Pricing strategy

Pricing is three decisions, and they are usually made in the wrong order. Most teams start with the number. The number is the last decision and the least important one.

1. Value metric — what you charge for. Determines whether revenue grows with customer success or fights it.
2. Packaging — what is bundled, gated, and tiered. Determines who self-selects into which price.
3. Price level — the actual number. Easiest to change, and the one everyone starts with.

Work them in that order. A wrong value metric cannot be fixed by adjusting the price; it produces a business that has to fight its own customers for revenue.

1. The value metric

The right value metric satisfies three tests at once:

It scales with realized value. As the customer gets more out of the product, they pay more, and it feels fair rather than punitive. Seats fail this test for products where value concentrates in a few power users.

It is predictable enough to budget. A metric the buyer cannot forecast creates procurement friction out of proportion to the revenue it captures. Pure consumption pricing on a volatile workload is technically fair and commercially painful — which is why usage-based products so often ship a committed floor.

It is countable without argument. If you and the customer can disagree about the number at renewal, you have designed a recurring fight into the contract.

When these conflict, resolve toward predictability for enterprise buyers and toward value-scaling for self-serve. That single distinction explains most of the pricing divergence between otherwise similar products.

Candidate metrics worth evaluating explicitly: seats, active users, usage volume, outcomes delivered, capacity provisioned, entities managed, revenue processed. For each, ask what happens when the customer succeeds wildly, and what happens when they have a slow quarter. Both answers should be tolerable.

2. Packaging

Packaging decides who buys what, and it does more work than price level.

Tier on a dimension buyers can self-assess. A buyer must be able to look at the tiers and know which one they are, without a sales call. Tiers separated by feature lists nobody understands push every deal into a conversation, which is expensive at low ACV and is exactly why so many self-serve funnels underperform.

Gate on scale and risk, not on quality. Withholding SSO, audit logs, or permissions from lower tiers works because those needs correlate with willingness to pay. Withholding reliability or basic usability trains customers to distrust the product.

Three tiers is the default for a reason, not a rule. Two tiers under-serve the middle; five tiers make the choice hard. Add a fourth only when a genuinely distinct buyer exists, not to fill a grid.

Design the anchor deliberately. The top tier's job is often to make the middle tier look reasonable rather than to sell in volume. That is legitimate; just be aware of which tier is the anchor and which is the target.

3. Price level

Willingness to pay

Van Westendorp is the standard instrument. Four questions per respondent — at what price is this too cheap (you would doubt the quality), cheap (a bargain), expensive (you would have to think), too expensive (you would never buy) — then intersect the cumulative curves.

PYTHONPATH="${CLAUDE_PLUGIN_ROOT}" python3 -m gtmkit.pricing \
  --responses responses.csv

The CSV needs columns too_cheap, cheap, expensive, too_expensive.

The implementation validates monotonicity per respondent and drops rows where the four answers are not in ascending order — those respondents misread the question, and in real survey data they are routinely 5–15% of responses. Leaving them in is the most common reason two analysts get different answers from the same file. Rejected rows are listed rather than silently discarded.

What the output means, and what it does not. The acceptable band runs from the point of marginal cheapness to the point of marginal expensiveness. The optimal price point is where the fewest people reject the product on price in either direction. But all of it measures stated willingness to pay, absent a competitor, a budget cycle, or a procurement team. It bounds the decision. It does not make it.

Below roughly 50 usable responses the intersections move several percent with every added row, and the tool labels the result "indicative only". Respect that label — a price set from 22 responses has four decimal places of precision and none of the reliability.

Value-based anchoring

Where you have a business case, price against it. If the product delivers $300k of measurable annual value, a $60k price is a 5x return the buyer can articulate to their CFO. Pricing at 10–20% of quantified value is a defensible starting position for most B2B software, and it gives the seller a number to reason from rather than a competitor's pricing page.

This is why value-case and this skill belong together: the ROI model is also the pricing argument.

Competitive reference

Use competitor pricing to understand what buyers are anchored on, never to set your own number. You do not know their margins, their strategy, their discount reality, or whether their published price is what anyone pays. Copying a price copies a strategy you cannot see.

Price increases

The mechanics matter more than the percentage:

• Grandfather existing customers for a stated period, and say the period.
• Give the increase a reason tied to delivered value, not to costs. Nobody buys "our costs went up."
• Announce with enough notice for a budget cycle — 90 days minimum for enterprise.
• Expect and model churn. A price increase with zero churn was too small.

Discounting

A discount without a concession is a price cut with extra steps. Every discount should buy something: a longer term, an earlier payment, a reference, a case study, a broader deployment.

Publish the approval thresholds internally. Ad-hoc discounting does not just lose margin, it teaches the market what your real price is — and that number, not the published one, becomes the anchor for every subsequent negotiation.

Common failures

Charging for the wrong unit and papering over it with tiers. If the value metric is wrong, no amount of packaging fixes it.

Pricing from cost. Buyers do not care what it cost you to build.

One price for two segments. If enterprise and self-serve buy the same SKU at the same price, one of them is badly mispriced.

Never testing. Most companies that suspect they are underpriced are, and most never run the test. A price test on new logos only, in one segment, is low risk and settles the question.