Advanced Bidding & Budgeting

The day is not uniform

A flat daily bid treats 2am the same as 8pm. Shoppers do not behave that way — traffic, conversion rate and competition all move across the day, and a single bid quietly averages over all of it.

That is the easy half of the argument, and almost every article on dayparting stops there. The hard half is that your data is not uniform either. Splitting a week into hours turns one number you could trust into a hundred and sixty-eight you probably cannot, and most dayparting decisions are made on the difference.

This scroll is about doing it honestly: what genuinely varies by hour, where the data comes from, what Amazon’s own controls can and cannot do, and how to tell a pattern from noise.

Three things swing by the hour, and they do not swing together

It is tempting to talk about “good hours” as if a single quality rose and fell across the day. Three separate things move, and conflating them is the most common analytical mistake in the topic.

Traffic volume is when people are browsing. Conversion rate is whether those people buy. Competition — and therefore your cost per click — is when other advertisers are bidding hardest. An hour can be excellent on one and terrible on another.

The classic shape is a late-evening browse peak with high traffic, mediocre conversion and heavy competition, sitting next to a quiet weekday-morning window with a third of the volume and noticeably better economics. Raising bids on the busy hour because it is busy is buying the most expensive traffic of the day.

Where hourly data actually comes from

You cannot act on the hour without measuring the hour, and there are two sources with very different access requirements.

The advertising console now carries hourly performance reporting, which is enough to find gross patterns and is what most sellers should start with. The deeper source is Amazon Marketing Stream, a push-based feed that delivers hourly campaign metrics in near real time across Sponsored Products, Sponsored Brands, display and DSP.

Stream is not a report you download. It pushes messages to you, which is why it requires an Amazon Ads API integration and an AWS destination to receive them. That puts it out of reach for a seller working in the console alone and squarely in the territory of agencies, tools and in-house teams with engineering.

The 168-bucket problem

Here is the arithmetic nobody puts in front of you. A week has 168 hours. Split by hour and day of week — which you must, because Sunday 9pm is not Tuesday 9pm — and every week of data lands in 168 separate buckets.

A campaign taking 500 clicks a week averages three clicks per bucket. At a 10% conversion rate you would expect a sale roughly every third bucket. The resulting heatmap will be vividly patterned, and almost all of that pattern is randomness.

This is the same click-threshold discipline as the Green Belt, applied to time: you need enough clicks in a bucket to expect several sales before its conversion rate means anything. Most accounts never reach that per-hour and have to aggregate.

• Aggregate hours into blocks — morning, working day, evening, overnight — before you trust anything.
• Pool many weeks rather than reading one. A pattern that is real will still be there in a quarter of data.
• Judge at account or portfolio level first. A single campaign almost never has the volume.
• If a bucket has no sales, that is usually not a dead hour — it is an empty sample.

Your hourly data is incomplete, in exactly the way your daily data is

Amazon attributes a sale back to the click that caused it, and the attribution window runs 7 days for Sponsored Products and 14 for Sponsored Brands. A click at 9pm on Monday can collect its sale on Thursday and still be credited to Monday 9pm.

That is good news for the analysis — the sale lands in the hour that earned it — and bad news for reading recent data. This week’s hourly heatmap is missing most of its conversions, and it is missing them unevenly: the newest hours are the emptiest.

Read a heatmap built from the last few days and you will systematically conclude that your recent hours are your worst hours. They are simply your least-settled ones.

Whose hour is it, anyway?

Your reports are timestamped in one time zone. Your shoppers are not in one time zone.

A campaign covering the United States is being read in a single account time zone while its buyers span the continent, so a three-hour spread is baked into every bucket before you start. The 8pm peak you can see is a smear of four different local evenings, and any schedule you set fires at one moment for all of them.

This does not make dayparting useless — the aggregate curve is still real and still actionable. It does mean the edges are softer than the chart implies, so schedule changes on generous blocks rather than sharp single-hour cuts.

Amazon can do this natively now

Most writing on Amazon dayparting predates the feature and still tells you the platform cannot do it. That stopped being true in late 2023.

Schedule-based bid rules for Sponsored Products let you raise bids at specific times of day, on chosen days of the week, or across a date range. Separately, hours of day arrived for schedule-based budget rules, so budget can be lifted at chosen times too.

Both are configurable in the console or through the API, and both are free. If your needs are simple — protect the evening, push harder on Prime Day — you may not need anything else at all.

Master Sifu — Before you buy a tool for this, check whether the native rules already cover what you actually want to do. A great deal of dayparting is “spend more during the good window”, and Amazon will do that for nothing.

The increase-only trap

Now the constraint that changes how you have to set all of this up, and which is almost never stated plainly: Amazon’s schedule-based rules only ever go up.

The bid rules describe increasing bids at chosen times. The budget rules describe increasing budget at chosen times. The performance-based variety raises budget when a campaign is doing well. Across the whole native rules surface, the documented direction is upward — there is no scheduled decrease.

So “bid down overnight” is not a thing you can natively schedule. The way to express it is inverted: set your base bid at the low number you would want in the dead hours, then schedule increases into the hours that deserve them. Your baseline becomes the floor rather than the norm.

Budget rules come in two kinds

Worth separating, because they answer different questions. Schedule-based rules raise budget at a time you choose — a date range, a recurring window, a recommended shopping event such as Prime Day. You are betting in advance that demand will be there.

Performance-based rules raise budget when the campaign clears a threshold you set on ROAS, click-through rate or conversion rate. You are not predicting anything; you are reacting to evidence that the money is working.

For a seasonal peak you know is coming, schedule. For a campaign that might get hot and you will not be watching, use performance. They are complementary, and the second is the better default for anyone who cannot babysit an account.

Does this fight with dynamic bidding?

A fair question, and the answer is that they operate on different information rather than competing.

Dynamic bidding up and down already adjusts your bid by up to 100% in either direction, in real time, based on how likely a given impression is to convert. It is reacting to the individual auction — the query, the shopper, the placement — not to the clock.

Your hourly rule is a blunt instrument applied to a whole window; dynamic bidding is a sharp one applied to each impression, and it works from the bid your rule has already set. They stack. That is exactly why aggressive dayparting on top of aggressive dynamic bidding can produce costs well above what either alone would suggest.

Three levers, and they fail differently

Hour-of-day control is usually discussed as one thing. There are three, and choosing the wrong one is a common way to spend money on a real insight and get nothing for it:

• Bids — changes what you pay per click and therefore where you rank. Use when the economics of the hour differ. Fails by overpaying for an hour that was merely busy.
• Budgets — changes how long you stay live. Use when you are budget-capped and running out before the good hours arrive. Fails by doing nothing at all if you were never hitting the cap.
• Placements — changes where you appear. Top of search behaves like a different product at different times of day. Fails by stacking a large multiplier on top of an already-raised bid.

Find your windows with revenue per click

Optimise on efficiency, not on volume. The hour with the most sales is very often not the hour where each dollar works hardest, and sorting a heatmap by sales is how people conclude their peak hour is their best hour.

Revenue per click — conversion rate times order value — is the honest measure, because it is exactly the number that sets the bid you can afford. An hour with half the traffic and double the RPC deserves the higher bid, however unimpressive it looks in a sales column.

Day of week is usually the bigger effect

People reach for hour-of-day first because it feels more sophisticated. In most accounts the day-of-week signal is larger, more stable and far easier to prove — and it comes in seven buckets rather than a hundred and sixty-eight.

Weekend behaviour genuinely differs from weekday behaviour in most categories, and business-to-business products often show a weekday pattern so sharp that no hourly analysis is needed to see it. Start there. If a weekly pattern does not survive contact with your data, an hourly one certainly will not.

What a “dead hour” actually is

Before cutting the overnight window, be clear about what you are cutting. Low overnight conversion is often not a lack of intent but a shift in behaviour: people browse in bed and buy the next morning.

Attribution partly rescues you here, because the sale is credited back to the click that started it. But not entirely — a shopper who browses at midnight, thinks about it, and searches again at lunchtime is two separate clicks, and only the second one converts. Cut the midnight bid and you may be removing yourself from the beginning of a journey you were winning.

This is the case for easing off dead hours rather than switching them off. A reduced presence still starts journeys; a zero presence cannot.

Peak events: the one case for everyone

Even an account with genuinely flat hours has a use for schedule-based rules, and it is the least controversial application in the topic: known demand spikes.

Prime Day, Black Friday, Cyber Monday and a category’s own seasonal peak are dates where you can say in advance that traffic will be higher and competition fiercer. Amazon’s schedule-based budget rules include recommended shopping events for exactly this, so the budget lifts itself on the day without anyone remembering to do it.

This is also the safest kind of dayparting, because the prediction is not really a prediction. You are not claiming to know that Tuesdays at 3pm are special — you are noting that Prime Day is busy, which it is.

Master Sifu — Set peak-event rules weeks early, while you are calm. The failure on a peak day is almost never the wrong bid — it is a profitable campaign hitting its cap at eleven in the morning and nobody noticing until evening.

When dayparting is worth it

Dayparting pitfalls to avoid

Run the schedule without babysitting it

Hour-level control is not hard to decide and is genuinely tedious to operate. The strategy is a handful of judgements; the execution is a change every day, in both directions, across every campaign you manage — which is precisely the kind of work that should not be done by a person.

Start with Amazon’s native rules if they cover what you need. Reach for a tool when you want the things they do not do: scheduled decreases, placement control by hour, one schedule applied across many campaigns, or a heatmap you can actually read.

Do this in SellerMateDayparting →Hourly bid and budget schedules in both directions, with four allocation modes (daily, weekday/weekend, all weekdays, custom ranges), a heatmap of performance by hour and day, and an activity log of every change. The decrease Amazon’s native rules will not do, and the visualisation that tells you whether you need it at all.