ASO for publishers and portfolios

Tier the catalog before touching anything

Most publishers discover during this exercise that a third of the catalog has not had a metadata change since submission. That is the cheapest install growth available to them.

Shared keyword and creative libraries

• Genre keyword sets built once and reused across every title in that genre, adjusted per app rather than rebuilt.
• Screenshot templates that carry brand consistency without requiring bespoke design per listing.
• Localization glossaries so the same feature is named the same way in every market and every app.
• A record of which creative patterns won tests, so the next title starts from evidence rather than taste.

Guardrails for automation at scale

1. Require human approval on any change to a flagship or to a legally sensitive claim.
2. Cap how many apps can receive the same change in one release wave.
3. Keep a portfolio-level change log; per-app history is not enough to spot systemic regressions.
4. Alert on rank drops by portfolio segment, not per app, or the noise buries the signal.

Worked example: a portfolio of thirty casual titles

A publisher with thirty live titles across three genres had no consistent view of listing quality. Some apps were maintained weekly by their own teams, others had not been touched since launch, and there was no way to tell which was which without opening each console by hand. Effort was going to the titles with the loudest producers rather than the titles with the largest recoverable upside.

Scoring the whole portfolio on one model changed where the work went. Four titles turned out to be carrying most of the available gain — long-tail games with real install volume and listings that had gone untouched for two years — while several of the titles receiving weekly attention were already close to their ceiling. The reallocation cost nothing and was worth more than any single optimization.

Cross-title learning followed. When the same icon treatment won tests on two puzzle games, the finding could be applied to the other four rather than rediscovered separately, which is the specific advantage a portfolio has over a single studio and the one most publishers never actually collect.

What the agent does in week one

1. 1

   Day 1 — score every title

   One model across the whole portfolio, so listing quality is comparable between studios rather than described differently by each.

2. 2

   Day 2 — rank by recoverable upside

   Combine current install volume with the size of the listing gap, so effort follows opportunity rather than internal noise.

3. 3

   Day 3 — group the shared work

   Cluster titles by genre so keyword sets and creative conventions can be worked once and applied across the group.

4. 4

   Day 4 — open the highest-gain listings

   Draft metadata and creative changes for the top titles, sequenced so no two changes on one app land in the same release.

5. 5

   Day 5 — start the shared test log

   Record every test and result in one place, so a win on one title becomes a default on the others.