# Lesson 15: How to prioritize ASO keywords

Make a repeatable prioritization method that exposes assumptions rather than hiding them in a score.

Prerequisite: Demand and competitive-feasibility notes for your shortlist.

Teaching examples are fictional unless explicitly identified as platform definitions.

## Foundations

### Use eligibility before weighting

First exclude unsupported, misleading or unresolved candidates. Then compare eligible terms on product relevance, audience value, evidence strength and feasibility. A weighted average can otherwise let a huge demand estimate compensate for a product the app does not provide. Eligibility is a boundary; weights are a judgment call inside that boundary.

### Write definitions for every rating

If you use a simple high/medium/low rubric, explain what each label means. High relevance might require a central shipped workflow, while medium relevance describes a secondary supported use case. Keep the evidence beside the rating. Two analysts should be able to explain their disagreement in terms of assumptions, rather than arguing about unexplained decimal scores.

### Test the stability of the decision

Change the uncertain ratings and see whether the same candidates remain priorities. A fragile ranking deserves more research or a smaller commitment. Also consider concentration: several highly rated phrases may all describe the same narrow job. The best final allocation is a coherent portfolio, not necessarily the first ten rows of a sorted spreadsheet.

## Apply the method

### Apply hard gates before weighted preferences

Create gates for supported capability, relevant audience and acceptable claims. A failure means reject or defer; do not let a high demand score compensate for a missing function. After the gates, compare candidates using criteria that match the objective, such as intent fit, demand evidence, differentiation and opportunity cost. Define what high, medium and low mean in words. A rubric helps consistency only when two reviewers can understand the same criteria and explain disagreements.

### Keep evidence confidence outside the score

A high rating based on a guess should not look equivalent to a high rating based on multiple observations. Add a confidence note and source reference for each important judgment. If reviewers disagree, identify whether they differ on the evidence or on business priorities. Those require different resolutions: more research may resolve the first, while a sponsor decision may resolve the second. Avoid decimal precision that implies a calibrated predictive model. The purpose is a defensible selection process, not a mathematically impressive ranking.

### Select a portfolio that fits the message

The highest-scoring individual terms may repeat the same concept or create an incoherent visible listing. Review the shortlist as a whole. Does it describe one useful promise? Does it leave room for the brand and natural language? Which concept is displaced when another is included? Record the marginal reason for each addition rather than optimizing each row in isolation. Freeze the rubric before evaluating the final candidates so weights are not quietly changed to justify a favorite phrase.

## Procedure

1. Apply hard exclusions before any comparative scoring.
2. Define a short rubric with written anchors for relevance, value, evidence and feasibility.
3. Rate each eligible candidate and attach the supporting observation.
4. Vary uncertain ratings and record which choices change materially.

## Worked example

Trail Notes rates private journaling high for relevance because it is the central workflow, but medium for demand evidence because only interviews support it. Route planning is ineligible rather than assigned a low relevance score. When the demand assumption changes, journaling remains central because it describes the product. The team can explain its choice without pretending a weighted number predicts a ranking.

| Candidate | Gate result | Priority rationale |
| --- | --- | --- |
| Hiking journal | Pass | Direct fit and clear promise |
| Walking diary | Pass | Useful variant; verify local preference |
| GPS navigation | Fail | Required functionality absent |

Both passing terms can remain research candidates without both occupying every visible field. If space is limited, choose the expression that best fits the audience and surrounding copy, then document how other concepts are covered elsewhere. The rejected term never reaches the weighted comparison. This is a deliberate design choice: no amount of estimated demand should turn an inaccurate product promise into a high-priority recommendation.

## Your ASO assignment

1. Write your rubric before rating the shortlist.
2. Have a second reviewer rate three candidates using the same definitions.
3. Resolve disagreements and record a sensitivity check.

Design a three-criterion rubric for a journaling app and apply a product-fit gate first. Then show how two relevant synonyms could tie on score while only one belongs in the app name.

## Handover

### Rubric

Define gates and comparison criteria before scoring, with examples of each rating.

### Evidence

Attach the basis and confidence of important judgments rather than presenting opinions as measurements.

### Selected set

Explain why the combination serves the objective and fits the available communication space.

### Exclusions

Record failed gates and displaced alternatives so future writers understand the tradeoffs.

### Approval

Name the reviewer, version and evidence that would trigger a revised shortlist.

## Check your work

- Unsupported promises cannot qualify through a high total score.
- Ratings have explicit definitions and evidence.
- The final portfolio communicates a coherent product.

## Common mistakes

- Adjusting weights until a preferred term wins.
- Reporting a subjective weighted score as measured market demand.

## Knowledge check

Two reviewers disagree sharply about a keyword’s relevance. Should you average their scores?

First identify the source of disagreement: different user intent, product capability or market context. Resolve or record that assumption before averaging anything. A numerical compromise can conceal an unresolved product truth that matters more than the score.

## Sources

- [Apple — App Store search](https://developer.apple.com/app-store/search/) — Search presentation, relevant keywords and metadata guidance.
- [Google Play — Metadata policy](https://support.google.com/googleplay/android-developer/answer/9898842?hl=en) — Accurate, relevant metadata and prohibited presentation.

Platform references reviewed 29 September 2026.
www.asoagency.com