ASO keyword audit

Turn a vague "which keywords should we target?" into a shortlist with numbers behind it.

Inputs

Ask for whatever is missing before starting:

• app — a package id (com.example.app) or a name to resolve
• store — ios or android
• country — a two-letter code, default US

Cost

Roughly 40–70 credits for a standard run. Say the estimate before step 2 and let the user stop you. user_balance is free — check it first if the balance might be short.

Steps

1 · Resolve the app. If you were given a name, app_search (2 credits) and confirm the match with the user before spending anything else. Skip when you already have a package id.

2 · Establish the baseline. app_metrics_metadata (3) for the current title, subtitle and category. The existing listing text is itself a keyword hypothesis — note which terms it already targets.

3 · Collect candidates from three sources, deduplicated:

• keyword_app_suggestions (6) — what the store associates with this app
• keyword_suggestions (5) on each of 1–3 seed terms from the title and category
• generate_keywords (12) — optional; use it when the obvious terms are clearly saturated, skip it when the first two already produced 40+ candidates

4 · Size the demand. keyword_metrics_sap_batch (9) over the whole candidate list in one call. Never loop keyword_metrics_sap per keyword — the batch tool exists for exactly this and costs a fraction.

Drop anything with negligible SAP. State the cutoff you used.

5 · Find the current position. keyword_app_metrics_rankings_batch (9) for the surviving list. This splits candidates into three groups that need different advice:

• already top-10 — defend, don't chase
• ranked 11–50 — the reachable wins, where effort pays fastest
• unranked or 50+ — the long game

6 · Check difficulty on the shortlist only. keyword_metrics (4 each) for the top 5–10 candidates by demand. This is the expensive step, so cap it and say how many you ran.

7 · Report. One table sorted by opportunity, columns: keyword · SAP · current rank · difficulty · verdict. Below it, three to five sentences of plain advice — which terms to put in the title, which in the subtitle, which to ignore and why.

Rules

• Batch tools over loops. _batch exists for lists; using the single-keyword tool n times is the most common way to waste a user's credits.
• Report the real numbers. SAP and difficulty come back as values — quote them. Never describe a keyword as "high volume" without the number next to it.
• Say what you skipped. If you capped step 6 at 10 keywords, say so; the user may want the rest.
• One country per run. Cross-market questions belong to keyword_suggestions_worldwide and keyword_app_metrics_rankings_worldwide, which are different, pricier tools — offer them as a follow-up rather than looping countries.