The first thirty days in a new ad account
Every decision made before data accumulates is a guess. Which metric to read in the first month, which change to postpone, and why the learning phase shouldn't be interrupted.
The first month of a new account runs on different logic from every month after it. The reason is simple: you do not yet have data that can decide anything, so every “optimisation” made in this period is really a guess. Guesses are not the problem — mistaking a guess for a measurement is.
Week one: is the setup correct
In week one you do not look at performance, you look at measurement. Are conversion events firing correctly, is the same event counted twice, are server-side events arriving late? Until those are answered, no cost figure you read is real.
- Make one test conversion and confirm it appears in the dashboard exactly once.
- Make sure campaign, ad set and ad names follow a scheme; reporting structure gets built in week one or never.
- Check that UTM parameters are attached to your links and that they land in site analytics.
Week two: do not interrupt the learning phase
Performance becomes visible in week two and the urge to intervene peaks. The best thing to do that week is usually nothing. Budget changes, targeting changes and creative changes can each push a campaign back into learning; doing all three in one week means the campaign never learns at all.
Week three: which metric to read
Conversion volume in the first month is rarely enough to decide anything. Intermediate metrics that arrive earlier and in greater numbers are more useful here: click-through rate, time on page, add-to-cart. They are not the goal, but they tell you early where the funnel is blocked.
Just do not confuse the intermediate metric with the final one. High click-through and no sales means the problem is the page, not the ad; low click-through with buyers among the clickers means the problem is the creative, not the targeting. Making that distinction is the real work of the first month.
Week four: write down what you postponed
At the end of the first month you should have two lists: what has been verified, and what has been postponed. The second is the more valuable one — anything filed under “I will look at it when there is enough data” never gets looked at unless it is written down. This is also where automation rules start: any earlier and you would be setting thresholds against a normal that does not exist yet.