Big CartelReverse Trial
How five experiments built the case for a trial-first acquisition model
At a glance
- Challenge
- Determine whether a reverse trial could outperform Big Cartel's existing freemium acquisition model without damaging overall conversion.
- Approach
- Led five experiments across the signup, trial, payment, onboarding, and pricing experiences, using each result to shape the next test.
- Outcome
- The combined evidence gave leadership the confidence to adopt a trial-first acquisition model for new signups.
- Role
- Principal Product Manager and product lead for Seller Acquisition
- Team
- Me, 1 engineer, marketing team, support, 1 designer
- Timeline
- Six months
Over six months, I led five experiments to determine whether a reverse trial could replace freemium as Big Cartel's primary acquisition model. No single test produced a perfect result. Together, they revealed which users were most likely to pay, when to collect payment information, and where the funnel was breaking. The evidence ultimately gave leadership the confidence to change the company-wide pricing model.
Big Cartel was the last major platform still running freemium as its primary acquisition motion. Every real competitor had moved to a trial-only model years earlier. Freemium had also hit a ceiling on its own terms: paid spend rose about 20% from November to December, and monthly new paid subscribers didn't move. The marginal dollar was buying almost nothing.
I developed the business case for testing whether trial-first could replace freemium, and for being honest about the answer even when it wasn't a clean win. It took five tests to get there.
The five experiments
A short orientation. The full story for each test is below.
| Test | Design | Result | Takeaway |
|---|---|---|---|
| Test 1 | Auto-enroll, no card, day-7 lock screen | 17% relative decline in paid conversion. Lock screen converted at ~20%, but fewer than 20% of the cohort ever reached it. | For a product with episodic usage, the intervention happened too late. Collect payment info at the moment of highest intent. |
| Test 2 | Card required at trial start | Trial slightly underperformed control in 30-day conversion to paid. Prepaid charges failed at 88%. | Payment failures ate conversion. Block prepaid cards and add retries over 7 days. |
| Test 3 | Retries + prepaid card blocking | Trial reached parity with control on 30-day conversion to paid. | Retries worked. Debit declines were mostly insufficient funds, a timing problem, not a dead card. Extend the retry window. |
| Test 4 | Platinum trial on the pricing page | Platinum sign-ups nearly tripled. Trial conversion held steady across plans. | The entry point shifted plan mix more than conversion. Concentrate choice by de-emphasizing Gold. |
| Test 5 | Gold demoted to a link on pricing page | Overall sign-up volume down ~33% due to fewer Gold sign-ups. Trial success rate 61% (trial start → paid). Volume parity with control on 30-day conversion to paid. | Removing Gold's visibility cost sign-up volume. Funnel leakage on the Platinum trial sign-up flow costs trial starts. |
Test 1: Auto-enroll, no card, day-7 lock screen
February
I went in with two hypotheses. First: exposure to premium features alone would drive payment attachment — sellers would use paid tools during the trial and want to keep them. False. Second: a hard lock screen at day 7 would force a decision. True, but with a large caveat.
Conversion on the lock screen itself ran around 20%, a solid number whenever someone actually hit it. The problem was getting people there. Sellers didn't come back on day 7. They trickled back over weeks, not days, and sellers who hadn't made a sale yet had little reason to hurry back and check on a shop with nothing happening in it. By the time enough of the cohort had actually seen the lock screen, the test's aggregate numbers read as a loss — even though the mechanism worked fine whenever it actually got in front of someone.
Learning: Collect payment information at the moment of highest intent, especially with price sensitive consumers.
A 17% relative decline.
Test 2: Card required at trial start
March
Test 1's lesson: collect payment info at the moment of highest intent — trial start, not seven days later. I designed the next test around an opt-out trial, invited straight from the welcome modal, and partnered with Design and Engineering to change every upgrade CTA across the product to read "start trial" instead of "upgrade," collecting card info right there.
It still lost, but not because people wouldn't start a trial. They did. It lost because a large share of those trials failed to convert to paid at the end for a fixable, unglamorous reason: payment failures at billing, concentrated in prepaid cards and insufficient funds, with no smart retry logic running on the declined charges.
Trial converted at 4.84% vs. Control at 5.31%.
Test 3: Retries and blocking prepaid cards
April
I kept the same design and partnered with Engineering to add four payment retries spread over 7 days. The retries worked, recovering enough failed payments to bring the test to parity with control on subscription starts.
Debit cards, still allowed, remained a large share of the base and still failed at a high rate — mostly insufficient funds rather than a dead card. That distinction mattered: a debit decline is often a timing problem, not a permanent one. That's the detail that led me to push the retry window from 7 days to 14 in the next test — more days meant more pay cycles for a paycheck or deposit to land before the card was tried again.
Test 4: Platinum trial on the pricing page
May
I designed this test with Design and Marketing to add the trial offer directly to the Platinum plan card on the marketing site. Bucketing happened at the marketing-site level across all traffic, so unlike earlier tests, this one measured overall subscription-start rate rather than Gold-specific conversion.
A cookie-consent and SDK issue complicated the read: the marketing site's experiment property didn't always make it into the app, especially once cookie-blocking tools automatically opted users out in the EU, California, and a few other US states. To keep the comparison clean, I partnered with Data to restrict the analysis to US-based sellers where the marketing-site and in-app experiment properties could be matched to each other.
The merchandising shift worked as intended: Platinum sign-ups nearly tripled. A decent share of sellers still chose Gold anyway. What mattered more than which plan someone picked at sign-up was what happened after — trial conversion held essentially steady no matter which door they came in through.
Test 5: Gold demoted to a link
June
This test pushed the pricing page further: Gold was demoted from a full plan card to a single line of text underneath Platinum and Diamond, with a free trial advertised on Platinum as the headline offer. Sellers who still chose Gold got the same trial offer and in-product experience as the prior two tests; sellers who chose Platinum went straight into trial checkout.
My hypothesis was direct: does Gold's pull as a genuinely free plan beat a trial offer, or does concentrating the choice on Platinum win out? Overall sign-up volume dropped about 33%, almost entirely from Gold sign-ups falling off — exactly the trade-off the team expected from removing Gold's visibility.
Trial success also slipped to 61% of trial starts converting to paid, down from the prior test, mostly on the same payment-failure pattern (debit declines, insufficient funds) that showed up in every test requiring a card.
The next chapter picks up directly from here: I recommended increasing trial-start rate through trial sign up funnel optimizations first, then trial-success rate.
Key Takeaways
The early experiments did not produce a clean conversion win, but they changed how I understood the problem. The opportunity was not simply to put more free users into a trial. It was to identify higher-intent users, remove avoidable payment friction, and give the company's marketing channels faster signals about acquisition quality. That shift in perspective shaped the recommendation I brought to leadership and our ultimate change to reverse trial as the primary acquisition motion.
Focus on high-intent users.
A trial start is a real commitment. A free sign-up isn't. Concentrating the funnel around people willing to start a trial means building onboarding and support for a smaller, more serious cohort instead of a much larger, mostly uninterested one.
A stronger, faster signal for paid marketing.
Trial starts happen at higher volume and much closer to day one than subscription starts do. That gives Google and Meta a bigger, faster event to optimize against than waiting on the rarer, more lagged “subscription started” signal — which should improve how efficiently paid spend finds the right people.
More case studies
Rebuilding Big Cartel's discount engine to drive merchant adoption and conversion.
An AI-powered system that surfaced product problems hidden in thousands of seller messages.
How lightweight fixes solved a duplicate-account spike without a multi-month rearchitecture.
A 17-skill AI system that processed 20,000+ pages of discovery and took the document-dump tactic off the table.
A photo and voice-first renovation punch list where AI does the filing and the app serves three tasks a day.
How solving a missing draft-product state and a decade of one-off themes shipped Big Cartel's strongest upgrade driver.