A FiscEdge Academy playbook — Build with AI series
Before you start
Pricing is the fastest lever you own. A 10% price increase drops almost entirely to the bottom line, while a 10% gain in volume drags acquisition cost, support load, and infrastructure along with it. Everyone knows this. Almost nobody acts on it, because acting requires a number, and the number feels like a guess.
So founders do one of three things. They copy a competitor, which imports that competitor's cost structure and customer mix along with the price tag. They pick a round number that "feels right", which is gut feel wearing a suit. Or they ask customers what they would pay, which is the single most useless question in pricing research.
That last one deserves a moment, because it is the one that sounds most reasonable.
Ask someone "what would you pay for this?" and you are not measuring willingness to pay. You are measuring how much they like you. People lowball to protect their budget, or they inflate to be encouraging, and neither number survives contact with a checkout page. Worse, the question anchors them. Say "$49?" and you have handed them the answer.
In 1976 a Dutch economist named Peter van Westendorp published a way around this. Instead of asking for a price, you ask four questions about ranges of acceptability, none of which can be answered with a strategic lie. Then you plot where the answers cross. Business schools still teach it fifty years later because it works, and because it never asks the question people cannot answer honestly.
The reason it was never standard practice for small companies is that the analysis was tedious. Four distributions, cumulative curves, four intersections. That is an afternoon with a spreadsheet, or an analyst you cannot afford.
That part is now a prompt.
This playbook is five moves. Each one is written the same way:
- The move — the concrete thing you do
- Copy this — the text you can use as-is
- Done when — the signal you can go to the next move
Move 1 — Pick the one thing you are pricing
The move. Before you write a single question, decide exactly what the respondent is pricing. Not your company. Not your roadmap. One specific offer, at one specific scope, for one specific unit of time.
This sounds obvious and it is where most surveys die. If half your respondents are picturing a solo plan and half are picturing a team seat, your four curves will cross in a wide, useless smear. The method assumes everyone is valuing the same object.
Write one sentence that a stranger could not misread. It needs three things: what they get, who it is for, and the billing unit.
Copy this.
[Product] gives [specific role] [the specific outcome], billed per [seat / month / project].
Example: "Ledger gives a solo SaaS founder a monthly close with categorized transactions and a runway forecast, billed per month."
If you cannot write that sentence, you do not have a pricing problem yet. You have a positioning problem, and no survey will fix it.
Done when you have one sentence, under 25 words, that names an outcome rather than a feature list.
Move 2 — Ask the four questions, exactly as written
The move. Send these four questions, in this order, with your one-sentence description at the top. Do not reword them to sound friendlier. The wording is doing real work: each question probes a different edge of the acceptable range, and two of them deliberately probe downward, which is what makes the method resistant to strategic answers.
Copy this.
Here is what I am asking you to price:
[your one sentence from Move 1]
Four quick questions. Answer with a dollar amount for each, per [your billing unit]. There are no wrong answers and I am not selling you anything today.
- At what price would this be so expensive that you would not consider buying it?
- At what price would this be getting expensive, but you would still think about buying it?
- At what price would this be a bargain — great value for the money?
- At what price would this be so cheap that you would question the quality?
Question 4 is the one people want to cut. Keep it. It is the floor of your range, and it is the question that catches the respondents who were going to tell you everything should be free.
Three rules that protect the data:
- No price anchors anywhere. Not in the intro, not in a screenshot, not in the follow-up. One number in view and the whole exercise collapses toward it.
- Ask in this order. Expensive first, cheap last. Starting at the cheap end drags every subsequent answer down.
- Free text, not multiple choice. Ranges to pick from are anchors with extra steps.
Done when you have the four questions in a form, an email, or a DM, with your one sentence on top and no number visible anywhere.
Move 3 — Survey the right twenty people
The move. You need 20 to 30 responses from people who plausibly buy this. Not 200 responses from a general audience. A tight sample of real buyers beats a large sample of bystanders, because the method finds the crossover points of a population, and a mixed population has no meaningful crossover.
Twenty is enough to see the shape. Thirty is comfortable. Past about fifty you are refining a decimal that your pricing page will round off anyway.
Who to ask, in order of data quality:
- Current customers, if you have them. Best signal by a distance.
- People who churned. Uncomfortable, unusually honest, and they will tell you where your ceiling actually was.
- Prospects who went dark after a demo. They already evaluated you.
- Cold ICP matches. Workable, lowest signal, needs the most screening.
Three people will poison your data. Screen them out:
- Your friends. They will price generously to be supportive. Every number is inflated and you will not know by how much.
- People who cannot buy. If they do not hold the budget or influence it, they are pricing a hypothetical.
- The wrong company size. A 200-person company and a solo operator do not share a price curve. If you serve both, run two separate surveys and expect two different answers. That is not a flaw in the data, that is your pricing tiers telling you they exist.
Done when you have 20+ complete responses, all four answers filled, from people who could actually sign a contract.
Move 4 — Run the prompt
The move. Paste the raw answers into Claude or ChatGPT with the prompt below. You do not need to clean, sort, or average anything first. Raw is better, because the model can flag the inconsistent responses that a pre-cleaned average would hide.
Copy this.
You are a pricing analyst. Below are raw responses to a Van Westendorp Price Sensitivity Meter survey for the following product:
[your one sentence from Move 1]
Each respondent answered four questions:
- TOO EXPENSIVE (would not consider)
- EXPENSIVE (getting pricey, would still consider)
- BARGAIN (great value)
- TOO CHEAP (would question quality)
RESPONSES:
[paste everything here, one respondent per line, in any format]
Do the following, in order:
- Clean the data. Flag and exclude any respondent whose answers are not in the logical order TOO CHEAP < BARGAIN < EXPENSIVE < TOO EXPENSIVE. Tell me how many you excluded and why. Do not silently drop them.
- Build the four cumulative curves and find the four standard intersection points:
- Point of Marginal Cheapness (PMC): the floor
- Point of Marginal Expensiveness (PME): the ceiling
- Optimal Price Point (OPP): where "too cheap" and "too expensive" cross
- Indifference Price Point (IPP): where "bargain" and "expensive" cross
- Give me the Range of Acceptable Pricing (PMC to PME) and state where OPP and IPP sit inside it.
- Recommend one number and explain the reasoning in plain English, in under 150 words.
- Tell me how much to trust this. Note the sample size, how tight or wide the range is, and whether the responses look like one population or two clusters. If it looks like two clusters, say so explicitly — that usually means two pricing tiers.
Show the four intersection values as a simple table. No charts.
Step 5 is the part most people leave out, and it is the part that keeps you honest. A tight range from 28 clean responses is a decision you can act on. A wide range from 12 responses with 5 exclusions is a signal to go collect more data, not a number to put on your pricing page.
Done when you have four intersection points, one recommended number, and an honest read on whether the sample supports it.
Move 5 — Read it like an operator, not a statistician
The move. You now have a range and a number. Here is what each part actually means for the decision in front of you.
The Range of Acceptable Pricing (PMC to PME) is your negotiating room. Anything inside it is defensible to a customer. Anything below the floor triggers quality doubt, which costs you more than the discount saves. Anything above the ceiling and you are outside the consideration set entirely, which is a different and worse problem than being expensive.
The Optimal Price Point minimizes the number of people who reject you on price in either direction. It is the safe answer. It is rarely the right answer for a young company, because it optimizes for breadth at exactly the moment you should be optimizing for margin and for the customers who value you most.
The Indifference Price Point is usually the more interesting number. It is where "this is a bargain" and "this is getting expensive" cross, which in practice is the price the median buyer considers fair. For an established product with a broad market, it is often close to what the category leader charges.
Three ways to act on it:
- You are under the floor today. Raise the price. This is the most common finding, and it is nearly always the highest-return action in this playbook. Grandfather existing customers, apply the new price to new signups, and stop apologizing.
- You are inside the range but below IPP. You have room. Test a raise on new signups and watch conversion, not complaints. Complaints are loud and free; conversion is quiet and real.
- The range came back split into two clusters. You do not have one price, you have two segments. Build a tier for each rather than averaging them into a number that serves neither.
Two failure modes to check before you commit.
The range is enormous — say $9 to $400. That is not a pricing answer, it is a positioning problem surfacing as a pricing problem. Your respondents are not picturing the same product. Go back to Move 1 and tighten the sentence.
The number is below your unit costs. The survey measures what this market will bear for this offer. If that is under your cost to serve, no pricing exercise fixes it. Either the cost structure changes, the offer changes, or the segment changes. Pricing below cost with the intention of "making it up in volume" is how you scale a loss.
Done when you have picked a number, written down the reasoning in two sentences, and set a date to revisit it.
The one-page version
- One sentence. What they get, who it is for, billing unit. Under 25 words.
- Four questions. Too expensive, expensive, bargain, too cheap. In that order. No anchors.
- Twenty to thirty real buyers. Customers, churned users, dark prospects. Not friends.
- One prompt. Clean, plot, intersect, recommend, and rate its own confidence.
- Read it. Under the floor means raise. Two clusters mean two tiers. Huge range means positioning, not pricing.
Twenty minutes of survey and one prompt beats another quarter of second-guessing your pricing page.
A note on what this does not do
Van Westendorp measures price acceptability, not demand. It tells you the band where a price feels reasonable to your market. It does not tell you revenue at each point, it does not model how many people buy at each price, and it does not account for a competitor cutting their price next month.
For most companies below Series A, that is fine. You are trying to stop being obviously mispriced, and this finds that fast. When pricing becomes the thing your growth actually turns on, graduate to conjoint analysis or a real price test with holdouts.
Until then: run the four questions. Most founders discover they are under their own floor.