What to know
Good keyword research starts with a user problem, not a giant list of popular words. Write down what the app helps people do. Collect phrases from Apple search suggestions, customer reviews, support questions, and competitor page language.
Start with a feature and benefit inventory. Write the action, the outcome, and the audience for each important feature. Turn those facts into plain phrases a person might use. Remove terms that sound attractive but do not describe the current product.
Use several evidence sources, then label each one. Apple’s keyword guidance emphasizes relevance and popularity. Reviews and support questions show how real users describe a task. A third-party score can help compare options, but it is still an estimate with a date and market attached.
Score each phrase on three questions: Is it relevant? Does the user appear ready for this result? Can a small app realistically compete? A broad word may have more demand but bring poor-fit visitors. A specific phrase may bring fewer searches but better users.
Keep a research table with the phrase, intent, storefront, country, language, date, source, popularity signal, difficulty signal, and decision. Do not merge scores from different markets as if they describe one audience. If the source does not explain its metric, record that limitation.
Place the final set across the name, subtitle, keyword field, and description without repeating the same phrase everywhere. Read the fields aloud. If the phrase makes the listing unclear or promises a feature the app does not have, remove it even if a tool gives it a strong score.
Recheck the list after a real product change or a meaningful shift in user language. Keep the old version with the date so you can connect a listing change to later results. Do not claim a keyword caused growth without comparable storefront data.
Use a decision matrix instead of one blended score. Give each candidate a relevance note, user intent, storefront and language, evidence source, popularity or discovery signal, difficulty estimate if available, product proof, and final decision. A term with a strong signal but weak product fit should not outrank a term the app clearly satisfies.
For example, a budgeting app might compare “expense tracker,” “shared grocery budget,” and “finance.” The first may describe the core job, the second may match a clear audience need, and the third may be too broad. The right choice depends on the product brief and evidence, not on which label sounds biggest.
Review candidates with a person who knows the build but did not create the initial list. Ask them to explain what the app would do for a person searching each phrase. If the explanation requires a feature, market, or outcome that is not in the product, remove or qualify the term before it reaches metadata.
Check the source before using its signal. Record whether the observation is a first-party data point, a user-language note, or a third-party estimate. Add the market, language, date, and method so a later reviewer does not treat an estimate as a fact.
Create a short research-to-review handoff with the selected terms, rejected terms, product proof, field assignment, storefront, and open questions. A reviewer should be able to approve the shortlist or ask for a specific missing fact without rebuilding the research process.
Refresh the list when the app changes its main job, audience, supported region, or terminology. Keep the prior version and decision notes. A term can remain popular or natural while becoming a poor fit for the product, and a new workflow can create better language that older research never captured.
Add a freshness review to the research table: product brief version, terminology source, storefront and language, date checked, current field assignment, and reason the term still fits. Mark phrases that depend on an old feature, retired audience, or changed policy for immediate review.
Run the shortlist through a research-to-build check before drafting metadata. For each selected term, point to the current screen, workflow, or support evidence that satisfies the search intent. If no one can show the proof, move the phrase to an open-question list instead of polishing it.
Reopen the handoff when a source changes its definition, the product adds a workflow, a locale expands, or a team member cannot explain a selected term. Preserve the evidence and rejected candidates so the next pass improves the decision rather than starting with a blank table.
ASOgenic can help turn this table into a reviewable draft and check the mechanical limits. Your team still decides whether each term is truthful, useful, and right for the release. The best shortlist is the one a product owner can explain and approve.