Work / our own app
Zip VPN
Our own app, measured the way we would measure yours.
A free VPN serving around 20,000 users a day, paid for by ads, with no clear answer to the questions that mattered: which locations people actually wanted, where a session ended, and how the ads were being met.
The figures
- 20,000
- users a dayour app
- 300,000
- events a dayour app
- 20%
- of users come backour app
In order
What we asked, recorded, learned and changed.
What we asked
Before touching anything we wrote down the questions. Which locations do people come back to? Where does a first session end? Does a failed connection cost us the user? How does someone behave around an ad, and does it push them out?
What we recorded
Every step from install to a working connection, with the outcome attached: which location was tried, whether it answered, and how long the user waited. Every ad shown, and what the user did next. Return visits read against all of it.
What it told us
The users we lost were not leaving because of speed. They left during the seconds a dead location took to fail. A small number of locations carried most of the returning users, and ad placements that interrupted a connection cost more sessions than they earned.
What we changed
Failover became the app's defining feature: it keeps trying until it connects, and shows it working. The locations people love are kept healthy first. Ads moved to the moments the data said were safe. Around 20,000 people a day use the result, and one in five of them comes back.
The product
The screens the events come from.
The lesson
The expensive guess was that users wanted speed. The data said they wanted certainty, and a particular short list of places. That is the kind of thing measurement finds, and it is why we do it to our own software first.