How we source and verify Irish software data

Each vendor profile carries a Last verified date stamp. On that date, an automated verification pass re-checked the vendor's public website and pricing page and re-confirmed the structured fields the profile is built on. Editorial copy (the verdict, pros and cons, the long-form overview) is reviewed separately on a slower cadence and is not re-stamped by the weekly pass.

We re-verify every vendor's pricing and compliance signals every week. An automated check compares each vendor's live website against our stored data and flags any drift. Every change is reviewed by a human before publication - no field is updated unattended.

The fields covered by the verification pass are:

• Pricing. Entry-level price, pricing model (per-user, flat, usage, contact), currency, and whether euro pricing is guaranteed. Checked against the vendor's public pricing page.
• Irish compliance booleans. Revenue Online Service integration, PAYE Modernisation support, Irish bank feeds, SEPA Direct Debit, Irish IBAN support, VAT handling, Irish-business-hours support. Checked against the vendor's feature pages and Irish documentation.
• GDPR data residency. Whether customer data is stored in the EU, the US, or both regions.
• Free trial availability. Whether a free trial is offered and the terms attached to it.
• Legal entity. CRO number, status, and registered address for Irish entities, refreshed on a separate weekly basis via the official CRO API.

Where a value cannot be verified from the vendor's public surface, the profile carries a Data estimated or Data pending verification chip and we say so. We do not infer Irish compliance status for a vendor that does not publish it.

A head-to-head comparison page draws its data from two underlying vendor profiles. Each profile carries its own verification date, so the comparison is only as fresh as the older of the two. The "Last verified" stamp on a comparison page reflects the older date - if BrightPay was re-verified yesterday and Sage on the first of last month, the comparison reads as verified to the older Sage date.

This is deliberately conservative. A reader looking at a comparison should be able to trust that every field shown has been checked at least as recently as the date they see.

Some vendor profiles carry a Market momentum reading - a 0 to 100 score that tracks whether a tool is gaining or losing ground. It is mechanical: there is no editorial opinion and no AI in the figures. The score is computed from public signals we measure on a recurring schedule.

Three signals can feed the score. Hiring demand is the number of Irish job postings that mention the vendor, from Adzuna. Web popularity is the vendor's position in the Tranco web-popularity ranking, where a rank that falls towards 1 means a site is climbing. Uptime (90d) is the count of incidents the vendor published on its own public status page over the trailing 90 days, where fewer incidents reads as steadier reliability. Each signal is scored on its direction of travel against the previous reading - a flat reading sits at 50, an improving one scores higher, a declining one lower - and the ready signals are averaged.

The uptime signal is read from the vendor's own public status page (Statuspage or Instatus). We do not ping vendor endpoints ourselves or use any paid monitoring source - we only count what the vendor publishes. Where a vendor has no public status page we can confirm, the uptime reading is left absent rather than guessed. Each vendor with a tracked status page carries a reliability timeline listing every published incident, one citable anchor per incident.

We never show a number we have not measured. Until a vendor has two readings for at least one signal, the module says so rather than inventing a score, and each reading carries the month it was taken. The series are append-only: a revised figure overwrites the month it corrects and the change is logged. We are rolling this out to one cohort first and widening it as the data accumulates.

The Irish SaaS Adoption Index measures which software tools are DETECTABLE on Irish business websites, by category. This is website technographics from our own self-crawl, not a reused third-party feed and not a survey - see the distinction from our Eurostat-sourced adoption report below. There is no AI in it: every figure is crawl-extracted and computed by a pure function, and all the prose around it is template text with the numbers slotted in.

Where the data comes from

We run a self-crawl of an Irish-RESOLVED business seed list - domains attributable to Irish-registered businesses, each with recorded provenance (which dataset it came from and when it was resolved). Vendors' own marketing domains are deliberately excluded from the seed: they are software makers, not Irish adopters, and crawling them would measure the wrong population. As of this methodology's last update, no in-repo dataset (company records, Central Bank of Ireland register entries, or eTenders supplier data) publishes a website field we can resolve deterministically, so the seed does not exist yet and the crawl has not run - see scripts/intel/build-irish-seed.mjs for the exact, re-runnable check. The index activates the moment a seed of at least 100 resolved domains lands.

How we fingerprint a site

Each site's homepage is fingerprinted with the Wappalyzer open-source engine, which reads the public markup, scripts and response headers a browser already receives and matches them against a library of technology signatures. No paid technographics API is involved and no headless browser runs any JavaScript. For each tracked category we record the single highest-confidence tool detected on a site, so a site counts once per category.

What "n", "crawl base", "detected sites" and "share" mean

n is always the crawl base - every site the crawler successfully fetched that period, regardless of whether any tool was detected on it. Within a category, detected sites is the narrower count of crawled sites where a tool in that category was found at all; a tool's share is the number of sites running it as their primary tool divided by detected sites, not by the crawl base. Because a site with nothing detectable simply drops out of that denominator, a category's shares summing to roughly 100% is a property of this detection-conditional model - it is not evidence that every install was found. Both the crawl base and the detected-site count are published alongside every figure, so the base is never hidden and never conflated with the detection count.

Crawl dates, revisions and missing periods

The index is a quarterly, append-only series: each crawl writes one period per category, stamped with the quarter it was taken. A re-crawl that finds the same picture changes nothing on disk. When a recomputed period genuinely moves, it overwrites that quarter and the change is logged. A quarter that has not been crawled is left absent - never carried forward and never filled with a guessed number. Until the first crawl lands, the page says so rather than showing a fabricated share. A period more than 90 days old is flagged with a dated-data notice on the page rather than shown silently as current.

Not a substitute for the Eurostat survey

Vendors.ie also publishes Irish Business Software Adoption Statistics, built from Eurostat's official enterprise ICT usage survey with an EU27 comparison. That is survey-based official statistics; this index is a homepage technographics self-crawl. The two measure different things (survey response vs. detectable homepage technology) and are never presented as interchangeable - each links to the other.

The crawl has real, structural blind spots. We disclose them here rather than let the figures imply completeness they do not have.

• No JavaScript execution. The fingerprinter runs without a headless browser and does not evaluate or load JavaScript, so any tool injected by a tag manager, or that only appears after client-side rendering, is invisible to it - even if the tool is genuinely running on the site.
• Homepage only. Only the homepage is fetched. A tool that only appears on a checkout page, a booking page, or any URL other than the homepage is never seen.
• Self-hosted, reverse-proxied and CDN-fronted installs may be invisible. Wappalyzer's signatures match publicly-observable markup, scripts and headers; a tool that hides its footprint behind a generic proxy or a heavily customised front end can go undetected even though it is genuinely in use.
• Detection, not census. A tool that is not detected on a site is not proof the site runs nothing in that category - it may run a tool this crawl cannot see. That is why shares are of DETECTED installs, not of every site crawled - see "What n, crawl base, detected sites and share mean" above.

None of this makes the figures wrong for what they measure - they are exactly what a plain, unauthenticated fetch of a homepage exposes. It does mean the index cannot be read as a full census of Irish software use, and we say so on the page itself, not only here.

Pricing changed, a compliance flag is wrong, or a vendor has gone out of business in Ireland? Use our contact form with the vendor name, the field, and a public-source link if you have one. Corrections from vendors and from readers are treated the same - we cite the source, we date the change, and the profile carries the new "Last verified" date once the update merges.

Vendors who want to take ownership of their profile (and add product detail beyond the structured fields) can claim it. Claiming a profile does not change the verification rules - claimed profiles are held to the same evidence standard as the rest of the dataset.

Every vendor profile and pillar page also carries a provenance footer with the source URL checked during the last verification, the relevant Irish regulator for the software category, our operating entity, and a link to the live Irish compliance matrix. If you want to inspect the raw evidence behind any field, that footer is where to start.

Methodology last reviewed: 2026-08-13.