Judge dayparting by blocks of hours, not single hours

Dayparting cuts bids or switches ads off by the hour. Before cutting an hour, count the orders behind its ACOS: 300 orders a month spread over the 168 hours of the week is under two per hour. Group hours into blocks, compare each block with your break-even ACOS, not your target, and cut only the blocks that lose money.

An hour with two orders in it does not have an ACOS. It has an anecdote.

What Amazon gives you to work with

Amazon added schedule-based bid rules for Sponsored Products in November 2023. They increase bids at chosen times of day, days of the week and date ranges, and Amazon points to its hourly performance reporting to find the hours when campaigns perform best. So hourly data exists. The question is how much of it each hour actually holds.

How few orders an hour holds

Take a product that sells 300 orders a month through ads, and spread them across the clock:

With under two orders, one sale more or less can double or halve an hour's ACOS. That is the same problem as judging a keyword on three orders, worked through in when to raise or lower a bid, only worse. Four-hour blocks across the whole week hold enough orders to say something. Single cells of the week rarely do.

Compare a block with break-even, not with your target

An hour's job is to make money, not to hit the target. Suppose a product keeps 40% of its sales before advertising, so its break-even ACOS is 40%, and its target is 30%:

Both blocks miss the 30% target. Switching off the night block saves $100 a month. Switching off the morning block loses $50 a month, because at 35% ACOS it still makes money; it just makes less than you hoped. A dayparting rule written against the target cuts both. The break-even ACOS calculator gives you the line to judge against.

Both answers assume the night shoppers would not have bought from you at another hour anyway. No report shows that, so treat the saving as the most an Off block can give, not a guarantee.

Trim before you switch off

Off removes a block's sales along with its spend. A lower bid for those hours keeps some of the sales and lowers the price of the clicks. Amazon's schedule-based rules only raise bids, so on Amazon alone the way to get there is a lower base bid, raised in the hours that earn it. In Vendlen's Dayparting rule, each of the 168 cells of the week holds either Off, which pauses the ad group for that hour, or a modifier from −100% to +900%. A modifier is always applied to the ad group's baseline bid, so a $1.20 bid in a −30% cell becomes $0.84, and the next hour's cell starts from $1.20 again; cells never compound. Hours are read in the marketplace's time zone, not yours.

Cut hours only while the campaign is running expensive

An hour that loses money in a bad month may pay in a good one. Instead of one fixed schedule, you can keep two: a full schedule as the default, and a stricter one with hours off that applies only while the campaigns are running above an ACOS line.

That is how Vendlen Automation chooses between schedules. You fill in every cell yourself; Vendlen does not read hourly performance and does not decide which hours to cut. What it does decide, each marketplace day, is which of an Objective's Dayparting rules applies: the highest-priority rule whose Criteria pass on the Objective's combined performance over both the last 30 and the last 60 days. A rule with no Criteria always passes, so it serves as the fallback. A stricter schedule with a Criteria of ACOS at 60% or more therefore runs only while the Objective is that expensive over both windows, and the full schedule returns when it is not.

Where this still needs a person

The hours are your decision, and they rest on things a report does not show: whether night shoppers come back in the morning, whether a holiday shifts the week, whether a budget that runs out at noon is really a morning problem or a budget problem. An Objective with no orders yet has no ACOS for a Criteria to read; protecting a new product from spend that never converts is a job for loss rules, not the clock.

What to do with this

1. Download 30 to 60 days of hourly performance for one campaign and add each hour across all seven days together.
2. Group the 24 hours into four to six blocks, and count orders in each. A block with a handful of orders cannot be judged yet.
3. Work out each block's ACOS and compare it with the product's break-even ACOS, not the target.
4. For a block above break-even, try a lower bid first, or a lower base bid with increases in the hours that earn it; switch the block off only if it still loses money.
5. Re-check the blocks every month. If they only lose money in expensive months, make the cut conditional on the campaign's ACOS rather than permanent.

← All articles