Every install starts as a randomized 50/50 A/B test on the brand's own Instagram traffic: bucket A is escaped to the real browser, bucket B stays in the in-app webview as the control. Same ads, same spend, same store. Below are completed test windows, reported with sample sizes, z-scores, and outlier-trimmed revenue — the same numbers the brands see on their dashboards.
A 17-day randomized 50/50 split on live Instagram traffic. Visitors escaped to Safari converted at 3.4× the rate of visitors left inside the in-app browser — on identical ads, identical spend, identical landing pages.
8 days, 24,426 randomized Instagram visitors on a concentrated herbal supplements store: +37% checkout conversion and +37% revenue per visitor for escaped traffic — a perfect 50/50 split with zero outlier orders to argue about.
20 days, 53,784 randomized Instagram visitors. The in-app browser wasn't just hurting checkout — it was killing product discovery: 17.5% of control visitors ever viewed a product, versus 48.7% of escaped visitors.
40 days, 31,382 randomized visitors: +83% checkout conversion and +88% revenue per visitor for escaped traffic — while Ads Manager was simultaneously reporting escaped conversions as a ROAS problem.
The longest high-volume window in the portfolio: 42 days, 54,598 randomized visitors. Escaped traffic converted +125% better and produced +142% more revenue per visitor — at z = 12.5, one of the most decisive results we've measured.
Eight days, 180,023 randomized Instagram visitors, 2,648 orders. At this scale the error bars vanish: escaped traffic converted +40% better (z = 8.7) and produced +42% more revenue per visitor, outlier-trimmed on both arms.
Ten days, 32,253 randomized visitors on a high-AOV protective-gear store for MTB, snow, and motorsports. Escaped traffic converted +111% better and produced +128% more revenue per visitor.
A one-week split across 127,077 Instagram visitors during a viral spike. Escaped traffic converted +46% better and produced +44% more revenue per visitor.
25 days, 63,173 randomized visitors. The first few days of this test read flat — then the full window settled at +41% checkout conversion and +40% revenue per visitor at z = 6.5. A case study in why we don't call tests early.
Six weeks, 4,270 randomized visitors. Even with an unusually strong 10.9% control conversion rate, escaping the in-app browser lifted checkout conversion 30% and revenue per visitor 37%.
Randomized at first touch. Visitors are assigned to escape or control before the page paints, so both arms see identical ads, prices, and merchandising.
Outlier-trimmed revenue. Single whale orders (> Q3 + 3·IQR and > 8× the median) are excluded from per-visitor revenue on both arms — in either direction. Conversion counts are never trimmed.
Significance, stated plainly. Conversion lift is tested with a two-proportion z-test and reported with its z-score and p-value. Brands are anonymized by category; several are under NDA.
Install is one script tag. The 50/50 test starts immediately, and in 7–14 days you have your own defensible number.