Every buying team runs a tracker, and every buying team has at some point discovered a problem days after it started. Those two facts are usually treated as a failure of attention. They are not. They are a property of where a tracker stands.
The click path
A tracker works by being on the route a click takes. Somebody sees an ad, clicks it, arrives on a page, and eventually converts; the tracker is threaded through that sequence, so it knows the click, the postback and the payout. This is a good position for the question it answers, which is what a campaign earned.
Now consider the events that decide whether an account keeps running:
- an ad is disapproved;
- a campaign is limited by a policy;
- billing fails;
- the account is suspended.
None of them produce a click. A disapproved ad does not serve, so nobody clicks it. A suspended account serves nothing at all. From the click path, these events are indistinguishable from a quiet day — and a quiet day is exactly what they look like in a tracker's dashboard, at the exact moment they are least quiet.
This is not a shortcoming of any particular tool. It is the difference between measuring an outcome and watching a system.
What the absence looks like
The failure mode is specific and worth naming, because it is how the days get lost.
Traffic falls. Someone notices, some hours later, that yesterday was down. The first hypothesis is always seasonality or creative fatigue, because that is what "down" usually means. Somebody looks at the creative, maybe rotates it. The next day is also down. Eventually someone opens the account itself and finds that half the ad group has been sitting disapproved since Tuesday.
The elapsed time between the event and the diagnosis is not attention. It is the shape of the question the tracker was asked: how much did we earn, not what state are the accounts in.
The second question
The other question needs a different vantage point — inside the account, reading what the account reports about itself: ad status, approval share, policy limitations, spend against yesterday, whether the account reported at all.
That last one is worth its own sentence. The most expensive account state is not "bad numbers", it is no numbers. A tracker cannot detect silence, because silence and zero look the same from the click path. Something reading the account directly can: an account that reported every hour for six weeks and then stopped is an event, and a loud one.
Both, not either
None of this is an argument for replacing a tracker. Revenue lives on the click path and always will — the payout is a fact only the tracker has, which is why account monitoring has to accept revenue from it rather than pretend to compute it.
The arrangement that works is boring: the tracker keeps answering what the campaign earned, something else keeps answering whether the accounts running it are still alive, and the two figures end up in the same row so that ROI is a column rather than a monthly reconciliation project.
What that costs is one integration each way — a postback pointed at the monitoring side, and a script inside the Google Ads account reporting out of it. What it buys is that the interval between "something broke" and "someone knows" stops being a function of who happened to look.