App store conversion rate optimization

Diagnose before you redesign

The creative hierarchy

1. Icon — must be legible and distinct at 60px, and distinguishable from category rivals in a row of similar tiles.
2. First screenshot — one benefit, six words or fewer, readable without zooming.
3. Second screenshot — the proof, or the differentiator against the obvious alternative.
4. Screenshots three to five — the remaining objections, in the order users raise them.
5. Video — optional; it helps games far more than utilities and can reduce conversion when it delays the first frame.

Testing so the result is readable

Use Apple's product page optimization and Google Play's store listing experiments — they randomise real store traffic, which no external mock-up survey can imitate.

Fix the decision rule before you start: minimum runtime, minimum installs per variant, and the confidence level you will accept. Peeking daily and stopping on a good morning is how teams ship losers with conviction.

Expect most tests to be flat. Two or three clear wins a year on a single listing is a good outcome, and each one keeps paying.

Worked example: fixing a listing that ranked but did not convert

A travel app ranked in the top five for several high-volume terms and converted at roughly half its category benchmark. This is the most misdiagnosed situation in ASO: the team assumed a keyword problem and spent two quarters chasing rankings that were already good, while the actual loss was happening after the impression.

The first two screenshots were the whole story. Both were UI shots with captions describing features — 'smart itinerary sync', 'unified booking view' — that meant something internally and nothing at thumbnail size. They were replaced with two benefit statements legible at a glance, and the feature tour moved to positions three through five where the users who swipe are already interested.

The order matters as much as the content, because most viewers never swipe. The two frames that appear without interaction carry nearly all the conversion weight, and treating them as the opening of a sequence rather than as a standalone argument is the single most common way a strong listing leaks installs.

What the agent does in week one

1. 1

   Day 1 — locate the leak

   Separate impressions, product page views and installs so it is clear whether you are losing people in search results or on the page itself.

2. 2

   Day 2 — grade the first two frames

   Check the opening screenshots at thumbnail size, on a phone, and mark any caption that describes a feature rather than an outcome.

3. 3

   Day 3 — rewrite the openers

   Two benefit-led frames that make the argument without a swipe, with the feature tour resequenced behind them.

4. 4

   Day 4 — set a single-variable test

   One change, a fixed run length and the metric that decides it, so the result is attributable.

5. 5

   Day 5 — baseline everything else

   Record the pre-change numbers for every metric you are not testing, because the comparison is worthless without them.