Most software companies set price once, during a late-night pre-launch discussion, and revisit it only when a board member asks. Meanwhile acquisition costs rise, the product doubles in capability, and the price stays where a founder guessed it two years ago.
Start with the value metric
The value metric is what the customer buys units of: seats, contacts, transactions, gigabytes, API calls, revenue processed. Get this right and pricing largely solves itself, because revenue grows automatically as customers succeed.
| Metric | Aligns with value? | Predictable for buyer? | Watch out for |
|---|---|---|---|
| Per seat | Sometimes | High | Password sharing; penalises collaboration |
| Per active user | Good | Medium | Definition disputes at renewal |
| Per transaction | Strong | Low | Bill shock in a spike month |
| Per contact / record | Good | High | Encourages data deletion |
| Percentage of revenue processed | Very strong | Medium | Resented as the customer scales |
| Flat platform fee | Weak | Very high | No expansion path |
The testAsk: "If a customer doubles their business using us, does our invoice grow?" If not, you have a flat fee wearing a metric costume, and every expansion dollar will require a renegotiation.
Model a price change before you ship it
The break-even churn figure is the one to internalise. A 15% increase survives roughly 13% customer loss before revenue declines — and real-world churn from a well-communicated increase is usually a small fraction of that.
Packaging: three tiers, three jobs
- Entry tier — qualify, don't monetiseIts job is to get the right customers in with a clear ceiling that creates natural upgrade pressure.
- Core tier — the one most should buyDesign everything around this tier. It should be the obvious choice for your primary segment and carry your best margin.
- Enterprise tier — capture the top of the marketSecurity, compliance, SSO, audit logs, SLAs and support. Price on request where deal sizes vary widely.
- Differentiate tiers by capability and scale, not by artificially removing basic functionality.
- Put administrative and compliance features high — that is what larger buyers pay for.
- Keep the tier count at three or four. Every extra option reduces conversion.
- Show annual pricing prominently; annual billing improves cash flow and reduces churn.
Research that beats intuition
Van Westendorp questions. Four questions — too cheap, cheap, expensive, too expensive — asked of 40+ target buyers produce a defensible acceptable price range in a week.
Willingness-to-pay by segment. The same product can be worth 5× more to one segment. If your pricing page shows one number to everyone, you are leaving that spread on the table.
Win/loss interviews. Ask lost prospects what they bought instead and why. "Too expensive" almost always means "value was not established", which is a positioning problem, not a pricing one.
Discounting disciplineEvery unstructured discount is a permanent revenue reduction and a signal that list price is fiction. If you must discount, exchange it for something: annual prepayment, a case study, a longer term, a reference call.
Executing a price increase
- Ship value firstAnnounce the increase alongside real improvements delivered in the preceding quarter.
- Give 60–90 days' noticeAdequate notice converts a surprise into a planned budget line.
- Grandfather with an expiryTwelve months at the old price for existing customers. Permanent grandfathering creates a legacy base that never funds the roadmap.
- Offer a lock-in pathLet customers prepay a year at the old rate. Cash now, goodwill preserved.
- Brief support with real answersEvery rep should be able to explain the rationale in one sentence and know the escalation path.
- Measure for 90 daysTrack churn, downgrades, expansion and win rate. Reversibility is a legitimate part of the plan.
The metrics that judge your pricing
Net revenue retention is the single most informative number. Above 100%, your existing base grows without new logos — the compounding engine every durable software business runs on. Below 90%, pricing and packaging are leaking value faster than sales can replace it.
Limits, errors and borders of the standard advice
Critical reviewPricing advice in software is dominated by consultants selling frameworks and by investors optimising for a growth pattern that suits their return profile rather than your business. The confident maxims are mostly extrapolations from a handful of famous companies.
"Charge more — you're underpricing."
The claimFounders systematically underprice, so raising prices is nearly always correct.
Where it breaksIt is directionally right often enough to be useful and it is stated as a law. The evidence base is heavily selected: we hear from companies that raised prices successfully, not from those that priced themselves out of a market where a cheaper competitor then captured the segment. Price also has to fit the acquisition channel — a high price usually requires a sales motion, and a company built on self-service growth may not be able to support one. In price-sensitive or commoditised categories, and in markets with strong free alternatives, the advice can be simply wrong.
The borderRaising prices is usually right when you sell to businesses, your product creates measurable financial value, churn is low and objections are rare. Be far more careful with consumers, in commodity categories, or where a free competitor is credible.
"Use value-based pricing, never cost-plus."
The claimPrice on the value delivered to the customer, not on your costs.
Where it breaksSound in principle and frequently unimplementable. Value varies enormously between customers and is usually unmeasurable in advance, so "value-based" in practice becomes a plausible story used to justify a number chosen for other reasons. It also ignores the constraint that competitor pricing sets a ceiling regardless of your value case, and that costs set a floor you cannot ignore — a value-priced product sold below its true delivery cost, which happens with usage-heavy infrastructure, fails despite perfect pricing theory.
The borderValue framing is essential for how you communicate and package. Use competitor pricing to bound the range, unit economics to set the floor, and value analysis to position within that range — not as a substitute for either constraint.
"Offer three tiers — good, better, best."
The claimThree tiers exploit anchoring and capture different willingness to pay.
Where it breaksThree is a convention borrowed from consumer retail psychology and applied indiscriminately. Tiers only capture willingness to pay if the differentiating dimension actually correlates with value received; when it does not, tiers merely confuse buyers and cap revenue from customers who would have paid more. Badly chosen limits also create adverse incentives — capping a feature that drives adoption suppresses the usage that creates value. Multiple tiers additionally multiply support, engineering and migration complexity, a cost that falls hardest on small teams.
The borderTiering works when you have identified a genuine value metric that scales with customer benefit. Start with one plan until you can name that metric from evidence — a single well-priced plan beats three arbitrary ones, and adding tiers later is easy.
"Never discount — it devalues your product."
The claimDiscounting trains customers to wait and erodes perceived value.
Where it breaksOverstated as an absolute. Discounts have legitimate uses: annual prepayment that improves cash flow and reduces churn, early-stage design partners who provide product direction, entry into a new market segment, and volume commitments. The purist position also produces worse outcomes than a disciplined policy, because sales teams under quota discount anyway — informally, inconsistently and without extracting anything in return. Undisciplined discounting is the actual problem, not discounting.
The borderDiscount against something you receive: a longer commitment, prepayment, a case study, a reference. Keep list prices stable and publish the discount rules internally so concessions are consistent — that preserves perceived value far better than a rule everyone quietly breaks.
Almost everything published about software pricing, including this, generalises from a small number of visible successes — the survivorship problem we criticise applies to our own reasoning too. We also give no benchmark figures for conversion, churn or price points, because they vary so much by segment that a benchmark invites the wrong comparison, which leaves you without the reference point you probably wanted. And our preference for simplicity suits small teams; a company with a mature sales organisation can execute complexity we advise against.
Frequently asked questions
Should we publish prices?
Publish for self-serve and mid-market — buyers increasingly disqualify vendors who hide pricing. Reserve "contact us" for enterprise tiers where scope genuinely varies. A hybrid page with three published tiers plus a custom option works for most.
How often should we revisit pricing?
Review quarterly, change at most annually. Continuous change erodes trust; never changing erodes margin as your product improves and costs rise.
Is usage-based pricing better than seats?
It aligns better with value and expands automatically, but introduces forecasting anxiety for buyers. Hybrid models — a platform fee plus usage — capture most of the benefit while keeping the invoice predictable.
What about a free tier?
Only if it has a defined conversion mechanism and a cost to serve you can absorb. A free tier that consumes support and infrastructure without producing paid conversion is a marketing expense you have not budgeted.
The bottom line
Pick a value metric that grows with customer success, package three tiers with a deliberate centre of gravity, research willingness to pay instead of guessing, and run price changes as a communicated programme rather than an email. Then watch net revenue retention — it will tell you within two quarters whether the pricing works.
Revision log. Jul 2026: added pricing impact model and expanded the increase playbook. Reviewed twice yearly.