ZeroCount vs Countfire — an honest comparison.

Countfire counts by pattern matching against a symbol library — you tell it what a mark looks like and it finds every other instance of that exact mark across the set. Fast, exact, and free by the fortieth sheet. The cost sits at the front of the job: every practice draws to its own house style, so a library that is complete for one consultant is partial for the next, and the marks that go missing are the ones that matter — the emergency variant, the second socket type, the fit-out overlay device.

A trained model works from the other end. ZeroCount reads the convention rather than matching a stored shape, so there is nothing to build before the first count and nothing to rebuild when the next consultant draws it their way. The trade is real: a model returns a judgement rather than an exact match, so it will occasionally want a correction where pattern matching would have been certain. What you get is a first drawing from an unfamiliar practice costing roughly what the tenth does — the electrical and mechanical pages go through what it reads, sheet by sheet.

Countfire’s published documentation describes vector PDFs — drawings exported straight from CAD, where every line in the file is still a line the file knows about. That is the right input for pattern matching: exact geometry, exact match. On anything else Countfire is clear. A scan or a photograph is a drawing made of images, and Countfire’s changelog says: “Since images lack the information for Countfire’s automatic counting, you’ll need to use manual counting tools.”

What an input constraint costs you is easier to describe. The clean CAD export exists, but it stays near the design team. By the time a package reaches a subcontractor it has been through a print, a scan and an email, and the sheet with the red pen across the riser is a photograph somebody took in the site office. On insulation packages, furthest down the chain, that is the normal case.

A format limit rarely announces itself as an error. It arrives as an afternoon redrawing a corridor so the software has something to match, or a count that ran cleanly on eleven sheets and skipped the twelfth. Those hours are unbillable, and they come out of the checking time at the end of a bid. ZeroCount runs vector PDFs, scanned PDFs and photographed plans through the same engine at the same price, so bring the worst drawing on your desk to the demo rather than the tidy one.

Countfire does not publish pricing. Third-party trackers report a tiered model in the region of $99–$999 per seat per month, which we cite as public reporting rather than as fact. Their price is theirs to publish, and not publishing it is an ordinary decision in this market — most takeoff platforms make the same one.

The practical difference is where the number enters your process. With a published price you can put a line in a budget on Tuesday, get a nod from whoever signs, and arrive at the demo asking whether the software reads your drawings. Without one, the first call is partly a qualification call and the number lands after it, once you have already spent the time. Neither is dishonest. One is a shorter route to a decision.

ZeroCount is £55 per licence per month, published, before you speak to anybody. The fuller pricing comparison — what the trackers report, what a tiered per-seat model does as a team grows, what you would sensibly budget — is on the Countfire pricing page.

Both products get you from a drawing to a price; the difference is where the price lives. Countfire has a built-in estimating module, which is the simpler shape and worth saying plainly: one product, one login, one support conversation, nothing to integrate. For a team whose work stops at the tender, that is often the right answer.

ZeroCount hands off into Ensign instead, with 300,000+ live UK prices behind it. The case for that shape is what happens after the tender. A takeoff is not the end of a quantity — it is the first version of a number that gets priced, submitted, won, built, and argued about in month four. When the count, the rate and the job share one data spine, a variation is a question you answer by walking back to the drawing it came from. When they sit in separate systems joined by a spreadsheet export, it is a morning of reconciliation.

On project management the table says “not stated publicly” for Countfire, and that is deliberate: their public material does not address it, and an absence of published claim is not evidence of absence. ZeroCount links through to project management via the Ensign group, which is the same argument again — the measurement keeps its identity all the way through.