Work / Mix VPN
VPNAnalytics
Two platforms behaving so differently they needed separate funnels
2,545 Indonesian Android installs converted at 6.76% to trial, while iOS across the UK, Canada and Australia produced paid subscribers with almost no trial events — two funnels a blended report would have averaged into one misleading number.
2,545
Indonesia Android installs
172
Trials started, Indonesia
0
Refunds across all markets
The challenge
A single blended install-to-paid rate would have been actively harmful here — Android and iOS weren't running the same funnel, and the markets weren't comparable at wildly different install volumes.
The approach
Every market reported at its own platform and row, denominators visible, so a 100% rate on 2 trials read as the small sample it was. iOS paid conversions with no preceding trial event were flagged as a tracking question to resolve, not a performance verdict to act on.
What we did
01Split by platform firstAndroid and iOS reported as separate funnels rather than blended into one conversion rate.
02Keep denominators visibleEvery rate shown against its install count, so a 100% rate on 2 trials reads as the small sample it is.
03Flag funnel gaps as gapsiOS paid conversions without trial events treated as a tracking question, not a performance verdict.
04Report refunds either wayZero refunds stated across all five markets rather than left out of the table.
Geo and platform split
| Platform | Market | Installs | CR to trial | Trials | CR to paid | Paid | Refund |
|---|---|---|---|---|---|---|---|
| Android | Indonesia | 2,545 | 6.76% | 172 | 5.23% | 9 | 0 |
| Android | Japan | 107 | 1.87% | 2 | 100.00% | 2 | 0 |
| iOS | United Kingdom | 180 | — | 1 | — | 1 | 0 |
| iOS | Canada | 216 | — | 0 | — | 2 | 0 |
| iOS | Australia | 33 | — | 0 | — | 1 | 0 |
