Everlove 在 2.4 万用户上测试缩短引导流程,完成率升 8.6 个百分点但试用率降 18%
Everlove tested shorter onboarding on 24,000 users. Completion went up, trial starts went down.
情侣应用 Everlove 在 2.4 万名用户上把 28 步的新用户引导与 19 步版本做 A/B 测试,引导完成率从 78.1% 升到 86.7%,多出约 700 人看到付费墙。但看到付费墙者的试用启动率从 3.83% 降到 3.13%,初始转化率 3.35% 对 3.11% 无显著差异,两项效果相互抵消。测试从 2026 年 6 月 14 日开始,第二伴侣的 12 步引导未改动。
情侣应用 Everlove 把新用户引导从 28 步砍到 19 步,完成率涨 8.6 个百分点,但看到付费墙的人试用率降 18%,最终转化没动。

Everlove, a couples app with 150,000+ users, A/B tested its 28-step onboarding against a 19-step version across 24,000 users. The shorter flow lifted onboarding completion by 8.6 percentage points and put roughly 700 more users in front of the paywall. But trial starts among paywall viewers fell 18%, and initial conversion didn’t move: the two effects cancelled out.
The takeaway: onboarding length isn't the first thing to decide. Work out where your funnel leaks before you decide how long onboarding should be.
Here's a look at what we tested, and what shorter onboarding actually changed for Everlove.
Key findings
- Cutting onboarding from 28 to 19 steps raised onboarding completion by 8.6 percentage points (78.1% to 86.7%)
- The shorter flow put roughly 700 more users in front of the paywall, despite a smaller cohort
- Trial starts among paywall viewers fell 18% (3.83% to 3.13%)
- Initial conversion showed no significant difference (3.35% vs. 3.11%)
- Find where your funnel loses people before you decide how long onboarding should be
A test was designed to settle the onboarding length question
The prevailing onboarding advice has swung back and forth over the years, but is currently heavily leaning toward longer onboarding being better. You only have to look at the success of Noom’s 113-step onboarding. Ask more questions, build more investment, let the user commit before you make the offer. The reasoning is sound, and there's real evidence behind it: a user who has spent four minutes telling you about themselves is more invested than one who spent 40 seconds.
Everlove, a subscription app for couples, had built exactly that flow. But it wasn't working. So we tested it against something much shorter.
Interestingly, the result came back split. The shorter flow won the top of the funnel decisively. The longer flow won at the paywall. Net conversion was flat.
So the question became not which approach is better, but which one their funnel was stuck on at that moment.
Why we tested it
Everlove gives couples 15 minutes and something to talk about. Pick a topic, pick how deep you want to go, and the app serves up questions, quizzes and games about the things partners usually skip.
The onboarding was long, but length itself wasn't necessarily the problem. The flow included a substantial number of ‘About You’ screens focused on the individual user, alongside heavier questions about their relationship and what was or wasn't working.
The hypothesis was simple: were these additional questions necessary to deliver the Everlove experience, or were they adding friction before users had experienced any value?
Two things made the onboarding worth testing:
1. Too heavy, too early: a long series of questions about what's broken in your relationship — asked before you've played a single game together — heavily front-loads emotional effort at the funnel's most fragile point.
2. A two-person product was spending a lot of time on one person: the flow included an extensive ‘About You’ section focused on person one, while Everlove's value depends on both partners joining and playing together. Users who bring their partner in convert at roughly 2x the rate of solo users. But every additional screen spent profiling one partner delays the behavior most associated with conversion.
What we tested
We tested the existing long onboarding against a much shorter version. For the initiating partner, the flow dropped from 28 steps to 19, removing questions from the heavier ‘problems in your relationship’ and individual ‘About You’ sections to get users to the core experience and paywall sooner. The second partner's onboarding remained unchanged at 12 steps.
- We measured success with: PostHog for onboarding start through to trial start or direct purchase, and RevenueCat for revenue, plan mix, retention
- The test audience scope was: Worldwide — US volumes were too small to reach significance below the top of the funnel
- Our test window was: from June 14 2026 onwards, to the point where both variants became equally exposed to a concurrent paywall test; onboarding length is the only difference between the two arms
The outcome
The shorter flow won activation, definitively. Onboarding completion rose 8.6 percentage points — an 11% relative lift, statistically significant, and it holds in the US too (+5.3pp) where almost nothing else reaches significance. Despite a cohort that was 581 users smaller, the shorter flow put roughly 700 more users in front of the paywall.
The longer flow won the paywall. Trial start rate among people who reached the paywall fell from 3.83% to 3.13% — an 18% relative drop, and the one clearly significant negative in the set. This is the commitment argument showing up exactly where its advocates say it should: users who invested more were likelier to buy when asked.
Initial conversion showed no significant difference. 3.35% against 3.11%, comfortably inside the noise, because those two effects roughly cancel each other out.
If conversion per viewer had merely held, the reach gain alone would have produced roughly 444 conversions against 412. But instead, the rate fell far enough to absorb it.
Metric |
Long onboarding |
Short onboarding |
Delta |
Significance |
|---|---|---|---|---|
Onboarding completion rate (of users who started onboarding) |
78.10% |
86.70% |
+8.60pp |
Strong |
Paywall reach rate (of registered users) |
74.26% |
83.92% |
+9.66pp |
Strong |
Trial start rate (of paywall viewers) |
3.83% |
3.13% |
-0.70pp |
Strong |
Initial conversion rate (of registered users) |
3.35% |
3.11% |
-0.24pp |
None |
Trial conversion rate (of trials started) |
14.29% |
15.91% |
+1.62pp |
None |
Average days to subscription |
4.01 |
3.58 |
-0.43 days |
None |
Average LTV per registered user |
$0.1163 |
$0.1592 |
+36.9% |
Weak (p=0.093) |
Partner-joined rate |
31.26% |
31.98% |
+0.72pp |
None |
What we'll do next
Here's what we're doing next. Each change can be applied to any funnel.
Test the paywall, not onboarding again: the paywall now gets around 700 more users per window, and those users are less committed than the ones it used to get. The next opportunity is to make the paywall work harder: reinforce the value and address objections that the shortened onboarding no longer handles.
Put the removed questions back later, not never: the questions were fine. Asking them early was the problem. Asked after someone has used the product, they should still make the experience personal without costing users at the start. We'll test that on its own.
Look at what people bought, not just how many users bought: the short flow may be producing fewer buyers who are each worth more. If that's true, it will show up in which plan they choose. Re-running plan mix will tell us whether that explains the revenue difference.
Run this on your own funnel
In this test, the longer onboarding built commitment: it won on trial starts among people who saw the paywall. But it also cost onboarding completions. So length isn't the thing to decide first — you need to understand where you’re losing people in the funnel, and then adjust onboarding to counteract that.
Here’s what I’d suggest:
- Find where you lose the most people. Onboarding, or the paywall. Answer that before you touch length.
- If you're losing them in onboarding, cut it. Shortening buys real volume, and that's the effect we measured.
- If your paywall already gets plenty of traffic and converts badly, don't cut onboarding. You'll just send it more of the same problem.
- Check every onboarding question against what the user has got out of the product by that point. If a question arrives before someone understands why it's worth answering, move it later or drop it.
- Measure both halves. Onboarding completion rate and conversion per paywall viewer. One going up while the other goes down is the normal result, not a broken test.
Long and short onboarding can both be right for your app, but the time each one is most effective can vary drastically based on who your audience are, what they’ve experienced of your app, and what you’re asking from them. So if you’re wondering about adjusting your onboarding length, don’t start with onboarding — start by looking for leaks in your funnel.
来源:RevenueCat 订阅应用报告 · revenuecat.com