Why yesterday's "best hour to advertise" becomes next quarter's worst

Dayparting means adjusting how your ads are delivered, or how much you bid, based on the time of day or day of week. If your ads convert best between 7pm and 9pm, dayparting is the mechanism that lets you raise bids in that window, or shift budget toward it, or in some setups stop showing ads outside it entirely.

The standard way people arrive at a dayparting rule looks like this: pull a few weeks of performance data, break it down by hour, notice that one window consistently converts better or costs less than the rest, and configure the account to concentrate spend there. It's a reasonable instinct. The data shows a pattern, so you build a rule around the pattern and move on to the next problem.

The issue isn't that this analysis is wrong when you run it. The issue is what happens after you set the rule: it gets treated as a fixed, durable fact about the audience, something you've now 'figured out' about how this business's customers behave, rather than a snapshot of one specific slice of time.

When an analysis finds that 7-9pm converts best, it's easy to hear that as a statement about people: 'our customers are more likely to buy in the evening.' But that number is actually the joint result of several separate things, each of which can shift independently, and none of which show up in the dayparting rule itself.

Seasonal routines are one. A hypothetical example: a home-services business might see evening conversions spike in winter, when people are home from work in the dark and browsing on their phones, but see that pattern flatten out in summer when the same people are outside later and shopping earlier in the day. The 'best hour' from a January analysis isn't wrong in January, it's just not a year-round truth.

Who the platform is actually showing your ads to is another. Meta's and Google's delivery algorithms continuously adjust who sees your ads based on who's converting, budget pacing, and account-level signals. The audience the platform was reaching at 8pm three months ago is not necessarily the same audience it's reaching at 8pm now, even with an unchanged targeting setup, because the algorithm's own learning process shifts the mix.

Competitive dynamics are a third factor, and this one is easy to miss entirely because it's invisible in your own account. Ad auctions are relative: what you pay and how often you win depends on who else is bidding at that hour and how aggressively. If a hypothetical competitor starts running evening promotions and bidding up that same window, your previously-cheap 7-9pm slot can quietly get more expensive, while some other hour that used to be crowded opens up and becomes the better buy. None of this shows up as a change on your end.

Platform pacing logic is the fourth. Both Meta and Google periodically change how they pace budget delivery across a day, which can shift when in the day your ads actually get shown even if your targeting and bids haven't changed at all. A pacing update alone can move your effective 'best hour' without a single input on your side being touched.

Dayparting rules have a specific failure mode that other campaign settings don't: they tend to sit there unrevisited, because the work of figuring them out already happened. Once you've run the analysis and configured the rule, the question feels closed. There's rarely a trigger that brings you back to re-check it, the way a sudden CPA spike would force you to look at a campaign again.

That's the trap. The rule doesn't decay visibly. It keeps enforcing whatever the account looked like during the original analysis window, silently, indefinitely, while the actual pattern underneath it drifts. In the worst case the rule is now actively working against you: it's concentrating spend in a window that has quietly become worse, and starving a window that has quietly become better, and there's no alert for that because the rule is doing exactly what it was told to do. Nothing about the campaign looks broken from the outside. It's just optimizing for a version of your audience that no longer exists.

The correction isn't to stop dayparting. It's to stop treating a dayparting conclusion as a fact you only need to establish once. Re-run the same hour-by-hour analysis on a fixed cadence, roughly every quarter or two, and update the rule to match what the data currently shows rather than what it showed when the rule was first built.

If a business's patterns turn out to be genuinely seasonal, that's a better outcome than it sounds. It means the fix isn't one dayparting rule reviewed periodically, it's a small set of seasonal rules (a winter-evening rule, a summer-midday rule, whatever the actual pattern is) that get swapped in on a schedule instead of one rule left to quietly go stale for the other three seasons of the year. Either way, the underlying discipline is the same: a dayparting rule is a hypothesis with an expiration date, not a settled conclusion.