What is Dayparting?

Dayparting weights or restricts advertising delivery by hour of day and day of week. In its simplest form it switches campaigns off during chosen hours; in a more nuanced form it adjusts bids up or down by time block.

The idea is intuitive and the intuition is frequently wrong, which is what makes it worth a careful treatment. Not every hour performs the same, and it does not follow that the underperforming hours should be removed.

The first justification is a fact about the business. A sales team answers phones between nine and five, so leads generated at midnight sit until morning and convert worse. A restaurant takes bookings during service. A delivery promise has a cut-off. A category is restricted by law or by platform policy during certain hours.

These require no analysis. The schedule encodes something already known, and the only decision is where the boundaries sit.

The second justification is a performance report showing a worse cost per result in certain hours. This is the common case and it is where most dayparting mistakes originate, because hourly reports are among the least reliable in an ad account.

Splitting an account's data by hour divides it twenty-four ways. Splitting by hour and day of week divides it one hundred and sixty-eight ways.

Most accounts do not have the conversion volume to support conclusions at that granularity. An account with a thousand conversions a quarter has, on average, about forty per hour-of-day bucket and six per hour-and-weekday bucket. At those counts, the difference between a good hour and a bad one is frequently four conversions, and four conversions is noise.

The reports do not indicate this. They present a cost per conversion for 3am with the same confidence as the one for 3pm, and the human reading them sees a pattern because humans see patterns in small samples.

The remedy is to aggregate before deciding: group the day into three or four blocks, look at whether a difference persists across several months rather than one, and treat any conclusion resting on a handful of conversions as unproven.

Even where a genuine difference exists, switching an hour off has effects the report that prompted it cannot show.

Upstream contribution. Cheap impressions in low-competition hours fill retargeting pools and build recognition that converts later, during the hours credited with the conversion. Removing them lowers the cost of the retained hours on paper and the volume of the account in practice.

Compression. The budget does not disappear with the hours. It is now spent across fewer hours, which raises the spend rate in those hours, pushes further down the inventory quality curve, and typically raises the price per result. A schedule that removes a third of the day is an instruction to buy the remaining two thirds harder.

Signal loss. Automated strategies learn from what they observe. Hours never served are hours never learned about, and the restriction becomes self-confirming.

Where a campaign runs an automated strategy with adequate conversion volume, the bidder already varies what it pays according to predicted value, and time is one of the inputs to that prediction.

It does so at a granularity no manual schedule can reach: not per hour block, but per auction, combining time with device, audience, query, placement and everything else it knows. An hourly schedule layered on top overrides that with a much blunter instrument, and does so in one direction only — removal.

The defensible pattern with automated bidding is to use the schedule for hard business constraints and to leave performance variation to the model. Where the model does not have enough volume to be trusted, the account probably does not have enough volume to support hourly analysis either, which argues for leaving the schedule alone in both cases.

Hourly reporting is only meaningful if the time zone is understood, and this catches accounts regularly.

Platforms report in the account's configured time zone, which may not be the time zone of the customers, particularly for a business selling across a country with several zones or internationally. A schedule set against the account's zone applies uniformly, so a 9am start is 9am in one place and 6am in another.

Before any hourly conclusion is drawn, confirming the account time zone and thinking about where the customers actually are prevents a whole class of confident and wrong decisions.

Where the business case is genuine, a few practices make the implementation less costly.

Prefer bid adjustments to hard exclusions where the platform allows them. Reducing bids in low-value hours keeps the account present and cheap rather than absent.

Set boundaries slightly wider than the literal business hours. Somebody enquiring at ten minutes to five is still worth having, and hard cut-offs lose the edges.

Review the schedule on a cadence rather than leaving it forever. Business hours change, support coverage changes, and a schedule set three years ago tends to survive every other change in the account.

And keep the always-on segments out of it. Brand defence in particular rarely benefits from being switched off, because somebody searching your name at midnight is not going to be cheaper to acquire in the morning.

An account reports a cost per conversion of A$30 between 9am and 5pm and A$52 between midnight and 6am, and switches the overnight hours off.

The overnight window held 41 conversions across three months against 780 in the working day. Those 41 conversions cost 41 × 52 = A$2,132, so four more of them would have brought the figure to 2,132 ÷ 45 = A$47 — the whole apparent penalty is about five dollars, and it rests on a sample that four conversions move. Meanwhile the hours removed were also filling the retargeting pool that the daytime conversions were drawn from, and nothing in the hourly report shows that.