Explainer · Coverage
The map of what we watch
Eight failure modes. Watched daily. Across every property
Detection only matters if the coverage holds. ainpulse runs continuous checks across the eight failure modes where marketing data tends to break — from outright tracking loss to subtle composition shifts inside a steady aggregate. Here's the full map of what we watch, and where each signal applies.
What we watch
Eight failure modes, in detail
Each card is one bucket of the catalog. Number, name, one-line description, concrete example anomalies, and the data sources it applies to.
Tracking & data loss
A property stops sending data — broken tag, OAuth disconnect, paused account.
- "Zero sessions for 24 hours after a GTM publish"
- "Google Ads stops reporting any conversion actions"
Traffic volume
Sessions and users drop or spike outside the property's normal range.
- "Sessions dropped overnight — 4,200 → 540"
- "Unexplained 280% spike from a single referrer"
Channel composition
Mix shifts inside a steady aggregate — organic regression hidden by paid lift, or direct spike from broken UTMs.
- "Organic −34%, paid +29%, total sessions flat"
- "Direct traffic +340% overnight (UTM tagging broke)"
User engagement
Engagement rate or pages-per-session drops out of pattern.
- "Engagement rate fell from 62% to 41% (page reverted to draft)"
- "Pages-per-session dropped on mobile after a layout deploy"
Conversions
Conversion events stop firing, or conversion rate drops at the channel or account level.
- "Key event 'purchase' stopped firing on Thursday"
- "Paid CVR −45% sustained for two days"
- "Account-wide conversion drought — zero key events in 24 hours"
Revenue & AOV
Average order value drops or ROAS regresses on paid spend.
- "AOV dropped 28% — discount stuck in cart"
- "Campaign ROAS regressed 35% week-over-week"
Ad cost & performance
Campaign cost spikes or drops; CTR and CPC drift outside normal range.
- "Campaign spend up 220% in 24 hours (bid strategy change went wrong)"
- "CPC spike +180%"
- "CTR dropped 60% week-over-week (creative fatigue)"
Campaign-level signals
Per-campaign anomalies — conversion droughts, CVR drops within campaigns that are still active.
- "Active campaign hasn't recorded a conversion in 7 days"
- "Single campaign CVR dropped 50% while account-level held"
How one looks end-to-end
Tracking broke on Tuesday. Flagged Wednesday morning
Here's how a real failure surfaces. A team pushed a GTM update Tuesday morning. The 'purchase' conversion event silently stopped firing — the rest of the tracking kept working, sessions and traffic looked normal everywhere, but every order from Tuesday onward looked like it didn't happen. Without a watcher, that silence typically runs for days while campaigns keep optimizing against zero conversion data.
A hard stop doesn't need a long pattern. The next daily check flagged it.
Daily conversions · acme-store.com
Illustrative. A hard stop doesn't wait for a long pattern — the next daily check flags it.
Detected: 2026-05-06
Event 'purchase' has not fired in 24 hours vs a normal 90–145/day. Last fire: Tuesday 2026-05-05, 09:40.
A hard stop like this matches a tracking break — typically a tag manager publish or page change.
Silence in the data. Alert out the next morning. Same flow, every property, every day.
Every card on this page becomes a live check
Connect one GA4 property or Google Ads account — coverage starts the next day.
Where we stop
What we don't watch yet — and why
Coverage is a deliberate boundary, not a maximum. There are signals we deliberately leave alone — either because another category of tool covers them better, or because acting on them would mean making assumptions about your stack we shouldn't make.
Site performance
Core Web Vitals, page load time, server errors. Why: APM tools cover these — we'd add noise without value.
Cross-property comparisons
"Property A vs Property B" benchmarks. Why: every property has its own context; comparing creates false confidence.
Root-cause diagnosis
Why a metric moved — was it the GTM publish, the bid change, the feed shift? Why: every regression has its own causes from your stack. We don't know your stack.
Predictive forecasting
"Conversions will be X next week" projections. Why: forecasting requires assumptions about you that we shouldn't make.
Detection, not diagnosis. Coverage, not surveillance.
Sources
Connected to where the data lives
Google Analytics 4
Google Ads
Meta AdsGA4 and Google Ads are live today. Meta Ads is next.
Questions
Common questions
Now you know the catalog.
See it run on yours
Connect one property in under five minutes. Every card on this page becomes a live check on your data the next day.