We classified 2,844 App Store complaints across 21 field-service apps. Crashes are a top-3 complaint in 18 of them.

Every count here comes from a published rule set (taxonomy.py) applied to review text — no human read them into buckets, and no model guessed. Coverage ran 36.0%–79.6% per app. Unmatched reviews are reported as unclassified rather than forced into a category. What this category complains about is not what the star ratings alone would tell you.

The ranking

Complaint rate = share of an app's all-time ratings that are 1–2 stars, taken from Apple's published star distribution. It is deliberately not computed from our review sample: that sample comes back in Apple's own review ordering, which skews critical (but not uniformly across apps), so its internal ratio would overstate the rate several times over. You can recount any figure here on the app's own App Store page.

Samples run from 94 to 549 reviews per app. Read the tiers as meaningful and the exact ordering inside a tier as noise — a 100-review sample carries real error bars.

What they actually complain about

These 21 apps gave us 5,331 one- and two-star reviews. Of those, 2,844 matched a named complaint type and the rest stay unclassified rather than being forced into a bucket (per-app coverage ranged 36.0%–79.6%). The classified subset clusters into a small number of recurring categories. Each review is assigned one primary category by a fixed priority order; the counts below use the same method as Research No. 2, so the two articles can be cross-checked line by line.

In this category, crashes are not a technical footnote — they are the product experience. A field-service app runs in a truck, in a basement, on a roof, often on a phone with a cracked screen and one bar of signal. When it fails, the technician cannot do the job. Crashes account for 590 classified complaints here — about 1.8× pricing (320), the thing SaaS teams usually worry about most.

Three things worth knowing before you read the table

1. Complaint rate is a property of the business model, not of code quality

Two lead marketplaces — Thumbtack for Professionals, Angi for Pros — carry the largest share of billing complaints in the entire set — and most of those are not about subscriptions at all. They are pros saying they paid for job leads that never turned into jobs. Read Research No. 2 before you treat a high complaint rate as an engineering verdict.

2. A small number of apps have effectively no billing complaints

Fleetio Go - Fleet Management, ServiceTitan Field, FastField: Field Service Forms each have at least ten negative reviews in the sample and zero about cancellation or billing. Those apps are not nicer companies. They charge once, or not at all, and they removed the surface where this anger forms. That is the strongest available evidence that billing complaints are structural, not cultural.

3. Users name competitors out loud — rarely, but decisively

Only 66 reviews across the whole set name a rival directly. But when they do, they say exactly which feature won them over, and they say it in public. For a product team that is the most expensive sentence in the dataset: it is a conversion decision already made, written down.

Method, limits, and what we deliberately did not do

• Source: Apple's public customer-review feed. No private data, no scraping behind a login, no App Store Connect access.
• Two different denominators, always labelled. Complaint rate comes from Apple's all-time star distribution (a complete count). Category breakdowns come from our most-critical-first sample (a partial count). The first answers "how many people are unhappy"; the second answers "what are they unhappy about". Mixing them produces numbers that look precise and are wrong.
• We publish counts and recurring functional phrases only. No review text is reproduced, no usernames, no avatars. A count is a fact about a public dataset; a copied review is someone's words — and redistributing it is the line we do not cross.
• Classification coverage runs 36.0%–79.6% per app, depending on how cleanly complaints map onto known categories. Unclassified reviews are excluded from the category breakdown, never forced into one. Where a category has fewer than five matches we do not generalise from it.
• This page is regenerated from live data. The numbers above were computed on 19 Sep 2026 against 21 apps and 7,677 sampled reviews. When the corpus changes the page changes with it — which is why we print the date and the corpus size on every edition rather than quietly editing in place.
• This is a cross-sectional snapshot, not a trend. It says what this batch of reviews said. It does not say whether things are getting better or worse.