What construction app users actually complain about
Published:
We scraped 737,216 app-store reviews and analysed the 208,539 that were in scope: 235 apps, 28,412 of them critical (one to three stars). These pages report what that corpus says.
Every figure below is a computed count, not an estimate — each article ships a provenance ledger naming where in the corpus each number came from. Where the data disconfirms something we expected, the disconfirmation leads. Every page here links a plain-markdown mirror of itself from its ; all of them are listed in /llms.txt.
- When the calculator app went subscription, users went back to the hardware
Tenon Tools counted the alternatives Construction Master Pro reviewers name after its move to subscription. The most-named one is hardware, not an app.
- What 16,158 critical reviews say about field-camera apps
Ads lead field-camera complaints in the Tenon Tools review study: 1,906 matched and 2,286 emergent mentions in 16,158 critical reviews. Crashes run flat.
- The German accuracy gap: wrong results are 16.7% of German-storefront calculator criticals and 1.0% of US ones
Tenon Tools' 208,539-review study: wrong-results complaints are 16.7% of German-storefront calculator criticals (34 of 204) vs 1.0% in the US (96 of 9,169).
How was this corpus built?
Reviews were scraped from both app stores across the construction, field-service and jobsite-camera categories, then filtered to the apps actually in scope. 737,216 reviews came in; 208,539 across 235 apps survived the scope filter, 28,412 of them critical at one to three stars. The remainder were dropped because their apps turned out to be off-domain — a general-purpose photo editor is not a field camera, however often it appears in the same search.
Matching is plain regular expressions over review text, not a model. That choice is why the error rate is knowable: a hand spot-check of 180 reviews put the false-positive rate at 8.3%, and that figure is published beside the counts rather than kept in a drawer. Two hypotheses failed that audit outright and were rebuilt from scratch; the numbers they produced first are named as void in the article that used them, next to the corrected ones.
Why are the review tiers never blended?
Because blending them produces a number that describes nobody. Advertising complaints are a Tier-1 phenomenon — 15.16% of Tier-1 criticals against 1.84% of Tier-3 — while subscription anger is the mirror image, 16.72% of Tier-3 criticals against 2.73% of Tier-1. Average those and you get a middling figure for a problem that is severe at one end of the market and nearly absent at the other.
So every count on these pages names its tier and its denominator. A percentage without the denominator beside it is not a finding, and it is not quotable.
What did the data disconfirm?
Three expectations went into this study and did not survive it. Resentment at PlanGrid being folded into Autodesk turned out to be stale — three mentions across 2025–26 — and that positioning was retired rather than softened. Users churning from a calculator app that went subscription did not move to a rival app; the most-named alternative was a physical calculator, which changed what a competing app has to beat.
The third is the one worth reading carefully, because it is a null result and not a refutation. Small German crews being priced out by per-seat site-diary software produced zero supporting mentions in 608 incumbent reviews. That is a wrong-instrument result: people who never buy the software never install it, and people who never install it never review it. It is recorded here as unmeasured, and no claim on this site rests on it being true.
This page as plain text: markdown mirror.