Estonian software publishers’ homepages, 2026

Key numbers

Who was audited

The list was not bought, scraped or hand-picked. It comes from the Estonian e-Business Register open data (annual reports as of 31 August 2026, register as of 1 October 2026), so anyone can rebuild it. The audit covered every homepage that met the published selection rules and could be measured.

1. Active companies whose latest annual report (2024 or 2025) names EMTAK 58291, other software publishing, as main activity, with at least €100,000 of revenue from it144
2. …with a website on the register, or a company e-mail domain that is not a free mail service110
3. …minus domains shared by more than one company (e-residency service providers)100
4. …whose homepage answered over HTTPS92
5. …minus e-mail domains that do not match the company name (accountants, law firms, personal forwarding)73

Of the 73, 10 are left out of every number below, because what the browser saw was not the homepage:

• 3 — The browser was shown an error page or a bot wall (HTTP 4xx/5xx), not the homepage
• 6 — The browser could not render the page; an HTML-only reading would score an empty script shell, so it was not counted
• 1 — The address did not return an HTML page

What each number is out of

• Overall score, findings and the basics: out of 63 measured homepages.
• Mobile, accessibility and performance: out of 61. On the other 2 the browser rendered the page but the measurement pass did not complete, so those three read “not measured” for them — never a zero — and they are left out of those sections only.

“Software publishing” is the register’s category, not a definition of SaaS. It is chosen by each company in its annual report, so it includes product companies and some whose main income is development or other services. The revenue floor keeps out dormant companies.

Overall score

The overall score is the same 18–96 score every ProUXAudit report carries: deductions for findings by severity, capped by the published ceilings. The median homepage scored 71; the middle half sat between 56 and 82. How the score is built, ceiling by ceiling, is on the example report.

• Score 18–5931.7% (20 of 63)
• Score 60–7934.9% (22 of 63)
• Score 80–9633.3% (21 of 63)

Mobile: small tap targets

Each page was laid out again at 375px wide. The median mobile score was 85; 45.9% (28 of 61) scored below 85. Most of the gap is one check: whether any visible link or button has a box smaller than 24×24 CSS pixels — on a phone, the difference between tapping the menu and tapping the logo.

• At least one link or button under 24×24 px (exceptions not applied)85.2% (52 of 61)
• Body text smaller than 16 px21.3% (13 of 61)

Read this number as “at least one target to review”, not as a WCAG failure count. The check measures each element’s box and flags a page when any one is under 24×24 px. WCAG 2.2 success criterion 2.5.8 also accepts smaller targets that have enough space around them, and links inside a sentence; this check does not apply those exceptions, and no result was reviewed by hand.

Accessibility: contrast and structure

axe-core 4.10.2 ran on every rendered page. The median accessibility score was 82; 55.7% (34 of 61) scored below 85. The most common problems are not exotic: text too faint to read against its background, content that sits outside the page landmarks a screen reader uses to jump around, and icon links with no name.

• Content outside any landmark (header, nav, main, footer)63.9% (39 of 61)
• Text below the WCAG AA contrast ratio59% (36 of 61)
• Links with no accessible name, such as icon-only links41% (25 of 61)
• Heading levels that skip a level37.7% (23 of 61)
• No main landmark34.4% (21 of 61)
• Landmarks that cannot be told apart11.5% (7 of 61)
• No level-one heading for assistive technology11.5% (7 of 61)
• Scrollable areas a keyboard cannot reach11.5% (7 of 61)

Only violations axe-core confirmed are counted; results it marks as needing a human decision are left out, and a page with forty low-contrast labels counts once. An automated rule set catches only part of what WCAG asks for. A page with no violations here is not thereby accessible, and none of these numbers is a compliance position.

The basics: title, description, heading, next step

These are read from the rendered page and graded on ProUXAudit’s own severity scale, which is not axe-core’s impact scale. 23.8% (15 of 63) of homepages had at least one finding ProUXAudit grades high or critical. The two high findings that reached 5 sites were a missing H1 and no label anywhere that tells a visitor what to do next. Neither is a WCAG requirement in itself; they are graded high because a visitor loses the page’s opening statement or its next step.

• Page title under 20 characters or under 4 words22.2% (14 of 63)
• No meta description20.6% (13 of 63)
• No H1 heading14.3% (9 of 63)
• No call-to-action label found on the page12.7% (8 of 63)

Performance: why it is not a headline

The median lab performance score was 100, and the number of homepages whose Largest Contentful Paint went past the 2500 ms “good” boundary was fewer than 5 sites; time to first byte past 800 ms: fewer than 5 sites.

That says more about the test than about the sites. Every page was loaded once, from a data-centre browser on a fast, unthrottled connection. Your visitors on a phone in a train are not that browser. We report the lab number because it was measured, and decline to turn it into “Estonian software sites are fast”, because it cannot support that.

What this study does not show

• What your real users experience

  Every timing is a lab measurement from one browser run in one location. Field data from your visitors is not collected, so nothing here is presented as it.

• Interaction to Next Paint

  INP measures how fast the page answers an interaction. The audit scripts no interactions and collects no data from real visitors, so there is no INP to report, and none is estimated.

• Accessibility compliance

  An automated rule set catches only part of what WCAG asks for. Keyboard paths, screen-reader behaviour and whether content makes sense need a person, so no score here is a compliance position.

• Your funnel and your revenue

  No analytics are connected, so no drop-off is measured and no lost-revenue figure is produced.

• Which company scored what

  Per-site results are kept privately and are not published, ranked or sent to anyone. A homepage is one page on one day; it is not a verdict on a company.

Method

• One run per homepage on 2 October 2026, with the same engine and scoring as a ProUXAudit report: a real-browser capture, then mobile at 375×812px, axe-core 4.10.2 and lab timings in the same session.
• Only homepages the browser actually rendered are counted. A capture that met an error page or bot wall, or fell back to reading raw HTML, is excluded and reported above by reason.
• Shares count sites, not occurrences: a homepage with forty low-contrast labels counts once.
• Audit results are published as aggregates. A result group of fewer than 5 sites is merged with a neighbouring group or shown as “fewer than 5 sites”, so no result can point at one company. Counts of excluded or unmeasured homepages are given exactly.
• A homepage address comes from the register: the company’s registered website, or else its registered e-mail domain if that domain matches the company name. A product domain that differs from the company name can be dropped by that rule; it errs towards leaving a company out rather than auditing someone else’s site.
• Who made it: ProUXAudit, a product of LEISSON OÜ, which sells the audit tool used here. The study was not commissioned or paid for by anyone else, and no result was reviewed by an independent expert.
• The quality bars used for “below” (85 mobile, 85 accessibility, 80 performance) are the thresholds ProUXAudit applies to every report.

Citing this study

Quote any number with a link to this page. Suggested citation:

ProUXAudit (LEISSON OÜ), “Estonian software publishers’ homepages, 2026”, 63 homepages measured 2 October 2026. https://prouxaudit.com/research/estonian-software-homepages-2026