Analytics Methodology

1. What This Page Explains

TrialConnect publishes activity figures — page views, trial views, match starts, contact clicks, and similar counts — to sponsors, researchers, and on our own reporting surfaces. This page explains what those figures actually measure, how they are calculated, what they leave out, and why a number quoted from TrialConnect might not match a number from another system.

It describes our own first-party product analytics, and how that differs from other measurement layers (like third-party web analytics tools, or our operational referral records) that also exist alongside it. It does not cover matching-algorithm accuracy, which is a separate subject.

2. Measurement Principles

We hold ourselves to a small number of rules when we publish a number:

• We only describe metrics that exist in our production analytics today, as implemented in code — not aspirational or planned metrics.
• Observed activity on TrialConnect is not the same as epidemiological prevalence, national treatment demand, or a confirmed pool of eligible patients.
• Activity recorded against a trial or sponsor is not the same as activity at a specific hospital or recruiting site, unless stated otherwise.
• We do not claim to have removed all bots, or all internal/staff traffic, from our counts.
• Consent choices reduce what we are able to count. We disclose that undercounting rather than working around it.
• A number on this platform describes recorded events, not a verified, step-by-step journey for a specific individual.

In short: these metrics indicate observed activity within TrialConnect under the stated measurement rules below — they are not a census of every visit or action, and they are not clinical or epidemiological evidence.

3. Metric Definitions

The metrics below are the ones we currently calculate from recorded product events. Each is defined by what is counted, when, and what is deliberately excluded.

Page views

Counts recorded page navigations across the site during the selected period. This is recorded regardless of analytics consent, because it carries no session or user identifier. Repeat navigations by the same visitor are not deduplicated into a "unique visitor" count — each navigation is counted individually.

Trial views

Counts recorded views of an individual trial's detail page. This is only recorded for visitors who have given analytics consent (see Section 4) — no trial view is recorded for a visitor who has not opted in.

Match starts and match completions

"Match starts" counts recorded starts of the TrialConnect matching questionnaire. "Match completions" counts recorded completions that produced a results screen. Neither represents confirmed eligibility, a submitted application, or enrolment in a trial — they describe engagement with the on-site matching tool only. Where we publish a match conversion rate, it is:

match conversion rate = match completions ÷ match starts, over the same time window

This is a ratio between two independently counted totals for the same period — not a tracked, one-to-one journey for a single visitor. See Section 8.

Contact clicks

Counts recorded clicks on a trial's contact or interest action. Where we publish a contact rate, it is:

contact rate = contact clicks ÷ trial views, over the same time window and grouping (e.g. per condition or per trial)

A contact click on its own does not mean a patient's email or identity was captured. See Section 7 for how this differs from a confirmed patient referral.

Patient interest (combined signal)

A combined count of six distinct on-page actions on a trial page: saving a trial, opening the contact form, emailing the site team, generating a draft enquiry, sharing a trial, and following the link to ClinicalTrials.gov. It is a sum of several action types, not a single tracked action — a visitor who performs more than one of these actions on the same trial is counted more than once.

Condition demand

For a given condition, the number of recorded page views, matching-tool starts, and trial-related actions associated with that condition during the selected period. This is a signal of activity on TrialConnect. It is not a measure of how many people have that condition, how much treatment demand exists nationally, or how many patients would be eligible for any given trial.

Trial / sponsor activity

For a given trial, the number of recorded contact-intent actions (contact form opens, "email site team" clicks, contact clicks, and draft-enquiry generations). Because our analytics events are recorded against a trial and a free-text sponsor field rather than a distinct facility identifier, this activity is attributed at the trial or sponsor level, not at the level of an individual hospital or recruiting site, and sponsor matching is a partial text match rather than a verified organisational registry. Treat it as indicative, not exact.

4. Consent and Coverage

Visitors choose between two options in our cookie banner: allow all cookies and analytics, or essential only. If a visitor chooses essential only, the product-analytics events behind the metrics in Section 3 — trial views, match starts and completions, contact clicks, and patient-interest actions — are never sent from their browser, and none of them are recorded for that visit.

A small number of anonymous counters (page navigations, and clicks on a general contact link) are recorded regardless of consent choice, because they carry no session identifier and cannot be linked to an individual or their activity elsewhere on the site.

Because most of what we measure depends on an active, current consent choice:

recorded analytics activity ≤ total platform usage

Visitors who decline analytics, who have not yet made a choice, or whose browser blocks tracking scripts are undercounted relative to actual usage — they are never secretly tracked to make up the difference.

5. Bot and Internal Traffic Handling

We do not currently run any automated bot detection, search-engine-crawler filtering, or internal/staff-traffic exclusion inside our analytics pipeline. Every event that passes our basic validation checks is included in every count, rate, and chart described on this page — whether it came from a member of the public, a crawler, an uptime monitor, or TrialConnect's own team browsing the live site.

We do apply request-rate limiting at the network layer to protect the platform from abuse, but that is a traffic-safety control, not a bot-classification or analytics-filtering step, and it does not remove any traffic from the figures above. If dedicated bot or internal-traffic filtering is added in future, this section will be updated to describe exactly what is excluded and how.

6. Why Numbers May Differ Between Systems

TrialConnect's own first-party analytics (described on this page) is one of several measurement layers active on the platform. We also run third-party web analytics tools (currently Vercel Web Analytics and Plausible Analytics) that operate independently, with their own definitions, their own approach to bots and consent, and their own reporting dashboards that we do not control and cannot fully describe here. Separately, our application database holds operational records — such as confirmed patient referrals — that are not analytics events at all.

Numbers from these systems will not always agree, for reasons that include:

• different consent and cookie handling between systems;
• different (or no) bot filtering — see Section 5;
• different definitions of what counts as a "view" or a "visit";
• different time zones or reporting windows;
• events that arrive late, are retried, or are blocked by a browser extension;
• an operational record (e.g. a referral) being created separately from, and later than, the web event that led to it.

Disagreement between these systems is expected, and is not automatically a sign that one of them is broken — they are measuring related but distinct things.

7. How Metrics Are Corroborated

We do not merge our different data sources into one blended "truth" number, because they measure different degrees of certainty about patient interest. Instead, we treat them as a hierarchy:

• Web activity signal — page views, trial views, match starts/completions and contact clicks, as defined in Section 3. This is the earliest and most anonymous signal, and the noisiest.
• Confirmed patient referral — a separate, distinct record created only when a patient voluntarily submits their email address and gives explicit consent to be contacted about a specific trial. This requires a deliberate, identifying action and is a stronger signal than a contact click; it is stored and audited separately from anonymous analytics events.
• Site or sponsor-confirmed outcomes — screening or enrolment status, where a referral has a recorded follow-up update. These come from operational follow-up records, not from web analytics, and are only available where that follow-up has actually taken place.

A contact click is not treated as equivalent to a referral, and a referral is not treated as equivalent to confirmed screening or enrolment.

8. Important Limitations

• Funnel and conversion figures are ratios between independently counted totals for the same time window, not a verified step-by-step journey for one visitor. The same person can appear in more than one step, and a drop-off between steps cannot be attributed to a specific cause.
• Where a rate's denominator is zero for a given period but some related activity was recorded, our systems display that rate as 100% rather than "no data". This is a known artefact of the calculation and is most likely to appear in low-volume periods or narrow breakdowns — treat rates built on small counts with caution.
• "Patient interest" (Section 3) sums several different actions, so one highly engaged visitor can contribute more to the total than several visitors performing a single action each.
• Trial and sponsor attribution comes from fields recorded on each event rather than a verified registry, so sponsor-level figures in particular should be treated as indicative rather than exact.
• No metric on this page implies confirmed trial eligibility, a completed application, or enrolment. Those determinations are always made by the trial's own team, never by TrialConnect's analytics.
• Bot and internal-traffic handling is limited to what is described in Section 5 — no broader filtering should be assumed.

9. Methodology Version & Contact

Methodology version: 1.0

Last updated: August 2026

This page is reviewed and updated when a metric definition described here materially changes. If a figure you have been given does not match what is described on this page, or you have a question about how a specific number was produced, contact us at admin@trialconnect.co.uk.

Related documents: Privacy Policy · Data Protection