Dayparting: When Restricting Hours Helps

The short version

Dayparting means restricting delivery to chosen hours or days. The logic is intuitive — stop paying for hours that do not convert.

It is sometimes right. More often it removes volume without improving efficiency, because the hourly data it was based on could not support the conclusion.

Why the data usually cannot support it

Splitting a week into 168 hourly cells divides your conversions into groups small enough that most differences are noise.

A campaign with 200 weekly conversions averages roughly one per hourly cell. Any ranking of best and worst hours from that is describing randomness. Cut the "worst" hours and next week a different set looks worst.

Before dayparting, ask: how many conversions are in the hours I am about to cut? If it is single digits, you are acting on noise.

When it genuinely helps

A real business constraint. A sales team that only answers phones during business hours makes leads at 2am worth less. This is not a performance judgment, it is an operational fact.

A large, stable, explainable pattern. B2B advertising performing poorly at weekends is durable and has a reason. Trust patterns you can explain and that persist across months.

A hard capacity limit. Restaurants, appointment businesses, anything where demand outside operating hours cannot be served.

Compliance requirements. Some categories restrict hours by rule.

When it hurts

Attribution lag makes the pattern false. Someone sees an ad in the evening and converts the next morning. Last-click reporting credits the morning; the evening impression did the work. Cutting evenings removes the ads that created the conversions you attributed elsewhere.

It shrinks the auction. Fewer hours means a thinner auction and often a higher CPM in the hours you keep. The efficiency you gained by removing bad hours can be lost to the price of the good ones.

It fragments learning. Restricted delivery windows slow the accumulation of optimization events, which can keep a campaign in learning longer.

It hides the real problem. Poor performance in some hours is often a creative or landing page issue that also exists in the hours you kept.

How to do it if you do

Cut coarsely. Whole dayparts or whole days, not individual hours.
Require an explanation. If you cannot say why an hour underperforms, treat the pattern as noise.
Check it against a long window. Weeks, not days.
Watch CPM in the retained hours. If it rises enough to offset the gain, revert.
Prefer bid adjustments to hard exclusions where available. Reducing presence keeps the option open; removing it does not.

Practical rule

Daypart for business reasons, not for performance data, unless the pattern is large, persistent, and explainable. Most hourly performance differences are sample size, and cutting on them costs volume for nothing.