The dilemmas
The playbook runs in order when you’re pricing from scratch. Real pricing thought rarely starts there — it starts stuck, on one specific question, usually one of these ten. Each gets the shortest resolution that actually resolves it, and a link to where it’s derived. If an answer surprises you, that’s the link to follow.
Trial, freemium, or nothing free at all?
Trial by default. Freemium only if you can name the mechanism and put numbers on it — free users mechanically recruit paid ones, or their content improves the product for payers, or the free tier beats your next-best marketing channel on cost per conversion. Most products pass none of these tests, and a free tier that passes none is a subsidy to non-buyers plus a price anchor at zero you’ll fight forever (Decision 5).
The worry that keeps people on freemium anyway: if nothing is free, how does anyone experience the product? But free-to-try and free-to-keep are different things, and conversion only needs the first. A trial is free-to-try with the price visible from day one. So is letting someone use the product before signup — real value in one anonymous session, then account, then payment (Decision 6). So is an interactive demo on sample data, a three-minute recorded walkthrough, or a no-questions refund policy that moves the risk to you without setting your reference price at $0. Nobody needs a permanent free plan to find out whether your product works.
Which revenue model is the right one?
Wrong question — and the feeling that a right answer exists is the distortion to fix. Subscription, pay-once, the Sketch model, credits, lifetime deals: these aren’t ranked options, they’re different configurations of the same arithmetic, and which configuration wins depends on two facts about your product: whether the value is ongoing or a one-time transformation, and whether your costs recur. A pay-once tool with AI inference in it is a subscription you forgot to charge for; a subscription on a job that finishes is churn with extra steps. Run your own numbers through each model — LTV, payback, the cash curve over 36 months — and the “right” model usually reveals itself as two viable ones and three that would quietly bankrupt you (the catalog, with when each dies).
Should I charge from day one, or wait until it’s “ready”?
Charge from day one. Price is information, and a free beta buys none of it — worse, it sets every early user’s reference price at zero, so your eventual launch price arrives as an infinite increase to your warmest audience. If the product is rough, keep the price real and discount it honestly: “founding customer price, first 50, locked for a year” gets you revenue, commitment-weighted feedback, and an anchor at the real number. The fake door works even earlier — real prices on a landing page before the product exists.
My users say it’s too expensive. Do I lower the price?
Check who’s complaining before touching the number. “Too expensive” from people whose alternative is free is a positioning readout, not a pricing one — you’re being compared to the wrong reference, or you’re talking to a segment whose alternative is cheap (Decision 1). Watch behavior instead of comments: if conversion is healthy and churn is low, the complainers aren’t your buyers. And remember the asymmetry — underpricing reads as a quality confession to the segment you actually want. Lower the price only when the right segment’s behavior says so, and even then consider a cheaper tier with a real fence before a cheaper product.
Seats or usage?
Neither, usually — hybrid. A flat base for predictability plus prepaid credits on the dimension that costs you money. Seats are honest only when value really is per-human; the moment your product does work instead of people, seat pricing meters your own value proposition backwards. Pure usage tracks cost but makes every unit hurt, so users self-ration away the habit your retention depends on. The four tests, in order: grows with customer value, predictable bill, grows with success not pain, cheap to measure and hard to game (Decision 3).
Monthly or annual?
Both, with the toggle defaulted to annual and a 15–20% discount — about two months. Annual floors churn at one decision a year (a hyperbolic LTV gain), collects the cash before you’ve paid to serve it, and hurts once instead of twelve times. Never annual-only at self-serve prices; monthly is the low-commitment door the wary walk through first. B2C is the exception to aggressive annual discounting — deep prepay harvests money from people who won’t use month three, and the refunds and resentment arrive on schedule (cross-cutting calls).
The competitor is cheaper. Should I undercut them?
No. Cheaper-same-job is the default instinct and the worst move: price signals quality, so the cheap clone reads as the knockoff, and you’ve opened a war that whoever can bleed longest wins — which is them, especially if they’re funded. Go premium in a narrower segment they serve shallowly, or reposition sideways so the comparison stops applying. If the rival is another bootstrapper, the real contest is distribution, not price (the competitor section).
Can I raise prices on existing customers?
Yes — with the fairness rules, because customers punish exploitation, not increases. A justification they can repeat (“costs rose” and “we shipped X” both work; “you’re locked in” never does), 60–90 days notice, and an escape valve: lock the current price by going annual. New customers are a separate case entirely — raise on them freely and continuously; they have no reference price with you (Decision 8, fairness).
Should I do a lifetime deal?
Only as launch financing, and only if your marginal cost rounds to zero. You’re selling a customer’s entire future for a few months of revenue — rational exactly once, in capped quantity, for cash and an initial user base. With AI COGS it’s a debt that compounds: every lifetime customer is a small negative annuity you can never cancel (the model catalog).
Should I ever discount?
Keep one public discount: annual prepay. Beyond that, discounts leak margin at your highest-intent moment and train customers to wait for the next one — a reflex that’s brutal to untrain. If a segment genuinely can’t pay your price, that’s a tier with a fence or regional pricing, not a coupon. At cancellation, one honest save-offer at most; a discount ladder there is a retention maze wearing a bow.
Pricing something end to end rather than un-sticking one call: the runbook.