What remediation really costs: 792 findings are 284 decisions

First step: raw hits versus consolidated findings. A contrast failure in the footer appears on every page of a site. A testing tool therefore reports it as often as it tested pages. In our measurement across 20 sites, 1,862 raw hits stood against 792 consolidated findings — a factor of 2.4. On individual sites the gap was larger still: 396 raw hits against 157 actual places.

Second step: findings versus decisions. Even 792 is too high. Many findings are repetitions of the same correction: the same colour pair, the same CSS class, the same icon in five places. Add those together and 794 analysed findings come down to 284 independent decisions — an average of 2.8 findings per decision, considerably more on the large jobs. The site with 157 findings came to 29 decisions.

Anyone quoting by raw hits is selling the same work up to five times over. Anyone quoting by findings is still selling it 2.8 times on average. That is no imputation of intent — it is simply the number that falls out of the tool if you do nothing further with it.

For the sake of clean figures: the remediation analysis ran on 25 August 2026 against the then-current stock of 19 sites with 794 findings. The accessibility measurement in the article on the most common barriers now covers 20 sites with 792 findings, because one site was added and several scans were refreshed. The two figures are therefore similar but not identical. We do not harmonise them artificially.

The reason 794 findings become 284 decisions has a name: the batch correction. Every proposal gets a key derived from the actual change at attribute level — for contrast from the colour pair, for a removed CSS class from that class. Everything sharing a key is the same act.

An example from the measurement: on an outdoor retailer’s site, a single act — removing the class no-underline from links in running text — cleared 42 findings at once. One colour value in a central CSS variable clears every finding with the same colour pair, no matter how many pages they sit on.

Measured in the browser, in the worksheet we use to work through findings: a job with 152 findings was empty after 37 key presses, one with 13 findings after 9. That is expressly not the working time — it is its floor. The time goes into reading and deciding, not into ticking off. But it shows the order of magnitude by which an honest calculation differs from one based on findings.

For comparison, the prices we read off publicly available provider pages on 25 August 2026. Ongoing automated testing services ranged from 19 to 299 euros a month; one provider adds a “from 999 euros” tier for very large sites. Some net, some gross — the comparison is only worth making if you read that along with it. One-off testing services run from a quick check at 79 euros to a full audit between 490 and 1,250 euros.

For remediation there are no comparable list prices. Agencies state on their own pages roughly 2,000 to 8,000 euros for retrofitting an existing corporate website, 8,000 to 15,000 euros for sites with a shop or many interactive components, and 200 to 800 euros a month for ongoing support. These are providers’ own statements, not surveyed market data — we reproduce them as exactly that.

Put side by side, the contradiction becomes visible: the most expensive item is remediation, and it is precisely for that item that nobody states a traceable unit of calculation. The range from 2,000 to 15,000 euros is a factor of 7.5. The range of findings is just as wide — except that one of them is measurable and the other is guessed.

A fuller breakdown by the four cost blocks — testing, remediation, monitoring, documentation — is on our cost page.

Conformity is not a state but a process. Every theme update, every new product page, every additional plugin can introduce new barriers — and regularly does. A one-off remediation applies to today’s state.

That is not our sales prose but the statutory text. § 14(3) BFSG requires the service provider to meet the accessibility requirements at all times, and expressly names changes in the way the service is provided, changes in the requirements and changes in the harmonised standards as matters to which “due regard” must be given. And § 17(3) requires anyone invoking a disproportionate burden to repeat the assessment at least every five years and whenever the service changes.

In practice: an honest calculation contains a recurring item. How large it is depends on how often the site is worked on — different for a shop with weekly changes than for a site touched twice a year.

Technically the effort for it is small: a second scan compares finding by finding using the same fingerprint and separates them into fixed, still open, new. That is at once the evidence for third parties — and, frankly, the point at which you notice whether a remediation was really installed or merely ticked off.

What that costs with us, and what the test mark is tied to, is on the pricing page. In short: the mark is valid for as long as testing continues — that is not a contract clause of ours but the condition under which a test mark may be displayed at all.