You can't crowdsource a price
Post "would you pay $19/month for this" on Reddit and you'll get fifty answers from fifty different budgets, none of whom are your actual buyer. The honest way to answer "what should I charge" is Van Westendorp price sensitivity analysis: a pricing method that asks a sample of real target buyers four questions about the same product and finds the price range where demand is highest and resistance is lowest. It takes one survey, one panel of people who'd actually buy the thing, and about 15 minutes to build. Everything else, gut check, competitor matching, asking your cofounder's sister, is a guess dressed up as research.
What is the Van Westendorp pricing method?
The Van Westendorp Price Sensitivity Meter is a pricing research method that asks your target buyers four questions about the same product: at what price it would feel too cheap to trust, a bargain, expensive, and too expensive to consider. Plotting the answers across your sample produces an acceptable price range and an optimal price point, instead of a single guessed number.
The method was built in the 1970s by Peter van Westendorp, a Dutch economist, and it's held up because it sidesteps the biggest flaw in pricing research: asking people directly "would you pay $X" produces answers people think they should give, not answers that predict behavior. Asking about a range of prices from four angles lets the pattern tell you where the real ceiling and floor sit.
The output is a set of curves. Where they cross, you get a Point of Marginal Cheapness (below this, people start doubting quality), a Point of Marginal Expensiveness (above this, price kills the sale), and the workable range in between. That range is your answer, not a price plucked from a spreadsheet.
The four questions you need to ask
Every respondent sees the same product description and answers all four:
1. At what price would this be so cheap you'd start to question the quality?
2. At what price would this be a bargain, a really good deal?
3. At what price would this start to feel expensive, though you'd still consider it?
4. At what price would this be too expensive to consider at all?
The description matters more than people expect. "An app that turns a job posting into a written freelance proposal, priced monthly" gets you real answers. "An AI tool" with no context gets you noise, because respondents are pricing their own assumptions instead of your actual product. Write the description the way you'd actually market it, then hold it constant across every respondent.
How to read the results
Once the four curves are plotted, the range between the Point of Marginal Cheapness and the Point of Marginal Expensiveness is your workable zone. Inside that zone, you'll usually find an Optimal Price Point (where resistance to "too expensive" and "too cheap" are balanced) and an Indifference Price Point (where an equal number of people say "bargain" and "expensive," often the natural default if you want one number instead of a range).
This works the same way for a service business as it does for software. If you're pricing a two-person service business, you don't need a product screenshot, you need a tight, specific description of the deliverable and the four questions asked against it. The math doesn't care whether the thing being priced ships on the Play Store or gets delivered over a Zoom call.
Why an old subscription anchor doesn't answer a new pricing question
$9/month became the default SaaS price point because a specific set of products, at a specific moment, converged on it. Whether that anchor still holds for your product isn't a fact you can look up, it's a question your specific buyers answer when you ask them. Categories move. Buyer expectations move faster than founders update their pricing intuition. The only way to know if $9 is still the right anchor, or if your buyers would pay more (or notably less) for what you're building, is to run the test against your actual product description, not against the market's memory of what SaaS "should" cost.
The word "AI" changes what people will pay
This matters more than most founders pricing an AI product assume. In our own AI Markup study, we tested the same app twice, once described plainly and once relabeled "AI-powered," and median willingness to pay dropped about 25% for the AI-labeled version (p=0.02). Same product, same features, lower number, just from the label.
That's not a one-off finding. Our 2026 AI Tax Report, a survey of 1,017 U.S. consumers weighted to U.S. Census age and gender, found that 68% would choose a "human-made" product over an identical "AI-made" one at the same price, with only 10% choosing the AI version. 54% said they'd pay a premium specifically for human-made. 36% had taken a concrete action against a brand in the past six months because it felt too AI-driven, and 47% said they lose trust when they realize a brand's copy was AI-written. The effect skews young: 59% of 18 to 28 year olds have penalized a brand over AI, versus 18% of those 61 and older. You can see the live results here: the 2026 AI Tax Report.
If you're estimating realistic revenue for a subscription app with a per-use AI cost, your compute cost sets the floor, but the label you launch with can quietly cap the ceiling. Run your price sensitivity test with the exact framing and wording you intend to use in the app store listing, AI mention included. A test that prices an unlabeled hypothetical won't catch a tax that only shows up once your actual buyers see the word "AI" in your actual pitch.
Who should answer, and how many
The four questions only produce a usable range if the people answering them are your actual target buyers, not a general population sample and not your friends. A panel of people who match your buyer profile, verified rather than self-reported, is what makes the curves mean anything. For clean curves, aim for 100 to 300 respondents from your actual market. Fit still outranks count: forty well-matched respondents beat four hundred mismatched ones. Once you've got the panel set up, a verified panel typically delivers within 48 hours, which is worth knowing but shouldn't be the reason you run the study.
Cree un estudio, llegue a encuestados verificados en su mercado y obtenga respuestas en cuestión de días.
Todo lo que hace un equipo de investigación. Sin el equipo de investigación.
Con SegmentOS puedes diseñar el estudio, llegar a una audiencia verificada y obtener datos en los que realmente puedes confiar. De principio a fin, sin necesidad de tener experiencia en investigación.
Calidad de datos
Cada estudio pasa por la huella digital del dispositivo, detección de velocidad, controles de atención y descalificación del filtro.
Can you run a pricing study without a research agency?
Yes. The same Van Westendorp study an agency charges $5,000 or more for is four fixed questions and a panel of matched respondents, both of which you can run yourself for a few hundred dollars. The method isn't proprietary and it isn't complicated, it's just rarely built into a tool founders can run without a research background. What the agency really sells is the methodology and the sample; when those come pre-built, the price gap stops making sense for a small team.
If you want to run the exact study described above, describe your product, choose your panel, and get the price curves without writing the questions from scratch, you can do it here: Van Westendorp calculator.
Pricing a new product will never be a solved problem. But it doesn't have to be a guess either. Four questions, the right people, and a plain description of what you're actually selling will get you closer to the real number than any thread full of strangers guessing on your behalf.
Preguntas Frecuentes (FAQ)
Would you pay $19/month for an AI tool that creates freelance proposals?
Nobody outside your target buyers can answer that for you, and a single yes/no question will always undercount resistance. Run a Van Westendorp study with actual freelancers describing the exact product at $19/month, and you'll get a defensible price range instead of a guess.
How much should my sister and I charge for this service business?
You don't need a product to run this method, just a clear, specific description of what the client gets. Ask a sample of your actual target clients the four Van Westendorp questions against that description, and you'll get a range instead of picking a number out of the air.
What is the Van Westendorp pricing method?
It's a pricing research method that asks target buyers four questions about the same product: the price at which it would feel too cheap to trust, a bargain, expensive, and too expensive to consider. Plotted across a sample of 100 to 300 matched respondents, the intersecting curves give you an acceptable price range and an optimal price point instead of a guess.
Realistic revenue for a subscription app with per-use AI cost on Play Store in 2026?
Your per-use AI cost sets your floor, but your ceiling depends on whether buyers will pay more once they know it's AI-powered, and current data says many will pay less. Run your price sensitivity test with the exact framing you plan to launch with, AI mention included, so the number reflects the real market instead of a spreadsheet assumption.
How many respondents do you need for a Van Westendorp pricing study?
Aim for 100 to 300 respondents who match your actual buyer profile; that range produces clean, readable curves. Fit outranks count: forty well-matched respondents beat four hundred mismatched ones, because the method only measures willingness to pay among people who would actually buy.
