Guide · Monitoring
The Google Ads monitoring checklist
Written 8 October 2025 by Wout Peeters · thresholds re-tuned 19 July 2026
Every Google Ads account has five vital signs, and most accounts check none of them between Monday meetings. This guide is the full checklist we run as a service, laid out so an in-house team can build a serviceable version with scripts and rules for exactly zero euros. The parts that need paid tooling are labeled honestly.
Why monitor at all?
Because the failure math is asymmetric. An account improvement compounds over months; an account incident compounds over hours. The environment supplies incidents on schedule: 8.51% of paid traffic ran invalid in 2025 (an estimated $63 billion of global waste), Google removed 8.3 billion ads for policy reasons, and every site release your developers ship is a fresh chance to orphan a conversion tag. None of these announce themselves. The only question is whether the silence lasts 40 minutes or a weekend, and the €9,400 write-up prices the difference.
What are the five vital signs?
| Vital sign | Sane starting threshold | Buildable free? |
|---|---|---|
| Spend pacing | ±25% of the daily plan, per account | Yes, Ads script + email |
| Conversion heartbeat | 0 conversions for 3+ hours vs a baseline that says otherwise | Yes, with baseline logic in a script |
| Disapproval scan | Any new disapproval or limited status on an active ad | Yes, script over ad status |
| Landing-page health | Non-200 responses on top ad destinations | Partially; needs an external checker |
| CPC anomaly | 40% over the 28-day campaign average | Yes, script with rolling window |
The thresholds are starting points, tuned monthly. Accounts with lumpy conversion patterns (B2B with 20 leads a month) need wider heartbeat windows than a store converting hourly; a threshold copied blindly from a blog, this one included, is a false-alarm generator.
How do you build the free version?
- Pacing and heartbeat: one Google Ads script, scheduled hourly, comparing today against a 28-day same-weekday baseline and emailing on breach. This is an afternoon of work and covers the two most expensive failure classes.
- Disapprovals: a second script iterating ad status, because the interface tells you about disapprovals only when you visit. The cleaning franchise in this outcome had 11 of 84 ads silently dead at intake.
- Landing pages: any uptime checker pointed at your top ten ad destinations; the free tiers of most services cover this. Ads happily buy clicks to a 404 forever.
- Routing:alerts to a shared channel plus one named human per account, with a written "if no acknowledgment in 60 minutes" escalation.
The honest limits: scripts run per account and rot quietly when nobody owns them, cross-account rollups get painful past a handful of accounts, and 30-minute cadences with proper alert hygiene are where DIY stops being cheap. That boundary is where our €390 monitoring tier lives, and we would rather you build the free version than run nothing.
What does a weekly human hour add?
Judgment the scripts cannot fake. The weekly hour reads search terms against intent, checks that spend still tracks the plan in shape and not just size, and issues verdicts on running tests. Benchmarks anchor the judgment: the 2026 medians run $5.42 per click and $66.69 per lead across industries, and an account drifting away from its vertical's band has a findable reason. Monitoring keeps the account alive; the weekly hour keeps it aimed. Confusing the two is how agencies end up selling strategy decks to accounts with dead conversion tags.
Asked before signing, usually
Vitals continuously or near it, judgment weekly. Spend pacing, conversion heartbeat, disapprovals, landing-page health, and cost anomalies deserve automation on a sub-hour cadence, because incidents compound hourly. Search-term reading, test verdicts, and structural decisions deserve a scheduled human hour, weekly, with no phone nearby.
Google Ads scripts run on schedules and can email you about pacing, zero-conversion windows, and disapprovals; automated rules handle simple triggers; and Google's own notifications catch billing and policy events. The gaps are landing-page health and anything cross-account, which is where scripts get painful and paid tooling or a service earns its keep.
Anchor to baselines, never absolutes: spend pacing ±25% of the daily plan, conversions at zero for 3+ hours where the baseline says that's abnormal, CPC 40% over the 28-day average. Then tune monthly. An alert channel people mute is worse than no alerts, because it manufactures false confidence.
More than ever, because automated bidding amplifies bad inputs at machine speed. A dead conversion tag under manual bidding wastes money passively; under tCPA it actively retrains the account toward noise within days. Automation moved the risk from doing nothing to noticing nothing, which is a monitoring problem by definition.
At least two humans with authority to act, through a channel they actually watch, with an escalation path for silence. The saddest incident reviews all share one line: 'the alert fired, nobody saw it.' Routing is as much a part of monitoring as detection, and it is the part teams skip.
Rather not babysit the scripts?
The service version runs every 30 minutes, cross-account, with a human on the other end. €390 a month, no contract theatre.
Get a free strategy call