The most expensive distinction in paid search
Conversions fell. Is that a performance problem or did your tag stop firing? Getting this wrong costs more than any bidding mistake, and most dashboards do not help you tell.
Monday morning, conversions are down 40% week on week. What do you do?
The instinct is to look at bids, budgets, competition, seasonality — the things you can control. That instinct is correct roughly half the time. The other half, the number moved because the conversion stopped being recorded, not because it stopped happening, and every hour you spend adjusting bids is an hour of spend against data that is not measuring anything.
Why this is hard to see
Three things conspire to make a tracking outage look exactly like a performance drop.
Conversion lag. Conversions are attributed to the click, not the day they occurred. For an account with a seven-day consideration window, yesterday's conversion count is always wrong and always low. A dashboard that does not model lag will show you a "drop" every single morning.
Partial failures. A tag that breaks on one template — the checkout confirmation page after a deploy, say — takes out a slice of conversions rather than all of them. The line goes down, not to zero, which reads as performance.
Sync freshness versus sync coverage. A data pipeline that ran twenty minutes ago looks healthy. If that run returned no rows because an API call quietly failed, the pipeline is fresh and empty. Freshness is a measure of when, not of how much, and a system that only checks freshness will report a green light over missing data.
The checks that actually separate them
A real tracking outage has a signature that performance drops do not:
- Spend continues normally while conversions go to zero or near-zero. A genuine performance collapse usually shows up in click and impression volume too.
- The drop is abrupt rather than sloped. Competitive pressure and seasonality move over days. A tag breaks between one deploy and the next.
- It correlates with a change on the site, not a change in the account. If nothing changed in the ad account and everything changed in the numbers, look outside the ad account.
- Conversion action configuration changed. Someone marked an action secondary, or excluded it from bidding, or changed the counting method. This is a settings change wearing a performance costume.
Build the alarm before you need it
The reason this failure is so expensive is that nobody is watching on a Saturday. Alerting on a flatline — conversions at or near zero while spend continues — is the single highest-value alert you can configure, because it is the failure mode where every hour has a direct dollar cost and no recovery.
Set it up before you need it. You will need it.
Read next
Why your search term report hides the waste it is supposed to show
A search term report has thousands of rows and almost no repeats. The pattern costing you money never appears as a single line item — which is exactly why n-gram analysis exists.
Your ad test is probably reporting winners that do not exist
Comparing the best of five ad variants against a threshold designed for two comparisons inflates false positives dramatically. Here is the arithmetic, and what to do instead.
Make your next change an informed one.
Tell us about your accounts. We’ll help you explore the findings, the workflow and the results with a 14-day trial.
14-day trial · No credit card · Approval required by default