Willingness to pay in SaaS: what it is and how to calculate it

Willingness to pay is the highest price a buyer will pay for a product before they walk away. Every pricing book tells you to price against it. Very few SaaS teams know what theirs is.

The usual advice is to ask. Run a survey, plot the curves, read off a number. I’ve done that, and early on at Toggl it went badly. We ran a Van Westendorp study for a new product, didn’t like the answer, and doubled the “optimal” price it calculated. The doubled price performed far better than the research said it would.

That stuck with me. Willingness to pay is real and it’s the right thing to price against. Asking people for it is the unreliable part. Surveys are good at ranking things: which feature matters, which tier a segment belongs in. They’re bad at the level, the actual number on your pricing page. The level comes from what buyers do when money is on the line.

This guide covers what willingness to pay means in economics and in B2B software, how to calculate a price from it with a worked table, every common research method with a verdict and the order I’d use them in.

What is willingness to pay?

Willingness to pay (WTP) is the maximum price a buyer will pay for a product or service. Economists also call it the reservation price. Below it, the buyer purchases. Above it, they go without or pick an alternative.

Every buyer has their own number, so across a market willingness to pay is a distribution, not a single figure. One company would pay $12,000 a year for your tool, another $1,200. Setting a price means picking a spot on that distribution and accepting who you lose below it.

Willingness to pay in economics

In economics, sorting buyers from highest to lowest willingness to pay gives you the demand curve. Alfred Marshall named the gap between what a buyer would pay and what they actually pay consumer surplus:

Consumer surplus = willingness to pay − price

If a customer would pay $9,000 and you charge $5,000, they keep $4,000 of surplus. Pricing strategy is mostly about how much of that surplus you capture and how much you leave so buyers feel they got a good deal.

Willingness to pay also depends on context. In a 1985 study Richard Thaler asked people on a beach what they’d pay for a cold beer a friend was fetching. When the beer came from a fancy resort hotel, the median answer was $2.65. From a run-down grocery store, $1.50. Same beer, same beach.

In software, the store is the budget line the buyer files you under. The same product sold as a developer utility and as compliance tooling gets two different ceilings.

Willingness to pay vs willingness to accept vs willingness to sell

Three terms get mixed up:

Willingness to accept runs well above willingness to pay for the same thing. People value what they already own more than what they could buy, which economists call the endowment effect. Meta-analyses, including Tuncel and Hammitt (2014), find the gap in study after study. It’s smaller for ordinary goods than for public goods like parks or clean air, but it’s always there.

That gap explains a familiar SaaS pattern. A prospect shrugs at a price that an existing customer will fight when you take it away from them. Downgrades, removed features and price increases on the installed base all trigger willingness to accept, not willingness to pay.

Willingness to sell sits at the bottom of what Harvard’s Felix Oberholzer-Gee calls the value stick. Price sits somewhere between willingness to sell and willingness to pay, and value-based pricing is about pushing the top of the stick up rather than adding a markup to the bottom.

What determines willingness to pay for B2B software?

Most lists of willingness to pay factors are written for consumers: income, age, weather, how trendy the product is. A company buying software doesn’t care about the weather. Five things set its ceiling.

Next best alternative

A B2B buyer’s willingness to pay is anchored on the next best alternative: a competitor, the incumbent tool, a spreadsheet and an intern, a contractor, a new hire or doing nothing. Tom Nagle’s economic value estimation, from The Strategy and Tactics of Pricing, puts it as a sum:

Economic value = reference value (the price of the next best alternative) + differentiation value (what you do better, in money)

That’s the theoretical ceiling. Change the alternative and you change the ceiling. An AI tool compared to a $20-a-month seat licence has a seat-licence ceiling. The same tool compared to a part-time contractor has a contractor ceiling.

Whose budget it is and who signs

A consumer is the user, the evaluator and the payer. In B2B those are different people. Gartner puts a typical buying group at 6 to 10 people. Forrester’s 2024 State of Business Buying put it at 13.

Each of them has a different willingness to pay. The user wants the product and spends none of their own money. The budget holder compares you to everything else on their budget. Procurement gets rewarded for taking a percentage off. Security’s willingness to pay is zero until you pass their review, whatever the product does.

Approval limits matter too. I’ve seen B2B deals bunch up just under signing thresholds. I still wouldn’t price to them. Price the product what it needs to be priced, not what slips past a manager’s limit.

Company size and use case

Willingness to pay for the same software varies a lot between segments, and not always the way you’d guess.

One productivity SaaS company I worked with had two tiers. The basic tier got data into the product. The advanced tier added data governance, analytics and reporting. The thesis was simple: the bigger the company, the more it needs the advanced tier.

When we surveyed customers, the result looked different:

• Small companies wanted the cheap tier. They had too little data to need governance or deep analytics.
• Mid-sized companies (20 to 200 people) needed the advanced tier. They had enough data to need governance and analytics and no internal tooling for it.
• Large companies (200+, including 1,000+) went back to the cheap tier. They already had governance and analytics tools in house.

Willingness to pay for the advanced tier followed a bell curve across company size, not a line going up. An “average willingness to pay” across all three groups would have described nobody.

Value experienced vs value promised

A prospect prices a promise. A customer prices something they use every day. Before buying, the buyer hasn’t seen the time saved or the workflow it replaces, so they guess, and they guess conservatively. After six months of use, removing the product costs them something real.

That’s one reason survey answers from prospects come in low, and why expansion revenue from existing customers is easier to win than the same dollars from new ones.

How predictable the bill is

Buyers pay extra to avoid surprises. Anja Lambrecht and Bernd Skiera (2006) studied internet customers who chose flat-rate plans that cost them more than pay-per-use would have. They called it the flat-rate bias and traced part of it to an insurance effect: people pay a premium to make the bill predictable.

So willingness to pay depends on what you charge for as well as how much. Willingness to pay per seat, per credit and per API call are three different numbers. In my experience B2B buyers accept variable bills more readily than consumers, because over the long run a business can come out ahead, and a seasonal business often prefers them. But a variable bill still costs you some willingness to pay with buyers who need to forecast.

How to calculate willingness to pay

There is no formula that gives you one buyer’s willingness to pay. You estimate the distribution across your buyers, then calculate the price from it. The second part is arithmetic.

Build a willingness to pay table

Economics textbooks teach this with a willingness to pay table. Gregory Mankiw’s Principles of Economics uses four collectors bidding for an Elvis album at $100, $80, $70 and $50.

Here is the same idea for a B2B SaaS product. The numbers are illustrative: ten prospective customers and the most each would pay per year.

Total willingness to pay across all ten is $52,200. No single price gets you all of it.

Turn the table into a demand curve and a price

A buyer purchases if the price is at or below their willingness to pay. Count the buyers at each price and you have a demand curve. Multiply by price and you have revenue. Subtract the cost of serving each customer and you have profit.

Say each customer costs $1,500 a year to serve: model inference, infrastructure and support.

Three things fall out of this table:

• Revenue peaks at $5,000. Five buyers, $25,000. Drop to $4,000 and you add a customer but give up $1,000 from each of the five who would have paid more.
• Profit peaks higher, at $7,800. Once each customer costs real money to serve, the cheaper customers stop being worth it. This is why AI products with meaningful inference costs should usually price above the revenue-maximising point, while classic SaaS with near-zero marginal cost can price closer to it.
• One price leaves a lot on the table. At $5,000 you capture $25,000 of the $52,200. That’s 48%. The rest is consumer surplus with the high-WTP buyers ($14,800) plus the five buyers you lose entirely.

Capture more with tiers

Now offer two tiers instead of one: $7,800 and $4,000. If the expensive tier holds something A, B and C need and the cheap tier doesn’t, they buy the expensive one and D, E and F buy the cheap one.

Revenue: 3 × $7,800 + 3 × $4,000 = $35,400. That’s 68% of total willingness to pay, against 48% with one price.

The “if” carries the whole result. If buyer A can get what they need from the $4,000 tier, they’ll buy it, and you’ve just discounted your best customer. The fence between tiers is packaging and the value metric. That’s why most willingness to pay work ends up being a packaging question, and why good-better-best tiers built around what different customers need beat feature lists split at random.

Willingness to pay formulas you’ll run into

• Consumer surplus = willingness to pay − price. The buyer’s gain from the deal.
• Revenue = price × number of buyers whose willingness to pay ≥ price. The calculation above.
• Feature willingness to pay = part-worth utility ÷ −price coefficient. The standard way to put a dollar value on a feature from conjoint analysis. Greg Allenby and colleagues (2014, Quantitative Marketing and Economics) showed it overstates what a feature is worth to the company selling it, because it ignores how competitors respond. Treat it as a ranking, not a price.

Rules of thumb like “price at 80% of willingness to pay” have no research behind them. The right share depends on how the distribution looks and what it costs to serve each buyer, which is what the table is for.

Pricing research methods for willingness to pay

Every method fills in that table a different way. Stated methods ask people what they’d do. Revealed methods watch what they actually did. Stated methods are good at ranking and bad at the level. Revealed methods give you the level but need volume.

Van Westendorp price sensitivity meter

Peter van Westendorp published the method in 1976. Respondents give four prices: too cheap to trust, a bargain, getting expensive and too expensive. Where the cumulative curves cross, you get a range and a point the method calls optimal.

It’s cheap and fast and it hands you a chart, which is why teams keep using it. At Toggl, doubling the price it called optimal worked far better than the research predicted, which is why I don’t trust the chart. I’ve written up the evidence and eight reasons the number misses in why Van Westendorp accuracy is worse than founders think.

Gabor-Granger pricing

André Gabor and Clive Granger developed this method in the 1960s (Granger later won a Nobel Prize in economics for unrelated work). Respondents see a product and a price and say whether they’d buy. Then the price moves up or down and they answer again.

A typical ladder looks like this:

1. “Would you buy this plan at $49 per user per month?”
2. If yes: “At $59?” If no: “At $39?”
3. Repeat until the answer flips.

The share saying yes at each price is your demand curve. Multiply by price and you get a revenue curve with a peak.

The weaknesses:

• Anchoring. The first price shown pulls every later answer toward it.
• No alternatives on screen. The respondent judges your price in a vacuum, which real buyers never do.
• Direction of error is unclear. Kristen Berman of Irrational Labs (2024) notes newer research suggesting price-list methods like this may underestimate willingness to pay, while hypothetical questions in general tend to overstate it.

Gabor-Granger fits best on self-serve plans where one person decides with their own card and you can get a large sample per segment. For a sales-led product with a buying committee, it’s a guess with a curve attached.

Conjoint analysis and MaxDiff

Conjoint shows respondents a few product bundles that differ in features and price and asks them to choose. MaxDiff shows a set of features and asks for the most and least important. Both measure trade-offs instead of asking for a number.

People are good at comparing. When Miller, Hofstetter, Krohmer and Zhang (2011, Journal of Marketing Research) tested methods against real purchases, standard conjoint overstated willingness to pay but still got the ranking of attributes right.

That makes conjoint and MaxDiff the right tools for packaging: what goes in which tier, which features are add-ons and which segment values what. The productivity SaaS finding above, where mid-sized companies wanted the advanced tier and small and large ones didn’t, came out of a customer survey. That’s exactly the kind of answer surveys are good at.

Direct questions and WTP interviews

“What’s the most you’d pay for this?” is the weakest question in pricing. The buyer has nothing at stake, no alternatives in front of them and a mild reason to answer low if they think you’ll use it.

Interviews still earn their place if you change the questions. Ask about the past, which people can answer accurately:

• What are you paying for this today, in tools, hours or headcount?
• What did you replace last time, and why?
• Who signed off, and what’s their approval limit?
• What got cut in your last budget review?

Those answers tell you the alternative, the budget line and the buying process. That’s most of what you need to choose a value metric and positioning. Madhavan Ramanujam and Georg Tacke’s Monetizing Innovation (2016) makes the case for having these conversations before you build.

Auctions and BDM

The lab gold standard makes people put money on the line. In a Becker-DeGroot-Marschak (BDM) setup, a respondent states a price, a random price is drawn, and if theirs is higher they must buy at the drawn price. Lying doesn’t pay, so stated and real willingness to pay match.

Miller and colleagues found incentive-aligned methods like this matched real purchases where hypothetical ones didn’t. You can’t run it on B2B software. No procurement team will bind a company budget to a research lottery.

Revealed preference: what customers already pay

Revealed preference means reading willingness to pay from behaviour:

• Win/loss. Why deals were lost, and how often price was the reason.
• Discount depth by segment. Where reps keep cutting, list price is too high for that segment. Where they never cut, it may be too low.
• Upgrades and plan mix. What customers move up for, and which tier everyone piles into.
• Churn and contraction after a price change. What the market tolerated.
• Price tests on new customers. A new price for a cohort or segment, with real money changing hands.

This is the only source of the price level I trust. The catch is volume. A self-serve product with thousands of signups a month gets clean answers. A sales-led company closing 50 deals a year gets directional ones, and has to separate price from the rep who negotiated it.

How far off are willingness to pay surveys?

Usually too high, sometimes too low, and in B2B nobody has measured it properly.

• Jonas Schmidt and Tammo Bijmolt (2020, Journal of the Academy of Marketing Science) pooled 77 studies and found hypothetical willingness to pay ran about 21% above real willingness to pay on average. The gap was bigger for expensive and specialty products, and bigger for indirect methods like conjoint than for direct questions.
• James Murphy and colleagues (2005) found a median ratio of 1.35 between hypothetical and real values across 28 studies, with a long tail of much worse results.

Every one of those studies is about consumer goods or public goods. B2B software adds a buying committee, a recurring commitment and buyers who know their answers may become your price list. The error can go either way, which means you can’t correct for it with a haircut.

How to measure willingness to pay in B2B SaaS

Here’s the order I’d use. It covers the full willingness to pay research methodology, from segments to a live price.

1. Segment first. Willingness to pay is a distribution per segment, and segments can surprise you. In the productivity SaaS case above, demand for the advanced tier peaked with mid-sized companies, not the largest ones. Split by company size, use case and buyer type before you measure anything.
2. Decide what you charge for before how much. Per seat, per credit, per workspace, per output. Willingness to pay per unit differs by unit, and the wrong unit caps the whole curve. I’ve covered how to choose it in finding your SaaS value metric.
3. Rank with research. Conjoint, MaxDiff and interviews about past purchases tell you which features drive willingness to pay for which segment. Use that to build tiers and the fences between them.
4. Read the level from data you already own. The signal I trust most in Potio’s pricing sprints is win/loss. If price isn’t the reason for roughly 20% of the deals you lose, you’re underpriced. If nobody complains your price is too high, it’s too low.
5. Test the level on people who have to pay. New-customer cohorts, a segment, a new tier. Read the results the way you’d read a willingness to pay table: who bought, who didn’t and what you earned at each price.

Most teams run this backwards. They start with a price survey, set a number, and only later find out the value metric or the tiers were the real problem.

How to increase willingness to pay

In B2B software, willingness to pay moves when the alternative, the proof, the risk or the packaging changes.

• Go more vertical. The more vertical your product, the closer to value you can price and the more you can charge per customer. A horizontal product has to price on a metric that suits a wide set of buyers, which pushes it back toward infrastructure-style pricing and lower revenue per deal. Industry and product type set limits here too.
• Change the alternative you’re compared to. Positioned next to a seat licence, you get a seat-licence ceiling. Positioned next to a hire, a contractor or a compliance fine, you get a bigger one. Your positioning decides which budget line the buyer files you under.
• Put the value in the buyer’s numbers. The budget holder has to repeat your case to their boss. Hours saved, revenue recovered or headcount avoided, in their figures, raises the ceiling they can defend.
• Lower the risk. Trials, annual plans with a way out and predictable bills all reduce the premium buyers mentally charge you for uncertainty.
• Add offerings. With one tier, raising willingness to pay means dragging every customer up at once. More tiers or add-ons let you slice the market and give each slice a concrete offer near its own willingness to pay. In the ten-buyer example, two fenced tiers captured 68% of total willingness to pay against 48% for a single price.
• Get customers to value faster. Willingness to pay rises once the product is part of someone’s week. Onboarding is a pricing lever.

Conclusion

Willingness to pay is the right target for SaaS pricing and a bad survey question.

Treat it as a distribution, calculate the price from that distribution, and check what each customer costs to serve before you pick the revenue-maximising point. Use surveys and interviews for what they’re good at: segments, tiers and value metrics. Then get the level from behaviour, starting with your win/loss data and ending with prices tested on people who have to pay.

Frequently asked questions

What is the willingness to pay formula?

There’s no formula for an individual buyer’s willingness to pay. It has to be estimated from research or behaviour. Once you have estimates for a group of buyers, the formulas are simple: consumer surplus equals willingness to pay minus price, and revenue at any price equals that price times the number of buyers whose willingness to pay is at or above it. The revenue-maximising price is the one where that product peaks. If each customer costs money to serve, subtract that cost per customer and find the profit peak instead, which is usually a higher price.

What is an example of willingness to pay?

A company that would pay up to $9,000 a year for a SaaS tool has a willingness to pay of $9,000. If the tool costs $5,000, the company buys and keeps $4,000 of consumer surplus. If it costs $10,000, the company doesn’t buy. Across a market, willingness to pay varies widely: one SaaS company found demand for its advanced data governance tier peaked among companies with 20 to 200 people, while both smaller and larger companies preferred the basic tier.

Is willingness to pay the same as price elasticity?

No. Willingness to pay is a level: the most each buyer will pay. Price elasticity of demand measures how much demand changes when price changes, calculated as the percentage change in quantity divided by the percentage change in price. Elasticity comes from the shape of the willingness to pay distribution. If many buyers have willingness to pay just above your price, a small increase loses a lot of them and demand is elastic. If willingness to pay is spread far above your price, demand is inelastic.

Is willingness to pay the same as demand?

Willingness to pay is the building block of demand. Sort every buyer’s willingness to pay from highest to lowest and you get the demand curve: at each price, demand is the number of buyers whose willingness to pay is at or above it. One buyer has a willingness to pay. A market has a demand curve.

How many respondents do you need for a willingness to pay study?

Enough per segment to compare segments, because willingness to pay is a distribution per segment and an average across segments can describe nobody. Consumer studies often run hundreds or thousands of respondents. B2B studies rarely can, and a few dozen responses from one customer profile give confidence intervals too wide to set a price from. With a small B2B sample, use the research for ranking (features, tiers, value metrics) and get the price level from win/loss, discount and pricing test data.

How does willingness to pay work for usage-based and AI pricing?

With usage-based, credit-based or AI pricing, willingness to pay is per unit: per credit, per API call, per output. It also depends on how predictable the bill is, since buyers pay a premium to avoid surprises (the flat-rate bias, Lambrecht and Skiera, 2006). AI products also have real cost per customer, from model inference. That pushes the profit-maximising price above the revenue-maximising one, because low-willingness-to-pay customers can cost more to serve than they pay.