​

Page health monitoring

Erdo checks live pages daily and reports changes in how visitors interact with the form and generate lead events. An alert identifies the measured change; traffic counts alone do not establish whether its cause is the page, the visitors, tracking or capture.

​

What the daily check measures

The default monitor counts distinct sessions for yesterday and the prior 14 days. Each counted session must also have a pageview in its measurement window. It reads two rates separately:

• First field interaction: form_start sessions divided by pageview sessions. The landing kit records form_start when the visitor first focuses a field or first types into one, before validation or submission. On a mouse or trackpad the lead modal puts the cursor in its first field on open; that is the page’s own doing, so it never counts — the visitor’s first keystroke there does. form_start does not prove that a visitor finished the form.
• Generated-lead evidence: generate_lead sessions divided by form_start sessions. A successful first-step partial capture can emit generate_lead; it does not necessarily mean the visitor completed every form step.

The monitor reports a rate more than 30% below the same page’s prior-14-day rate. A page needs at least 50 pageview sessions yesterday and 50 in the baseline. Comparing generated-lead evidence also requires at least 10 first-field-interaction sessions in each window. It reports counts behind the rates and names any window it could not judge. Missing generated-lead events call for investigation. Check captured lead rows or a persisted capture probe before concluding that the capture path is broken. Neither a form start nor its absence establishes capture success or failure.

​

CTA clicks and reaching the form

A CTA click records a click on the primary call-to-action. It does not prove that every historical click opened a form. Where available, modal_open reports form opening separately. For pages whose live behavior or opening events verify that the CTA exposes the form, the monitor also compares CTA-click sessions with sessions that recorded both a click and a form start. The comparison requires at least 10 CTA-click sessions in each window and at least 80% of baseline form-start sessions to also have a CTA click. Co-occurrence alone does not establish cause. Inline forms without this opening step are excluded from that comparison. A separate standing check looks for a step that has always had a low rate: over 14 days, at least 50 CTA-click sessions but fewer than 40% going on to start the form, with the same opening-behavior and co-occurrence checks. Erdo raises one organization-level proposal to test a simpler opening step against the current form. It preserves the live form until an experiment is approved. Campaign splits use their own denominators; small groups cannot establish a shared cause.

​

What you see

Findings appear in your attention feed, naming the page, measured stage, comparison window and counts. The daily monitor refreshes its existing finding while the condition persists and clears it when it no longer holds. Routine healthy checks and insufficient-traffic results stay in the run summary. Some staff workspaces also use a separate deterministic monitor with absolute rate floors. Those floors are separate from the default daily check above. The deterministic monitor counts each stage only within the same pageview-session cohort and measurement window. Its findings include the session counts behind each rate. When the baseline implies fewer than five recent outcomes, the finding calls out the limited evidence; it does not establish a conversion collapse or a broken capture path. Platform-installed deterministic recipes receive updates in place. Custom script edits and existing schedule, pause state, and measurement parameters are preserved. The upgrade also corrects the interpretation in legacy findings. It preserves their historical report, original confirmation time, counts, severity and open state. This correction claims no new measurement or recovery. In particular, traffic below the minimum sample leaves the condition unknown rather than proving that conversion recovered. The two monitors read existing attention items to reduce duplicate reports; that coordination is best effort and cannot guarantee one item across both. The finding prompts an investigation or a specific experiment proposal. The monitor does not pause ads or reduce spend on its own.

​

What you need to enable

The daily monitor is enabled for every organization. It needs your workspace’s page-events dataset; if analytics is not wired, or no page meets the traffic floor, it records that limitation and stops. It leaves the dataset’s default filters enabled so test traffic does not enter the comparison.

​

How it fits with A/B tests

An A/B test compares variants with each other. Page health monitoring compares a page with its own history, including pages that have no experiment running. An alert from either check is evidence to investigate before diagnosing a page or capture defect.