App store reviews as a product-research feed
Play Store and App Store review text as a qualitative feed. Not Maps reviews. Not YouTube comments.
App store reviews are what users say about a build: star ratings and text from Google Play and the Apple App Store. That is software ratings, not Yelp or Maps, not YouTube comments.
ASO ranks are a different job — neighborhood and storefront, not the sentences. Cross-host research adds Reddit. This note is only the store text.
People paste a 4.7 into a slide and call it research. The research is “too many notifications,” “streak anxiety,” “won’t sync with the calendar I already have.” Stars are a sort key. Sentences are the product.
Software ratings, not places
Both stores expose reviews paths. Pair with permissions and data safety on Play when the research is about trust. Keep place-review language out.
curl -H 'Authorization: Bearer bby_live_…' 'https://googleplay.fetcher.sh/api/apps/APP_ID/reviews'
Same idea on appstore.fetcher.sh /api/apps/{appId}/reviews. Pull both when the app ships on both. iOS-only or Android-only products should not pretend the missing store is a clean bill of health.
A research prompt that mixes a restaurant with a habit tracker is how you get nonsense. Play reviews and App Store reviews are two storefronts. Reddit is a community. YouTube is a thread under a video. Maps is a cafe.
How to read the text without drowning
Grep first, then read. “Notification,” “reminder,” “crash,” “refund,” “subscription,” “family sharing” — pick the words for this product. Do not sentiment-score the whole corpus and stop. Average sentiment will hide a 1-star cluster about a billing change.
Pull a competitor’s reviews with the same greps. Users are often clearer about what they hate in the other app than they are in yours. That is still public text. It is not a survey panel, and it is not recruited.
OS versions, device names, and “works on my iPad” live in this corpus in a way they do not on Reddit. Keep them. They are not noise; they are constraints.
A case: “too many notifications” in two stores
A PM keeps hearing about notification fatigue in sales calls. They pull Play reviews and App Store reviews for their app id and a competitor’s, and they grep for “notification,” “reminder,” “mute.”
They do not mix Yelp. They do not mix YouTube comments in this spreadsheet — those go in the market-research pass later. Star ratings on software are not Maps ratings.
The competitor’s iOS reviews are full of “I just use the Clock app.” Their own Android reviews are full of “make it louder.” Same category, opposite problem, two stores. A blended 4.6 would have hidden both.
A case: a subscription change that only showed up on Play
They changed prices on Android first. App Store reviews were calm. Play reviews were not. A weekly combined feed, split by store, is the difference between “reviews are fine” and “Android billing is on fire.”
They still answer reviews in Console and Connect. This host does not post replies. It only reads. The PM’s job is to bring the sentences to the product meeting before the 1-star average shows up in a ranking chart they do not even buy.
Refund and subscription words cluster after price changes. Crash words cluster after a bad build. Those clusters have dates. Store review ids so you can see new text, not a forever-average of 2019. A research pass that always pulls “the listing’s reviews” without paging or dating will repeat the same five quotes in every quarterly doc.
Play permissions stay out of this spreadsheet unless the question was trust. Then they belong in the Play access pass, joined by app id, not dumped into sentiment.
A monthly PM ritual: both stores, grep list from last month plus any new Reddit phrases, new review ids only. Do not reprint 2019’s “love the UI.” Stars without dates are a lie. YouTube comments wait for the cross-host research note unless the question was literally a video.
Keep the objects apart
Play reviews and App Store reviews are two storefronts. Reddit is a community. YouTube is a thread under a video. Maps is a place. If a sentence in your research doc could apply to a ramen shop, you mixed corpora.
Docs on both hosts. 402 or a prepaid key. ASO if the next question is “who else is in similar apps,” not “what did they write.” Date the quotes. A research doc full of undated stars is a mood board.