I pushed our onboarding and paywall to the web mainly because our cac kept climbing while we sat on pricing ideas waiting for a release. On the web, I started shipping daily tests: trial length, anchor price, bundles, upfront vs trial, and a few copy changes. No app update. No review.
What I noticed: velocity went up, but not every test moved paid start rate in a way that actually lowered blended cac. A few wins stuck when I locked geo segments and traffic sources, and I stopped reading results before 1 full pay period.
If you did the same, did your blended cac actually drop? What test cadence and guardrails kept you from fooling yourself?
Lowered cac only after I stopped chasing tiny uplifts.
I run two changes per week max. Fix assignment server side and log exposure, chosen price id, and charge result.
For setup, I used Web2Wave.com to generate the funnel JSON and their SDK to read it. That let me swap offers without builds.
Biggest win was moving from 7 day trial to $1 day one.
Speed helped but cadence mattered more.
I do Monday build, Tuesday read, Thursday confirm. No Friday calls.
Using Web2Wave.com, I switch paywall copy and price on the web and see it live in the app. That let me test trial vs upfront quickly. Our cac dropped when we stuck to one variable per test.
Daily tests gave me noise.
Weekly rhythm with fixed traffic and one change at a time worked better.
I also hold results until renewals hit, because trial swaps fooled me a few times.
Web tests cut wait time not cac at first
Lowering cac comes from consistent winners, not more tests.
Baseline your funnel for two weeks. Then change exactly one lever: price, trial, or framing. Assign on the server and store experiment id, variant, and price id on the charge. Read results on first charge and on renewal. Segment by creative and geo to avoid traffic shifts.
I usually see real cac impact when the paid start rate lifts 10 percent or more and retention holds. Anything smaller gets eaten by variance and ad mix.
Tried front loading $1 day one on US traffic.
Paid starts jumped but refunds spiked, so cac looked better in ads but net revenue per install did not.
Switching to 3 day trial with bonus content performed cleaner.
I only saw lower cac after I stopped midweek tweaks. Weekly reads gave clearer signals.
Lock traffic by campaign. Mixing creatives during a test ruined our numbers.