Price Is the Profit Lever You Never Measure

You will run a dozen creative tests and a hundred bid adjustments this quarter and spend almost no time on the one number that moves profit more than any of them. A 1 percent price rise, with volume held constant, lifts operating profit by about 8 percent in McKinsey's analysis of a typical S&P 1500 company. Marn and Rosiello's original Harvard Business Review study put it at 11.1 percent, and both are right, they use different samples, so treat the number as directional. Yet the average company spends under 10 hours a year on pricing, and 72 percent of new products miss their financial targets, because price is decided last. The fix is not to raise the price or cut it, it is to measure willingness to pay and build your architecture around it, and to know whether you have a price problem or a value problem, because the cures are opposite.

Why price is the highest-leverage lever you control

The reason price beats every other lever is contribution-margin math. Win an extra unit of volume and you pay to acquire and serve it, so only the margin is profit. Raise price and hold volume and there is no incremental cost, so almost the entire increment drops to the bottom line. The comparison is the point, a 1 percent volume gain is worth only 3.3 percent of profit, a variable-cost cut 7.8 percent, and a fixed-cost cut 2.3 percent, all below price. The lever is symmetric, which discounting forgets. A careless 1 percent price cut vaporizes the same 8 to 11 percent of profit, and the pocket-price waterfall shows how dozens of small, invisible concessions compound into a realized price about half of list. The honest caveat is that this assumes no volume loss, which real elasticity can violate, so it is a reason to measure, not a promise. And it is the least-resourced lever there is, most of the growth budget flows to acquisition, yet a dedicated pricing owner is worth a 14 to 30 percent EBITDA improvement on Simon-Kucher's vendor data. The leverage is largest where the attention is smallest.

Why the job is to measure willingness to pay, not guess

The good news is that willingness to pay is measurable, with three workhorse methods. Van Westendorp's Price Sensitivity Meter asks four questions, at what price the product is too cheap to trust, cheap, expensive, and too expensive, and the intersections of the cumulative curves give an acceptable range and an optimal point, good for new products with no benchmark. Gabor-Granger asks purchase intent at a series of prices, builds a demand and revenue curve, and the revenue peak is your revenue-maximizing price. Choice-based conjoint shows realistic competing bundles with a walk-away option and derives the dollar value of each feature and price level, the indirect gold standard, at a sample of at least 300. The catch is that the method moves the answer, in one Conjointly illustration on a product a naive how-much-would-you-pay question averaged 31.84 dollars, van Westendorp put the range at 5.75 to 7.67, and Gabor-Granger a revenue-maximizing 13.89. And every survey overstates. Stated willingness to pay exceeds real by roughly 35 percent in one meta-analysis up to 200 percent in another, so treat survey numbers as a ceiling, discount them, and validate with a live price test.

Why the architecture matters more than the number

The highest-order pricing decision is not the price point, it is the value metric, what you charge for. Per seat, per contact, per transaction, per compute-hour, it can save your strategy even when the number is wrong, because it scales revenue with value delivered, the way Slack charges per active user. On top of the metric, good-better-best is proven architecture. A Good tier captures budget buyers who would not otherwise buy, Better retains the core, and a Best tier priced 40 to 100 percent above current harvests high-willingness-to-pay customers, often 30 to 60 percent of a suite's revenue, with a fence attribute strong enough to stop trade-down. The discipline is to stop there, the paradox of choice is real, the canonical jam experiment drew 60 percent of passers-by to a 24-jam display but only 3 percent to buy against 40 and 30 at a 6-jam display, and conversion falls past four tiers.

Why you must diagnose a price problem from a value problem

Before you touch the number, diagnose which problem you have, because the cures are opposite. A price problem means you are mispriced against willingness to pay, losing deals on price, discounting reflexively, or charging on the wrong metric, and the fix is architecture. A value problem means the product does not deliver or communicate enough value, and cutting price will not fix it, only destroy price integrity. The signals of an architecture and value-communication problem, not a demand ceiling, are chronic discounting and poor realization, a 32 percent implementation rate and less-than-half capture. And much of what sales records as lost on price is not. The JOLT Effect, on 2.5 million recorded sales conversations, found 40 to 60 percent of qualified B2B deals stall in no decision, 44 percent from status-quo preference and 56 percent from indecision and fear. That is a value-clarity failure a lower price does not solve. Control discounting tightly and treat a price cut as a last resort, because under 70 percent realization the problem is the message and packaging, not the number.

Why AI makes pricing cheap but cannot replace a live test

The analytical steps that used to need a consultant now cost a prompt, which makes this the moment to fix pricing. Have an LLM draft a van Westendorp or Gabor-Granger instrument, field it in your survey tool, then feed the responses back to compute the acceptable range or revenue-maximizing price, segmented by persona. It can also estimate elasticity from your historical price and volume, where an absolute value under 1 means underpriced, simulate good-better-best tiers, and classify win-loss notes as price-driven or value-driven. The limits are strict and the model must state them, survey analysis inherits the say-do gap, historical elasticity needs real price variation and is confounded by seasonality and promotions, conjoint needs a sample of at least 300, and over-optimizing invites fairness backlash, since personalized prices read as less fair than segment prices. No model output is a decision. Any AI optimal price is a hypothesis, and the only source of truth is a live price test, discount the stated willingness to pay, then A/B test two or three real prices on new-customer cohorts.

The pricing diagnosis (copy this)

Run this before your next creative sprint, because it outranks it on profit.

• Run a willingness-to-pay baseline. Field a van Westendorp and Gabor-Granger survey to a few hundred target-segment respondents and have an LLM analyze it. If your price sits below the acceptable range or under the revenue peak, you are underpriced, raise it on new customers first. If no one ever complains about your price, you are too cheap.
• Fix the value metric before the number. If you charge per seat but value scales with usage or outcomes, migrating the metric beats any price tweak, and a base-plus-usage hybrid is the lowest-risk entry.
• Install good-better-best with a real fence, capped at three or four tiers. Add a Best tier priced 40 to 100 percent above current to anchor and harvest high-willingness-to-pay buyers, and kill any tier past the fourth unless it converts.
• Diagnose before discounting. Instrument win-loss and discount data, and if you are realizing under about 70 percent of list or discounting reflexively, change the message and packaging, not the price.
• Validate every survey or AI output with one live price test. Discount the stated willingness to pay, then A/B test two or three real prices on new-customer cohorts. The live test is the only truth.

The full workflow, designing and analyzing van Westendorp and Gabor-Granger surveys, modeling elasticity from your historical data, simulating good-better-best tiers, and diagnosing price versus value from win-loss data, is packaged as a reusable Claude skill. Get the free skill.

What to do Monday

Put price on the same footing as everything else you optimize. Field a willingness-to-pay survey to a few hundred people in your core segment this week, have an LLM turn it into an acceptable range, and check your price against it. If you are capturing well under list or discounting on reflex, treat it as a value-communication problem and fix the message and packaging before the number. Then design three or four tiers around the value metric with a real fence, and prove the new price with a live A/B test on new customers. The lever that moves profit most is the one you leave untouched, and it is measurable, so measure it.

Sources: McKinsey on the profit leverage of price in the S&P 1500, and the pocket-price waterfall; Marn and Rosiello, Managing Price, Gaining Profit, Harvard Business Review, 1992; ProfitWell and Price Intelligently, Patrick Campbell, on pricing time-on-task and monetization; Simon-Kucher Global Pricing Study; Ramanujam and Tacke, Monetizing Innovation; van Westendorp, Gabor-Granger and choice-based conjoint references, and a Conjointly method comparison; Murphy and List meta-analyses on hypothetical bias; Rafi Mohammed on good-better-best, HBR 2018, and Iyengar and Lepper on choice overload; Dixon and McKenna, The JOLT Effect.