Billing and cancellation complaints are the longest reviews users write. Here is the data — 354 of them, across 21 apps.

Billing complaints are 6.6% of every negative review in this category — not the biggest bucket by volume, but consistently the most furious. We counted them complaint by complaint and looked at what actually triggers them. The answer is less about subscriptions than you would expect.

The severity table

Every negative review in the sample is assigned one primary category. The table below orders categories by how angry the writing is, not by how much of it there is — read alongside Research No. 1, which orders them by volume.

One-star share and length both measure the same thing from different angles: how much trouble someone took. Billing wins both, and it wins them by a margin wide enough to not be an artefact of the sample.

Finding 1 — it is not "too expensive", it is "I stopped you and you took it anyway"

Of 354 billing complaints, the words cluster around money that moved rather than prices that were too high:

Read that top row again: 240 complaints mention a refund, and 8 mention still being charged after they say they cancelled. Very few complain that the price is too high — those get filed under pricing, a separate category with 320 complaints, and they read completely differently. People accept a price. They do not accept a charge they believe they ended. Every one of these is a trust failure, not a product failure, and no amount of roadmap work touches it.

Finding 2 — 37.3% of these complaints are not about subscriptions at all

Two lead marketplaces — Thumbtack for Professionals, Angi for Pros — account for 132 of the 354 billing complaints (37.3%). Almost all of them are pros saying the same thing: they paid for job leads, and the leads were fake, duplicate, or never answered. Our classifier files them under billing because the sentences are full of charged / refund / pay — but the complaint is about product quality wearing billing vocabulary.

We set that bar at 5 lead-mentions per app on purpose. 3 more apps mention leads once or twice (Vagaro Pro, Joist: Estimate Invoice Maker, LawnStarter for Providers) — noise, not a pattern. Counting them too would inflate 132 into something bigger and blur the finding into "everybody has billing complaints", which is the opposite of the truth here.

Separating them out matters. The remaining 222 complaints spread across the other 19 apps at a much lower rate, and they behave differently: those are genuinely about cancellation friction and auto-renewal. Read the headline count as "354 apps have subscription rage" and you will draw the wrong conclusion, then probably ship the wrong fix.

Finding 3 — billing complaints are structural, and some apps have none

Fleetio Go - Fleet Management, ServiceTitan Field, FastField: Field Service Forms each have at least ten negative reviews in the sample and zero billing complaints. They are not better companies and they do not have nicer users. They simply have no recurring charge to cancel. Meanwhile the two apps at the top of the table above are exactly the two lead marketplaces: the apps with the loudest billing anger are also ones where the entire product is a charge attached to an outcome the user cannot verify.

That is the actionable version of this research: billing anger is not a tone problem to be managed by a support team. It is created by a pricing decision, and it can be priced out — which no amount of "we hear you" replies will accomplish.

What else these complaints mention

Billing complaints rarely travel alone. Counting which other categories get mentioned in the same complaint:

support (89) · price (48) · missing (37) · crash (26) · ux (12)

The practical consequence: a user who cannot cancel does not file a tidy billing ticket. They write one long review in which the cancellation, the silence from support, and the refund that never arrived are all the same event. Fixing only the cancellation flow removes maybe a third of the anger in that review; leaving support unreachable means the same user writes again.

Method, limits, and what we deliberately did not do

• Corpus: 7,677 sampled reviews across 21 field-service apps, of which 5,331 are 1–2★. Reviews come back in Apple's own review ordering, which skews critical, so the mix of categories here describes what people complain about — not how many people complain. The rate question is answered by Apple's star distribution in Research No. 1, using a different denominator on purpose.
• We publish counts and ≤3-word functional phrases only. No review text, no usernames, no avatars. "cancel subscription" is a description of a feature, not someone's words.
• Classification is rule-based (taxonomy.py), fixed priority order, one primary category per review. Coverage per app varies; unmatched reviews are reported as unclassified rather than forced somewhere plausible.
• The lead-marketplace split in Finding 2 is a judgement call we made after reading the counts — "paid for leads that went nowhere" and "could not cancel my subscription" land in the same rule-based bucket and we separated them by hand. We are showing you the arithmetic (132 of 354, bar: ≥5 lead-mentions per app) so you can undo it if you disagree.
• This page is regenerated from live data against whatever corpus exists on 19 Sep 2026. Sample sizes change as we add and retire apps, so the counts here are not comparable to a previous edition unless the app count is the same — check the metaline before you quote.