Nov 13, 2026

Simplified Startup Business Framework: CP Squared, Revisited for the AI Era

In 2016 I wrote a one-page framework for early startups: Customer, Pain, Competition, Product. Now that AI lets anyone build anything, it matters more, because the scarcity is a problem worth solving.

Simplified Startup Business Framework: CP Squared, Revisited for the AI Era

In October 2016 I published a one-page framework called CP Squared ↗ and tweeted it ↗. The image on that post has since gone missing from the web, so this is a rebuilt version, along with what I think of it ten years later.

Back then I had been the Managing Director of the Microsoft Accelerator powered by Techstars, and I observed some startups succeed - but a lot more fail. There are a million ways to fail, but the few startups that succeeded either 1. had a team that people just wanted to acquire (if that is you - read no more!) or 2. had identified a growing customer problem and were building a product that addressed that problem. They were paddling into a breaking wave, not just paddling in the ocean.

The framework

CP Squared: Customer, Pain, Competition, Product, filled in that orderFour boxes. Fill them in order.

  1. Customer: job title, company size, company industry
  2. Pain: the words the customer uses to describe the pain
  3. Competition: company, product, annual revenue, business model
  4. Product: better than the competition; 1/10 the cost of the competition; business model

Download the blank form ↗ and fill it in for your own idea. Print it, or paste it into a doc.

The rules I wrote in 2016

  • Spend about 90% of your early effort on the Customer and Pain loop. Go back and forth between them until a real person, with a real title at a real company, describes a pain you can name. You can cheat a little if you are the ideal customer, but you still want others to validate that they too want the world that you are building for yourself. If you are not the customer and think you just have a good sense about what the customer wants, completely discount your opinion because you are just randomly guessing.
  • Then look at Competition. Whatever customers are paying today to solve the pain, your competition is collecting. That is the existing market. The market is what customers currently pay competitors, not what they should be buying. Odds are, you can shift money from your competitions, but you will not materially grow a market. So if the amount of money that your competition is earning is not that interesting to you, you are probably in a bad market.
  • Then build the Product. A new product should be 10x better than the competition, or 1/10 the price. Anything less and there is no reason to switch. This does not sound rational - customers should switch to save 10%, right? Wrong! Customers are loathe to switch for small gains. Call it inertia or laziness - but getting a customer to switch from something they already have in a somewhat working state is uncommon. You need to make the cost of switching to your product so obvious and so compelling that customers immediately and unquestionably want to switch - not after 10 calls of you wearing them down, but immediately. If you are finding yourself arguing with your customers about why they should care about the gains you are offering, you have no breaking wave.
  • A startup is a team aiming for roughly 10x revenue growth. Not every business is one, and that is fine, but the framework is built for that goal.

What changed

Almost nothing in the framework, and a great deal around it.

In 2016, building the Product box was the expensive, slow part. It took engineers, months, and money, and that cost acted as an accidental filter. You had to be fairly sure about the problem before you could afford to find out you were wrong.

Today I can describe an app to an agent in the morning and have something running on my phone by the afternoon. I am doing exactly that with the apps at apps.dyor.com ↗, and I wrote about the tools I use in How I Work ↗. Anyone can do the same. That is wonderful, and it also removes the old filter.

Why this matters more now

When anybody can build anything, building stops being the differentiator. The bottom-right box is cheap. The top-left and top-right boxes are as hard as they ever were, and nothing AI-assisted makes them easier on its own:

  • Customer: an agent can generate a persona in seconds. It cannot sit across from a real person and notice what they actually do.
  • Pain: a model can list plausible problems. Only the customer can tell you which one keeps them up at night, and in what words.
  • Competition: the 1/10 price rule is now harder to hit, because everyone else can also build cheaply.

The risk is a new version of the old failure. In 2016 founders burned months building before talking to customers. Now they can burn a weekend, ship five apps, and feel productive while skipping the 90%. Faster building without the Customer and Pain work gets you to the wrong answer sooner.

My own apps are a fair example. They are all at $0 in revenue. I can ship quickly, and the framework is a reminder that shipping is not the hard part.

How I ~~would use~~ am using it today

  1. Write the Customer and Pain boxes before opening an editor or an agent. If a box is empty, go talk to people.
  2. Use AI to speed up the loop, not to skip it. Draft interview questions, summarize what you heard, find competitors and what they charge. Keep the conversations themselves human.
  3. Fill in Competition from real numbers. What do customers pay today, and to whom?
  4. Only then build. Let the agent do what it is good at, and hold the result to the Product box: better, or 1/10 the cost.

The tools got dramatically better. The question underneath did not change: is there a customer with a pain worth solving, and will they pay you to solve it? The real skill now is finding problems worth solving.

If you try the framework, I would like to hear how it goes.

Matt Dyor
by Matt Dyor
More Posts
It Took Two Tries to Get Buen Día onto Google Play

It Took Two Tries to Get Buen Día onto Google Play ↗

Google rejected Buen Día for freezing on a low-memory review device. Here's what we changed, the timeline, and how long each build took.

Published Nov 21, 2026

Buen Día: From Native Swift to Kotlin Multiplatform, 82% Shared

Buen Día: From Native Swift to Kotlin Multiplatform, 82% Shared ↗

I took a native iPhone app written in Swift and rebuilt it in Kotlin Multiplatform and Compose Multiplatform. About 82% of the code is now shared, and the same app went to Google Play.

Published Nov 17, 2026

30,000 Lines in Two Weeks: Massive Changes in the Hunt for Product-Market Fit

30,000 Lines in Two Weeks: Massive Changes in the Hunt for Product-Market Fit ↗

Golf Sparks 1.1 added about 30,000 lines of app, web and backend code in two weeks, roughly a quarter of the whole codebase, and it reached both stores the same night. A new batch is already waiting. This is the new normal.

Published Nov 15, 2026

Appos: Operator, Runner, Publisher, and Tiny Projects in One Window

Appos: Operator, Runner, Publisher, and Tiny Projects in One Window ↗

I took four standalone Mac apps for building and shipping mobile apps and put them in one window. Appos is a single project list with Run, Publish, and Tasks one click away, and each tool still works on its own.

Published Nov 10, 2026

Going to Production on Google Play and the App Store

Going to Production on Google Play and the App Store ↗

What it actually took to take Golf Sparks from testers to production review on both stores in one weekend: payments you can test before launch, the store-console forms that block you, and a short checklist.

Published Nov 2, 2026

How I Work with Operator, Runner, Publisher, and Tiny Projects

How I Work with Operator, Runner, Publisher, and Tiny Projects ↗

A walkthrough of Operator, Runner, Publisher, and Tiny Projects: the small Mac apps I use to organize projects, build and test mobile apps, handle releases, and keep follow-up work visible.

Published Oct 30, 2026

Page generated at: 2026-11-22T00:22:06.988Z