There is a tidy way to set a software price that every pricing book recommends, and a messy way that a lot of early-stage products actually use, and this is an honest account of the second one. The tidy way is bottom-up. You measure the value your product creates for one user, you find the share of that value the user will part with, and you set a price somewhere inside the band that survey work hands you. The messy way is top-down. You start with a number you have promised an investor, divide it by a market you read off a slide, and work backwards until a per-seat price falls out the bottom. The well-log tool we helped build, a vertical product that turned scanned paper logs back into digital curves, was priced the second way. Its headline number was 1,200 dollars per user per year, and the most interesting thing about that number is that almost nothing about it came from a user.
We want to be precise about what is ours to claim here and what is not. The pricing methods we lean on are public and old. The arithmetic the price had to satisfy was specific to one product and one fundraising moment, and that arithmetic is the part worth writing down, because the pricing essays you can find rarely show the top-down walk in full. They tell you how to find willingness to pay. They are quieter about what you do when the price has a second master: a revenue target that was fixed before a single customer had been asked anything.
The number every pricing guide tells you to find
Start with the credit, because the field has earned it. The oldest practical tool for finding a price is van Westendorp's Price Sensitivity Meter, which asks four questions about when a product feels too cheap, cheap, expensive, and too expensive, and reads an acceptable band off where the answers cross [1]. It is crude and it is still everywhere, fifty years on, because it produces a defensible range from a short survey. The modern version of the same instinct is the value-based school: Nagle and Mueller's standard text draws the line between pricing off the value you create, pricing off your own costs, and pricing off a competitor, and argues hard for the first [3]. Ramanujam and Tacke push the timing further, insisting the willingness-to-pay conversation belongs before the product is built, so the product is designed around a price the market has already blessed rather than the other way round [2].
Every one of those methods shares a direction of travel. They move from the customer inward: from what a user values, to what a user will pay, to a number. That is the right direction when you have customers to ask and time to ask them. It is the direction a mature product should always eventually face. And it is not the direction the 1,200 dollar number was first derived in, which is the whole reason this is a field note and not a textbook summary.
The arithmetic that came first
Here is the top-down walk, stated plainly. The product plan opened not with a user but with a market. The oil and gas transactions market was sized at 134 billion dollars, and the slice that was actually addressable software, the oil and gas technology market, at 6.7 billion dollars [5]. Against that serviceable figure the plan set a serviceable-obtainable share of 3 percent to be reached by the end of year five. Three percent of 6.7 billion dollars is 180 million dollars, and that 180 million dollar figure was the load-bearing target the rest of the model hung from. The first-year revenue projection, a more modest 6 million dollars, was the near-term checkpoint on the way there.
Notice that none of those four numbers is a price. They are a market, a share, a far target, and a near target. The price is what you get when you divide the targets by the thing you sell, and the thing this product sold was a seat. So the question reshaped itself from "what will a petrophysicist pay for this" into "what per-seat price makes 180 million dollars a plausible number of seats." At 1,200 dollars a seat, 180 million dollars is 150,000 seats, and 6 million dollars is 5,000 seats. Those two seat counts, not the dollar figures, were the real test of whether the price was sane: 5,000 seats inside a year and 150,000 by year five had to look reachable to a working salesperson, or the price was wrong no matter what a survey said.
The ledger above is the top-down walk made draggable. Move the seats lever and the only multiplication in the whole model runs in front of you: seats times 1,200 dollars is the revenue, and the revenue lands somewhere on the road between the 5,000-seat first-year checkpoint and the 150,000-seat year-five target. The funnel on the right is the same destination shown the other way, as a 3 percent sliver of the 6.7 billion dollar serviceable market sitting inside the 134 billion dollar total. What the picture makes visible is the load the price quietly carries. Halve the price to 600 dollars and every seat target doubles, so 150,000 seats becomes 300,000, and a salesperson who could believe one number stops believing the other. The price is not really pricing the product. It is calibrating how many customers the revenue promise requires.
Why 1,200 survived the round trip
A top-down price is only worth anything if it survives a bottom-up check, and this is where the public methods come back in as a test rather than a source. The honest question is whether 1,200 dollars a year sits inside any plausible willingness-to-pay band for the buyer, because if it does not, the seat counts are fiction and so is the 180 million dollars. For this product the band was wide and the price sat low in it, which is what made the number defensible. The work the tool replaced was an interpreter pulling curves off scanned logs by hand, slow expert labour billed at professional rates; against the cost of that labour, 1,200 dollars a year per user is a rounding error, well under any threshold a van Westendorp survey of that buyer would have flagged as expensive [1]. The value-based logic of Nagle and Mueller would have permitted a far higher number [3]. The price was deliberately set well below what the value justified, and that is a choice, not an accident: a low per-seat price trades captured value for seat volume, and seat volume was exactly the variable the 150,000-seat target needed to be large.
There is also a structural reason a per-seat model fit, which the benchmark literature documents. Seat-based pricing scales revenue with a customer's own adoption, so a single logged-in account can grow from a pilot user to a whole interpretation team without a new contract, and the SaaS expansion data shows that land-and-expand motion is where durable revenue growth in these businesses actually comes from [4]. A 1,200 dollar seat is a small enough first cheque to clear a pilot budget without a procurement fight, and a per-seat meter means the account grows itself afterward. The price was low on purpose at the door precisely so the expansion the model assumed had room to happen.
What the top-down walk hides, and what it does not
It would be dishonest to present the top-down derivation as a better method than the bottom-up one. It is a more fragile method that happened to produce a robust number, and the fragility is worth naming because the pricing guides understate it. A price reverse-engineered from a market-share target inherits every soft assumption above it: the 134 billion dollar market is someone else's estimate, the 3 percent share is an aspiration dressed as a plan, and the 150,000 seats are an output of arithmetic, not a pipeline anyone had built. Change the share assumption from 3 percent to 1 percent and the same 1,200 dollar price now implies 50,000 seats for a third of the revenue, and the entire story the price tells about the business changes without the price moving a cent. The bottom-up methods are slower but they are anchored to something real, a buyer's actual willingness to pay. The top-down walk is fast and it floats.
What saved this particular number was that the two walks happened to meet. The top-down arithmetic demanded a low price to make the seat counts large, and the bottom-up value of replacing expert hand-digitisation permitted a low price without leaving obvious money on the table. When those two pressures point the same way you get a price you can defend from either direction, which is the only kind of reverse-engineered price worth trusting. When they point opposite ways, the top-down number is the one that has to give, because you can argue with a salesperson about a quota but you cannot argue a buyer into a willingness to pay they do not have.
Key takeaways
- The 1,200 USD per-seat annual price for the vertical well-log tool was derived top-down, not bottom-up: it fell out of a market size and a revenue target rather than out of a willingness-to-pay study. The pricing literature (van Westendorp's Price Sensitivity Meter, Nagle and Mueller on value-based pricing, Ramanujam and Tacke on pricing before the build) almost all walks the other direction, from the customer inward.
- The load-bearing number was not a price but a target: 3 percent of a 6.7B USD serviceable market is 180M USD by end of year five, with a 6M USD first-year checkpoint. The 134B USD total and 6.7B USD serviceable figures are Deloitte's market sizing carried into the plan.
- Dividing the targets by a seat gives the real sanity test: at 1,200 USD a seat, 6M USD is 5,000 seats and 180M USD is 150,000 seats. Those seat counts, not the dollar totals, are what a salesperson has to believe, and halving the price doubles every count.
- 1,200 USD survived the round trip because the top-down and bottom-up pressures pointed the same way: the arithmetic wanted a low price for large seat volume, and the value of replacing slow expert hand-digitisation permitted a low price without leaving obvious money on the table. A deliberately low seat price also fit the land-and-expand motion that SaaS expansion benchmarks show drives durable growth.
- A reverse-engineered price is fragile because it inherits every assumption above it: move the share target from 3 percent to 1 percent and the same price implies 50,000 seats for a third of the revenue, with the price unchanged. When the two walks disagree, the top-down number is the one that must give, because a buyer's willingness to pay is not negotiable the way a quota is.
A price like this one is best read not as a statement about the product but as a hinge between two stories that have to stay consistent: the story a founder tells an investor about a market, and the story a salesperson tells a buyer about a seat. The 1,200 dollar number was the single point those two stories were forced to agree on, and it earned its place by being a price both of them could say out loud without flinching.
References
[1] van Westendorp, P. H. NSS Price Sensitivity Meter (PSM): a new approach to study consumer perception of prices. ESOMAR Congress (1976). The survey method that derives an acceptable price band from four willingness-to-pay questions. https://en.wikipedia.org/wiki/Van_Westendorp%27s_Price_Sensitivity_Meter
[2] Ramanujam, M., and Tacke, G. Monetizing Innovation: How Smart Companies Design the Product Around the Price. Wiley (2016). The case for holding the willingness-to-pay conversation before the product is built. https://www.simon-kucher.com/en/insights/monetizing-innovation
[3] Nagle, T. T., and Mueller, G. The Strategy and Tactics of Pricing. Routledge, 6th edition (2018). The standard text separating value-based pricing from cost-plus and competitor-indexed pricing. https://www.routledge.com/The-Strategy-and-Tactics-of-Pricing-A-Guide-to-Growing-More-Profitably/Nagle-Muller/p/book/9780134308524
[4] OpenView Partners. 2022 SaaS Benchmarks Report (2022). Industry survey of net revenue retention, expansion, and pricing-model prevalence across private SaaS companies. https://openviewpartners.com/2022-saas-benchmarks-report/
[5] Deloitte. Oil and Gas Transactions and Technology Market Analysis (2022). The market-sizing source for the 134B USD oil and gas transactions total and the 6.7B USD oil and gas technology serviceable market carried into the product plan. https://www2.deloitte.com/us/en/industries/oil-gas-chemicals.html
