Dayparting on Amazon ads: how to set bids by hour

Why the hour of the click matters

Amazon runs an auction for every ad impression, around the clock. Your bid applies equally at every hour unless you change it. The shoppers behind those clicks do not behave equally across the day, though. Some hours bring browsers comparing options late at night. Others bring buyers who are ready to order.

When conversion rates differ by hour but your bid doesn't, you overpay in weak hours and underbid in strong ones. Dayparting is the practice of adjusting bids, and sometimes budgets, by time of day and day of week, so that what you pay for a click follows what that click is likely to earn.

What you will be able to do

• Get hour-level performance data for your campaigns.
• Tell a real hourly pattern apart from noise.
• Turn the pattern into a simple bid schedule.
• Measure whether the schedule improved results.

What dayparting is, and what it is not

Dayparting means changing how hard you compete in the ad auction based on the time. In practice that takes one of three forms.

• Bid adjustments by hour. Raise bids in hours that convert well and lower them in hours that don't. This is the most precise form.
• Budget timing. Make sure the daily budget is still available during your best hours, instead of running out by mid-afternoon.
• Pausing. Switching campaigns off in certain hours. This is the bluntest form, and the one most likely to backfire.

Dayparting is not a fix for weak campaigns. If a keyword doesn't convert at any hour, changing its bid at 2 am won't help. Clean up targeting, negatives and listing quality first. Dayparting then fine-tunes campaigns that already work.

Getting hourly data out of Amazon

Standard downloadable campaign reports summarise performance by day. That is fine for most decisions, but it hides exactly the pattern dayparting needs. For hour-level data, there are two routes.

Amazon Marketing Stream

Amazon Marketing Stream is an Amazon Ads API feed that delivers campaign traffic and conversion metrics hour by hour. It is the most complete source of hourly data. It requires API access, so sellers usually reach it through a developer or a tool that has already integrated it.

Manual snapshots

Without the API, you can approximate. Note cumulative spend, clicks and orders for the day at fixed times, say every three hours, and record the difference between snapshots. It is laborious and less exact, but a few weeks of it will show whether a strong pattern exists.

What to collect

For each hour, or block of hours, you want impressions, clicks, spend, orders and sales. From those you can derive the four numbers that matter:

Reading your hour-by-hour pattern

Collect at least two to four weeks of data before drawing conclusions. One week is too easily skewed by a sale event, a competitor running out of stock or a payday. Group hours into blocks if a single hour has too few clicks to be meaningful.

Illustrative numbers for one campaign over four weeks, at an average order value of ₹800.

Read the table for differences in conversion rate first, not spend. Spend simply follows traffic. Here, late night converts at under half the rate of the evening peak, while cost per click barely changes. That gap is what dayparting can fix.

Check that the pattern is real

• Does the same shape appear in each week on its own, or only in the total?
• Does the block have enough orders? A block with five orders can swing wildly on one sale.
• Do weekdays and weekends follow the same pattern? Many categories differ, and need separate schedules.

Turning the pattern into a bid schedule

A simple, defensible starting point is to scale each block's bid by how far its ACoS sits from your target. Because ACoS moves roughly with cost per click, this nudges each block toward the target.

Using the illustrative data from chapter 03 and a target ACoS of 25%.

Why cap the changes

The relationship between bid and results isn't linear. Cutting a bid by half doesn't halve ACoS. You may simply lose the placements that drove most of your orders. Moving in steps of 30% or less lets you see how each block responds before you go further.

Budgets, time zones and other traps

Most dayparting mistakes come from reading the data wrongly, not from the schedule itself.

Budget exhaustion looks like a weak evening

If a campaign runs out of budget at 4 pm, the evening shows zero spend and zero orders. That isn't a weak evening, it's a missing one. Check whether campaigns hit their daily budget before reading the hourly pattern. Raising bids in the morning can make this worse by spending the budget sooner.

Time zones

Ad data is reported in the marketplace's time zone. If you sell on a marketplace abroad, or if your tool reports in a different zone, make sure the hours in your analysis line up with the hours your schedule runs in.

Late conversions

Orders can be attributed to an ad click days after it happened. The most recent days always look worse than they will once the data settles. Leave the last week out of your analysis.

One schedule for everything

Different products have different buyers. A schedule built from a whole account's totals can be wrong for every individual campaign in it. Start with your highest-spending campaigns and build schedules for them one by one.

Testing whether it worked

A schedule is a hypothesis. Measure it before you roll it out further.

1. Pick comparable campaigns. Choose two similar campaigns, or split one campaign's keywords into two. Apply the schedule to one and leave the other on flat bids.
2. Run for at least two weeks. Longer if order volume is low. Avoid sale events and stock changes during the test.
3. Compare totals, not hours. Judge on total orders, total sales and ACoS across the whole day. Shifting spend between hours only helps if the daily result improves.
4. Iterate. If the test campaign wins, roll the schedule out to more campaigns and tighten the multipliers. If it doesn't, the hourly pattern may have been noise.

A dayparting checklist

Run through this before and after every schedule change.

• The campaign already converts and has clean targeting and negatives.
• There are at least two to four weeks of hourly or block-level data.
• The last week is excluded, so late conversions don't distort it.
• The campaign doesn't run out of daily budget before the evening.
• Hours are in the marketplace's time zone.
• The pattern repeats week by week, not only in the total.
• Weekdays and weekends are checked separately.
• Bid changes are capped for the first round.
• Weak hours are reduced, not paused.
• A comparison campaign runs on flat bids during the test.
• Results are judged on total daily orders and sales.

Done carefully, dayparting is one of the few ad changes that can cut spend and keep orders. Done hastily, it turns noise into rules. The difference is mostly patience with the data.