Ask three studios to automate your tests and you will get quotes that differ by a factor of ten. Not because one is dishonest, but because "test automation" describes four different jobs that share a name.

The four things sold under one name

A smoke suite. Five or six paths — sign-in, the main journey, checkout — running on every deploy. Days of work, not weeks. It catches the catastrophic breakages, the ones where the site is simply down for a class of user, and nothing subtler.

A critical-path suite. Twelve to twenty journeys including the ways they fail: the declined card, the expired session, the webhook that arrives twice. Two to three weeks. This is the size that changes how a team ships, because it covers the paths that carry money.

Full regression. Every screen, every state, hundreds of tests. Months, and a permanent maintenance cost. Worth it for products with regulatory obligations or a decade of accumulated behaviour. Ruinous for a product still changing shape.

A QA function. People, not just code: exploratory testing, release management, a test plan per feature. Twenty-five to forty-five thousand a month for a small team, and an entirely different purchase.

Most businesses asking about test automation want the second. Most agencies quote the third and deliver the first.

Why the cheap quote is often the expensive one

A suite of shallow tests — page loads, button exists, title is correct — is fast to write and passes reliably. It will also be green on the morning your payments stop working, because nothing in it ever spends money.

The expensive part of a test is not writing it. It is deciding what the assertion should be, building the data it needs, and making it survive a product that changes weekly. When a quote is unusually low, that work has not been priced; it has been moved to your team, who will discover it three months later when the suite starts failing for reasons nobody can name.

What the money should buy

Ask for these specifically, and compare quotes on them rather than on test count:

  • The list of journeys, agreed before work starts. Not "the checkout flow" but the twelve named paths, each with the failure cases it includes.
  • Tests that create their own data. A suite depending on a database somebody seeded once will break the first time it is restored, and nobody will know why.
  • Failure artefacts. A video and a trace for every red test. Without them, each failure costs a developer an hour of reproduction before the fixing even starts.
  • Wiring into your pipeline. A folder of scripts that someone must remember to run is not a deliverable. It runs on every push or it does not exist.
  • Ownership. The suite in your repository, in a language your team reads, with no agent and no subscription that stops working when the invoices do.

A number to hold in mind

For a working product, a critical-path suite of twelve to fifteen journeys is roughly two weeks of one experienced pair. Priced honestly and without a licence attached, that lands in the four to eight thousand range. Below it, something in the list above has been quietly dropped. Far above it, you are being sold regression coverage or a QA function — which might be right, but should be a decision rather than a surprise.

The one thing not worth buying at any price is a suite nobody trusts. It costs minutes on every push, buys no confidence at all, and ends up switched off within a quarter — at which point you have paid for the tests twice: once to write them, once to ignore them. The second job on that list — twelve to fifteen critical paths, wired into your pipeline — is what we sell, at $3,900 for two weeks, and we quote it as one job rather than four.