (straight from G2 reviews, and straight into our positioning). Why the incumbent won't add this: they sell dashboards to compliance officers, not fixes to developers in CI, a different buyer and product shape. Status quo and why-now: teams tolerate the pain because audits come yearly; the EAA enforcement deadline ends that.

C3, evidence plan: 5 people: 3 agency tech leads + 2 municipal web managers, asking about the last audit they went through, what it cost, what they did. The commitment: one agency agrees to a paid pilot on one client project. Cheapest test: a landing page “Pass your DIGG audit on the first try” posted in the Swedish web-dev Slack, then personally email every signup.

★

Pitch checklist: 3 minutes

BEFORE YOU PRESENT

Goal: compress the whole worksheet into 3 spoken minutes. No audience? Record your phone or pitch a friend. If a sentence needs an adjective instead of a number, the worksheet isn't done.

This page doesn't store anything. Your answers exist only in this tab until you save them. The PDF contains the worksheet only, never the reading pages.