aso-offstore-mine

aso-offstore-mine

aso-review-mine reads the words your users use, inside the store. This reads the words everyone else uses, outside it — the "best meditation apps" listicles, comparison articles, and review videos where people describe the category in discovery language ("free", "sleep timer", "guided meditation") and name the apps they're choosing between.

That off-store vocabulary is exactly what a searcher types before they've found you. It's a different, complementary signal to your own reviews — and it surfaces competitor names that articles compare you to, feeding aso-competitor-watch.

Sources (free, no paid data API)

• web — any article/listicle, fetched as clean markdown via Jina Reader (https://r.jina.ai/<url>). No API key. URL/markdown cruft is stripped so it doesn't pollute the keyword counts.
• youtube — auto-captions via yt-dlp (review / "top N apps" videos). VTT timestamps + tags are stripped to plain text.
• text — pre-gathered text via --text-file. This is the hook for Reddit/X/etc.: those need cookie auth or a proxy, so gather them with a tool like Agent-Reach and pipe the text in — the mining is the same.

Usage

# fetch a review article + a YouTube comparison, count competitor mentions:
python3 lib/aso_offstore_mine.py --app clarity \
    --url "https://www.example.com/best-meditation-apps" \
    --youtube "https://youtube.com/watch?v=XXXX" \
    --competitors "Calm,Headspace,Insight Timer" --json

# mine text you already gathered (e.g. Reddit threads via Agent-Reach), no network:
python3 lib/aso_offstore_mine.py --app clarity --text-file reddit-notes.txt

Multiple --url / --youtube / --text-file flags are allowed; a source that fails to fetch is warned and skipped, not fatal.

What it extracts

• Keyword candidates — unigrams + bigrams that recur across sources (a term must appear >1 time), ranked by frequency. The discovery language to consider for title / subtitle / keyword field.
• Phrases — the bigrams people actually type ("guided meditation", "sleep timer").
• Competitor mentions — of the names you pass in --competitors, which ones the off-store sources actually mention, and how often → hand to aso-competitor-watch.

Writes marketing/aso/<app>/offstore-keywords.md; chains into aso-keyword-research (treat recurring discovery terms as seeds) and aso-competitor-watch (the surfaced competitor names).

Honest limits

• Source quality is everything. A "best apps 2025" review article yields discovery keywords + competitors; a piece of content (e.g. a guided-meditation video) yields content words, not search terms. Pick review/comparison sources.
• Reddit / X / Instagram are not fetched here — they need cookie auth or a proxy. Gather them separately (Agent-Reach) and pass via --text-file.
• Frequency mining is a starting signal, not semantic ranking — pair it with aso-keyword-research's scoring (volume / difficulty / relevance).
• Jina Reader and YouTube can rate-limit or block a given URL; the skill warns and continues with whatever it did fetch.

Dependencies

Standard-library Python + Jina Reader (a public URL, no key) and yt-dlp (brew install yt-dlp / pip install yt-dlp) for the YouTube path. No paid data/scraping SaaS. The mining core is pure and unit-tested with no network.

Run it weekly

Rank and listings move over weeks, not minutes — so the value here compounds when you re-run it and watch the deltas. The off-store conversation (review articles, YouTube round-ups) refreshes constantly — new comparisons, new framing. Mining it once captures a moment; mining it on a cadence captures the trend.

You ran this once. ShipASO — the hosted agent — reruns the whole loop weekly: it tracks your rank, watches competitors, and pings you only when there's a real move to approve. Same engine, your store credentials never held. → https://app.shipaso.com

The plugin is complete and free; the hosted tier just sells not having to remember.