How should a startup decide its pricing and packaging?

Monetization is several fronts, whichever apply: who pays, for what, how much, how packaged. Each front is a hypothesis with a number and a date, tested in conversation and in the smallest possible market test before launch, not decided at launch. Pricing is a discovery question, and the companies that treated it as a launch question have had to apologize in public.

What are monetization hypotheses and how do you test them?

Monetization is several fronts, whichever apply: who pays, for what, how much, how packaged. Each front is a hypothesis with a number and a date, tested in conversation and in the smallest possible market test before launch, not decided at launch. Pricing is a discovery question, and the companies that treated it as a launch question have had to apologize in public.

Should pricing be decided before or after launch?

Monetization is several fronts, whichever apply: who pays, for what, how much, how packaged. Each front is a hypothesis with a number and a date, tested in conversation and in the smallest possible market test before launch, not decided at launch. Pricing is a discovery question, and the companies that treated it as a launch question have had to apologize in public.

Monetization as hypotheses · Business Design

§The hypotheses: on each monetization front, who pays, for what, how much, and how it is packaged, we have a specific claim with a number and a date by which a conversation or a test will have moved it.

§Ramanujam and Tacke's finding, from studying thousands of launches, is blunt: most products that fail commercially failed because monetization was designed after the , and by then the product had been built for a value the market would not pay for. This chapter puts monetization where the previous chapters have been putting everything else: before the build, as hypotheses.

§ 9.1The fronts#

§Monetization is not one decision. It is several, and a first version usually has a position on three or four of them.

§Who pays. The user, a buyer above the user, a partner, a side of a marketplace, an advertiser. Often not the segment that loves the product most; chapter 3's three piles said where the would-pay pile was.

§For what. Access, usage, outcome, status, data, a service around the software. The value metric, in pricing language: the thing the price scales with. Chosen badly, it charges for what people give freely or gives away what they would pay for.

§How much. A number. Not a range; a number, with a about what happens at half and at double.

§How packaged. Tiers, a free layer, seats, credits, a bundle with something the partner sells. Packaging is where the value proposition by segment comes back: each package is for a segment, and a package for everyone is for no one.

§Each front gets one line on the sheet: the hypothesis, the number, the evidence so far, the date and method by which it will be tested. Whichever fronts apply.

§ 9.2Tested in conversation, then in the smallest market#

§The first test is the 's conversation, with money in it. Not "would you pay for this," which gets a kind yes, but the questions Ramanujam and Tacke recommend: what would be an acceptable price, what would be expensive, what would be so expensive you would not consider it. Three numbers per interviewee, and a pattern emerges by the eighth.

§The second test is the smallest market that can show behavior: a pricing page with real numbers and a way to say yes, a pilot priced at the hypothesis, a concierge sold at the intended price to three customers. The sixth whitepaper owns experiments once the product is live. Before it is live, the test is a page and a conversation, and Kevin Hale's advice holds: charge from the beginning, because a price of zero teaches you nothing about value and teaches the customer that the product has none.

§ 9.3Public lessons#

Case

Two AI-era companies changed their monetization in public and the record is instructive. Cursor moved in mid-2025 from a request-based plan to a usage-based one and published a clarification and an apology within weeks, after users found the new packaging opaque and costs surprising; the company refunded and rewrote the plan. The lesson is not that usage pricing was wrong. It is that a packaging change shipped as a launch, without the conversations that would have found the confusion first, cost trust that a hypothesis test would not have. Midjourney, by contrast, ended its free trial in early 2023 when abuse and demand made the free layer's cost unsupportable, and said so plainly; it had learned in the market what its free unit cost, and acted on the number. Both companies had the same underlying problem, a unit whose cost rose with use. One found out from its customers after the fact; the other found out from its costs and moved first.

§ 9.4The gave-freely pile, again#

§Chapter 3 marked what people give away. Monetization must not charge for it and must not pay for it. In the case, the point economy would have paid guests for data they gave gladly, which was a cost with no return; and a subscription for being recognized would have charged for what a good restaurant does anyway, which was a price with no taker. The monetization fronts that survived were elsewhere: the occasional customer's table on a full night, and the manager's knowledge of regulars across brands and staff. Those went on the sheet as hypotheses with numbers. Recognition went on as the mechanism that made them possible, price zero, cost a glance.

§ 9.5What has to be true#

§Added to the running list: for each monetization front that applies, the hypothesis, the number, the evidence so far, and the date and method of the test. The value metric, and confirmation that it is not in the gave-freely pile. The three-number conversation results per segment, when they exist. And the smallest market test that will show behavior before launch, with an owner.

9. Monetization as hypotheses