Subscription vs One-Time Pricing: How to Test It
How to A/B test subscription vs one-time pricing: why first-purchase conversion misleads, and how to compare revenue fairly over time.
📚 This article is part of the guide How to A/B Test Pricing: The Complete Guide.
A/B testing a subscription against a one-time or lifetime-deal price for the same product is one of the trickiest experiments in the complete guide to A/B testing pricing, because the two offers are not different amounts of the same thing, they are different shapes of revenue entirely. This article covers what actually changes between the two arms of that test, why first-purchase conversion is the wrong metric to decide it (a one-time offer almost always wins there, regardless of which model is the better business decision), the subtlety that a subscription arm cannot be judged fairly until you have watched real churn play out, and a concrete framework, discounted cumulative revenue per customer, for comparing the two on the same footing.
What You Are Actually Testing
The surface question, “subscription or one-time,” hides three variables that a clean test needs to isolate:
- The payment shape itself. Recurring monthly or annual billing versus a single charge (a “lifetime” or “pay once” tier) for the same access.
- The price point chosen for each shape. A fair test does not compare an underpriced one-time offer to an overpriced subscription; both prices should reflect what the business needs from that customer over the horizon being tested.
- How the choice is presented. Whether both options sit side by side at checkout (letting the visitor self-select), or each visitor is shown only one shape as a single-arm test. Side-by-side tests measure preference and self-selection effects together; single-arm tests isolate the shape’s effect on conversion and revenue more cleanly, at the cost of needing two separate audiences.
Whichever design you choose, the test produces two kinds of customers that behave differently from day one, and comparing them fairly is the entire difficulty of this experiment.
Instrumentation matters as much as design here. Tag every converted customer with the offer variant they signed up under at the moment of purchase, not after the fact, so that revenue, refunds, and churn can be tracked back to the correct arm for as long as the test runs. Teams that only tag conversions and rely on billing exports to reconstruct the cohort later routinely lose the mapping once a subscriber upgrades, downgrades, or a one-time buyer is issued a partial refund, which quietly corrupts the exact comparison the test exists to make.
Why First-Purchase Conversion Is the Wrong Metric to Decide This Test
Here is the central point of this article: judging a subscription-vs-one-time test by first-purchase conversion alone almost always tells you to pick the one-time offer, regardless of which one actually makes the business more money. A single, bounded charge is simply an easier yes than an open-ended commitment, so it is standard, not a fluke, for a one-time or lifetime option to lift conversion the moment it appears at checkout.
Say your product currently converts 3.0% of checkout visitors under a subscription-only offer. You test adding a one-time/lifetime option alongside it, with 5,000 visitors analyzed on each side.
Subscription-only (A): 150 of 5,000 converted (3.0%). With the one-time option added (B): 210 of 5,000 converted (4.2%). Running those numbers through a two-proportion z-test gives a relative lift of +40.0%, z is approximately 3.22, and the p-value is approximately 0.0013, which is significant, with B winning clearly. Read on its own, this metric says “always offer the one-time option.”
Two-sided two-proportion z-test. "Not significant" almost always means not enough sample, not that the versions are equal.
The Subtlety: A Subscription Test Is Not Trustworthy Until You Have Seen Real Churn
Reaching significance on conversion is fast. Reaching a trustworthy answer about revenue is not, because a subscriber’s true cost to the business only reveals itself through how long they stay. Using the same 3.0% baseline and a minimum detectable effect of 15% relative (a difference worth acting on), the required sample is 24,193 visitors per variant. At 8,400 combined checkout visitors per week, that test reaches its required sample in roughly 41 days, under six weeks.
Forty-one days is enough traffic to call a winner on conversion. It is nowhere near enough time to know what that winner did to revenue, because the people who convert under a more persuasive one-time offer and the people who convert under a subscription do not behave the same way afterward, and that difference only shows up over months, not weeks.
Two-proportion normal approximation, traffic split evenly across variations. The date uses your timezone and updates live.
This is not a hypothetical concern. According to a case study compiled by Freemius (a SaaS licensing and payments platform), Jacob Kalvo, founder of the proxy service Live Proxies, compared cohorts of subscription customers against lifetime-deal buyers on the same product at the 90-day mark: subscription customers showed 78 to 82 percent retention, while lifetime-deal buyers showed only 45 to 50 percent active usage. The lesson Freemius draws from that comparison is that blending LTD and subscription cohorts into a single retention number hides which group is actually driving the business, since the gap between them is roughly double. The lesson for an A/B test is the same: a one-time buyer who stops using the product still counts as a “conversion” forever, while a subscriber who stops using the product eventually stops paying, and only one of those two facts is visible if you stop measuring at the point of purchase.
Retention analytics platform Churnkey describes a related pattern in healthy subscription cohorts, a “smile curve”: the retention line drops sharply in the first few months as casual or poor-fit users churn, then flattens as the surviving core stabilizes, and in the strongest businesses turns upward again as some of those churned users return to the product, a pattern Churnkey associates with network-effect businesses like Airbnb and Slack. None of that shape is visible from a snapshot taken at day 41, or even at day 90. Declare the comparison horizon (commonly 12 months for a first read) before the test starts, and hold the verdict until it closes.
A Framework for a Fair Comparison: Discounted Cumulative Revenue per Customer
The metric that resolves this fairly is cumulative revenue per starting customer, projected across a fixed horizon, and ideally discounted so that revenue arriving later counts for less than revenue in hand today. This discounted cash flow approach to lifetime value is the same one David Skok popularized for SaaS metrics at ForEntrepreneurs.com: instead of assuming a subscriber’s future payments are worth their full face value, each future month is worth a little less than the one before it, reflecting the time value of money and the risk that the future payment never arrives at all.
Take a subscription priced at $29/month against a one-time price of $299, with an illustrative 6% monthly churn rate (a mid-range assumption for the exercise, not a claimed industry figure) applied to the subscription cohort:
| Horizon | One-time revenue | Subscription revenue (undiscounted) | Subscription revenue (discounted, 10%/yr) |
|---|---|---|---|
| 12 months | $299 (paid on day one) | $238.11 | $227.53 |
| 18 months | $299 | ~$305 (crosses one-time here) | ~$287 |
| 24 months | $299 | $351.43 | $325.97 |
Two things follow from this table that a conversion-only read would never surface. First, the answer genuinely depends on the horizon: at 12 months the one-time price still leads even after all the extra conversions the subscription arm gave up; by 24 months the subscription has overtaken it, discounting included. Second, discounting narrows the gap without erasing it, because a dollar arriving in month 24 is worth noticeably less than a dollar in hand on day one, which is precisely why comparing raw, undiscounted totals overstates how much a subscription is really worth relative to an immediate payment.
Common Mistakes in Subscription vs One-Time Pricing Tests
| Mistake | Why it happens | Fix |
|---|---|---|
| Declaring a winner from first-purchase conversion alone | The conversion dashboard updates in weeks; revenue differences take months to appear | Set the comparison horizon before the test starts, and hold the verdict until it closes |
| Comparing undiscounted totals | Nominal dollars are easier to add up than discounted ones | Apply a discount rate to future subscription revenue, following a discounted cash flow LTV method, before comparing to an upfront price |
| Blending subscriber and one-time-buyer cohorts in the same retention chart | The two buyer types look similar in a signup table | Segment cohorts by offer type from day one, as recommended in the Freemius case study cited below |
| Picking an arbitrary horizon after seeing the data | A 12-month cut looks favorable to whichever offer is currently ahead | Commit to the horizon (and to discounting, if used) in the test’s written hypothesis, before launch |
Where a One-Time Price Tends to Fit, and Where Subscription Still Wins
The right shape is not universal, and the market has already run a large, informal experiment on this question through lifetime-deal marketplaces like AppSumo. According to reporting on AppSumo’s business (ppc.land, citing a disclosure by founder Noah Kagan), the platform’s partner revenue has fallen roughly 50% over a two-year span, after having generated more than $55 million in partner payouts historically. Consultant Chris Rempel, whose argument the same coverage summarizes, frames the structural problem plainly: a lifetime deal hands the software vendor a lump sum today while leaving it with indefinite support obligations, and no recurring revenue to fund ongoing development, server infrastructure, or customer service. Separately, the Freemius case-study roundup cited above documents refund rates around 16 to 17 percent on lifetime-deal marketplace sales, well above what most direct subscription checkouts see.
| Product shape | Tends to favor | Why |
|---|---|---|
| Low, mostly-fixed delivery cost (a static tool, a template library, a one-off utility) | One-time / lifetime price | Ongoing support and infrastructure cost per customer stays low, so a single payment does not leave the business exposed the way it would for a heavier product |
| Ongoing infrastructure or support cost per active user (hosted SaaS, anything with real-time compute, human support) | Subscription | Revenue needs to keep arriving for as long as the cost does; a one-time price collected once against a cost that recurs forever is the exact mismatch Rempel describes |
| Uncertain product-market fit, early-stage validation | One-time (short term), then re-test | Fast cash and fast signal on willingness to pay, at the cost of the deferred-obligation risk above; revisit the pricing model once retention data exists |
| Enterprise or team accounts with expansion potential (seats, usage tiers) | Subscription | Only a recurring relationship captures expansion revenue as usage grows, which a one-time price forecloses entirely |
Treat this table as a starting hypothesis, not a verdict: the entire point of this article is that the right shape for your specific product, price point, and support cost gets tested and measured over a real horizon, not assumed from a category.
Write the Hypothesis and the Horizon Before You Launch This Test
Like any other pricing experiment, a subscription-vs-one-time test invites “discovering” a winner by accident, usually by stopping at whichever metric moved first and looked good. Write down, before the test runs: which price points are being compared, what horizon will decide cumulative revenue (12 months is a reasonable default, 24 if your sales cycle or support cost profile calls for it), whether that comparison will be discounted and at what rate, and what the retention data needs to show for either offer to be declared the winner. That single piece of discipline is what turns “the one-time offer converted better” from a premature verdict into one data point inside a complete answer.
It is also worth instrumenting refunds and support-ticket volume by offer type from day one, not after the fact: the Freemius case data above shows real-world lifetime-deal cohorts carrying materially higher refund rates than subscription checkouts, and a test that only tracks revenue while ignoring that cost can call a false win.
Automate This in Donnu
The hard part of this test is not running it, it is refusing to call it early. Donnu A/B tracks conversion and revenue by variant for as long as the test needs to run, so a subscription arm can be compared against a one-time arm on the same cumulative-revenue basis once real churn has had time to show up, instead of settling for whichever offer converted faster in week one.
Start a 14-day free trial and design your next pricing test around the horizon that actually decides it. See also the complete guide to A/B testing pricing and freemium paywall experiments, which shares the same trap of judging a monetization test by the metric that moves first.
Read also: How to A/B Test Pricing: The Complete Guide · Freemium Paywall Experiments · Price Anchoring Experiments · Significance Calculator
References
- Freemius (Scott Murcott). SaaS Lifetime Deals: When to Run One & How to Structure It, case study comparisons including Jacob Kalvo (Live Proxies). freemius.com/blog/saas-lifetime-deals.
- ppc.land. AppSumo’s revenue crashes 50% as lifetime deal model faces existential crisis. ppc.land/appsumos-revenue-crashes-50-as-lifetime-deal-model-faces-existential-crisis.
- David Skok. What’s your TRUE customer lifetime value (LTV)? DCF provides the answer, ForEntrepreneurs.com. forentrepreneurs.com/ltv.
- Churnkey. Customer Retention Curve: Calculator, Definition, Examples. churnkey.co/blog/retention-curves.
Frequently asked questions
- Which metric should decide a subscription vs one-time pricing test?
- A pair, read together, never one alone: first-purchase conversion, and cumulative revenue per customer over a fixed horizon (12 months is a common starting point). A one-time or lifetime-deal offer almost always converts more people upfront, because it removes the recurring-commitment objection entirely. That says nothing about which offer produces more revenue per customer once you count what a subscriber keeps paying, and what a one-time buyer never pays again.
- Why does a one-time offer almost always convert more people than a subscription?
- Because it asks for a single decision instead of an ongoing one. A visitor comparing "$299 once" against "$29 a month, indefinitely" is weighing a bounded cost against an open-ended one, and the bounded option reads as lower-risk even when its expected total cost is similar or higher. This is a well-documented framing effect in pricing psychology, not evidence that the one-time offer is the better business outcome.
- How long does a subscription vs one-time pricing test need to run before the result is trustworthy?
- Long enough to observe the horizon you actually plan to compare, not just long enough for statistical significance on first-purchase conversion. A test can call a conversion winner in a few weeks while telling you nothing about 12-month retention, because churn among freshly converted subscribers takes months to show up. Ending the test at the conversion verdict is the single most common mistake in this category of experiment.
- What is "discounted cumulative revenue per customer" and why does it matter here?
- It is the total revenue a customer is expected to generate over a chosen horizon, adjusted so that a dollar received later is worth less than a dollar received today, following the discounted cash flow approach to SaaS lifetime value popularized by David Skok. Comparing a subscription to a one-time price without discounting overstates the subscription, because it treats revenue arriving in month 18 as equally valuable as revenue arriving on day one.
- Is a lifetime deal ever the financially smarter outcome of this test?
- Yes, over a short enough horizon or for a product with low ongoing delivery cost. Over a longer horizon, a subscription that survives multiple renewal cycles routinely overtakes a one-time price, which is exactly why lifetime-deal platforms have run into sustainability problems: the cash arrives once, but hosting, support, and further development do not stop.