Short answer: with ten or more clients, you only notice a broken setup when someone asks about the numbers, unless you monitor it. Watch four layers: is the server up, do requests go out, do they arrive at the platforms, and does the outcome match the backend. Signals for each layer already exist in Google Cloud, your host, GA4, Google Ads and Meta. The work is in bringing them together and keeping a fixed rhythm.
By the TrackLead team. Technically reviewed by Pieter. Updated: October 2026.
What breaks after go-live
A server-side setup that works at handover doesn’t keep working on its own. These are the usual causes:
The server can’t keep up. Google recommends at least two instances, so an outage doesn’t immediately cost data. When the maximum number of instances is reached, requests are refused, and those don’t show up in the standard request metric.
The custom domain. The certificate renews automatically, but that can fail, for example when a CDN intercepts the validation.
A publish in the web container. Someone changes a tag or trigger, and the event or parameters the server container expects no longer come through.
A change on the website. A new checkout or theme changes the data layer, so values such as order value go missing.
Consent. An update to the consent platform changes the signals that are passed on. That directly affects what Google tags in the server container may send.
A token. A Meta Conversions API access token from a system user can expire after 60 days, depending on how it was created.
The platform itself. Meta names platform updates and new privacy rules as causes of critical issues in Events Manager.
Four layers to monitor
The server: is the domain reachable, is the certificate valid, and does the server return no errors (5xx)?
Sending: do the tags in the server container fire, and do the platforms respond with a success status code?
Receiving: do GA4, Google Ads and Meta see the events arrive, with the right data and without double counting?
The outcome: do purchases or leads in the platforms follow the same line as in the store or CRM?
A green check on layer 1 says nothing about layer 4. A server can run fine while order value has been empty for weeks.
The signals you already have
Google Cloud (Cloud Run)
In Cloud Monitoring, create an uptime check on the server container’s /healthy path with certificate validation on. It returns ok when the server is running. Attach an alerting policy, and add a second policy on the number of 5xx responses. Notifications can go out by email, Slack or a webhook, among others. The time_until_ssl_cert_expires metric warns you before the certificate expires.
Your host
Stape has a Monitoring tab with successful and failed requests per platform, and email alerts from the Business plan up. TAGGRS includes an overview of incoming requests and the impact of tracking prevention in all plans.
GA4
Create a custom insight per property on the conversion that matters, such as a drop in purchase compared with the previous period. Turn on email notifications. For web data this can run hourly, and you get up to 50 custom insights per property.
Google Ads
Under Goals › Conversions › Summary you see the status per conversion action. A status saying there have been no recent conversions isn’t always an error: sometimes the clicks simply didn’t convert. The Diagnostics tab also shows how enhanced conversions are performing.
Meta
In Events Manager, the Diagnostics tab shows critical issues and warnings. An issue left for three days moves to Previously detected and loses its fix instructions. Event Match Quality gives a score up to 10 over the last 48 hours; Meta recommends 6 or higher. Event Deduplication shows whether Pixel and Conversions API events are merged correctly. If you manage many pixels, the Dataset Quality API returns these figures per dataset, so you can put them in one overview.
Compare with the backend, not with a norm
There is no official norm for how far GA4 or Meta may differ from the store. Consent explains part of it: in basic mode GA4 sends nothing when a visitor declines, while the backend counts every order. So set up a fixed comparison per client: same time zone, same definition of an order, per week. If the ratio suddenly changes, something changed on the site, in consent or in the setup.
Tracking changes across dozens of containers
Publish notifications: in GTM, under User settings › Default Container Notifications, choose to get an email for every published version, for all your containers at once.
Limit publishing: give the Publish permission only to those who need it. Edit lets you make changes but not publish, Approve lets you create versions but not publish.
Roll back: every publish is a version. If something breaks, publish the previous version again under Versions.
Google Ads per client: alerts in a manager account don’t mention conversion issues. A Google Ads script run from the manager account can check up to 50 accounts at once.
How to set it up
Make a list per client: server container URL, host, platforms, key conversion and who has access.
Put the server under watch: an uptime check on
/healthy, alerts on 5xx and certificate expiry, or your host’s monitoring.Set one alert per platform on the key conversion: a GA4 insight and a regular look at the status in Google Ads and Meta.
Turn on publish notifications for all GTM containers.
Plan a weekly round through Diagnostics in Meta and Google Ads, and a monthly comparison with the backend.
Agree who picks up what when an alert comes in, and ask the client to announce website changes in advance.
Still starting a setup? Read what to sort out before you start with server-side tracking and how to set up Consent Mode v2 first.
With TrackLead you see all your clients’ implementations in one central overview, instead of per host and per platform. Take a look at tracklead.eu.
Frequently asked questions
How often should I check a client’s tracking?
Monitor the server and the key conversions automatically, so you get an alert when something stops. On top of that, check the diagnostics in Google Ads and Meta every week and compare with the backend every month.
What difference between GA4 and the store is normal?
There is no official norm. Consent, ad blockers and differences in time zone or in what counts as an order always cause a gap. So watch the ratio per client over time: if it suddenly changes, something has changed.
Does monitoring in Google Cloud cost extra?
Very little. Per project, the first 50 GiB of logs per month and 1 million uptime checks are free, and Google won’t charge for alerting policies before 1 September 2027. Google does recommend turning off request logging for the server container to keep logging costs low.
Do I see conversion issues in the Google Ads manager account?
Not reliably. Alerts in a manager account cover things like payments, budgets and campaigns, not conversion tracking. Check the status of conversion actions per client account, or let a Google Ads script do it for you.