It's 11 P.M. on Black Friday: How to Read Hourly Amazon Data Before You Cut a Bid
The biggest spend decisions of the year get made at 11 p.m. on a few hours of data, usually alone. This post covers how to read those hours fairly, what to check when a campaign goes quiet, and when to add budget, hold or pull back.
Start with what those hours are. Spend and clicks post almost as they happen. The sales they earn are still arriving, because Amazon attributes an order back to the click that drove it, sometimes days later. So the newest hours on your screen are always missing sales they have already earned, and the worse an hour looks, the younger it usually is. That is the single fact this post keeps coming back to.
It matters more now because the night repeats. Black Friday is November 27 and Cyber Monday is November 30, but Amazon's sales window has stretched well past the four headline days, and many brands run deals from mid-November into December. No single hour decides your quarter, and the read you make at 11 p.m. tonight is one you will make again tomorrow night, and the night after. A written method beats a gut call because you will use it a dozen times before the window closes.
Quick Answer
- What to check first: before you cut a bid on a weak hour, confirm the hour's time zone, whether its sales have landed yet, whether the budget ran out, whether the Buy Box and stock are intact, and whether a deal started, ended or was suppressed inside it.
- The opportunity: most quiet hours have a fixable cause such as a spent budget, a lost Buy Box, a suppressed deal or a switched-off variation, so the demand is still there to catch tonight.
- The risk: sales are reported after the spend, so the newest hours overstate ACoS, and a cut made on them takes bids down on ads that may be converting.
- The limit: an hourly read cannot forecast the weekend, and the add, hold or pull-back table below is a default for your own targets, not a prediction.
- Our read: compare the hour with the same event day last year, change one thing at a time, and keep every change reversible.
The Weekend's Biggest Calls Get Made on a Few Hours of Data
Cyber 5 compresses a year of decisions into a handful of nights. A budget raise at 11 p.m. Friday or a bid cut at 6 a.m. Saturday moves more money than most weekly optimization passes, and it happens with a thin sample, a tired operator and nobody to ask whether the number on the screen is normal.
Hourly data is useful here and easy to misread for three reasons. Attributed sales arrive after the spend that earned them. The hour on your screen may not be the hour in the marketplace. And a flat hour has several causes that look identical in a report. The sections below take them in that order.
Why the Morning ACoS Looks Worse Than It Is
Spend and clicks post quickly. Sales and conversions post later and keep changing. An Optmyzr study that re-pulled the same campaigns over 17 days found clicks and cost were stable early, while impressions, conversions and sales were more susceptible to change. For the most affected 5% of campaigns, attributed sales grew by at least 18.75% after the first report. The study also says that, according to Amazon, it can take up to 14 days for conversion and sales data to settle. Amazon publishes no fixed number of hours after which a given hour is final, so treat any hour-count you hear as a rule of thumb, not a guarantee.
Treat spend as final and sales as provisional
The practical rule is to read ACoS on the most recent hours as a floor on sales and a ceiling on the ratio. The last two or three hours will always look worst. An hour that looked poor at 11 p.m. often looks better when you reopen it at 7 a.m., which is why the same hour should be checked twice before it drives a cut.
How long to give an hour before you trust it
Split the question by metric. Spend and clicks are close to final within the hour. Sales need longer, and Amazon does not say how much longer, so measure the lag on your own account before the event: pull Tuesday's hourly report on Wednesday morning and again on Thursday, and the difference between the pulls is your account's settling time. Until you have that number, use three defaults. Make no cut on hours younger than your next-morning reopen. Treat a day-old hour as settled enough for weekend decisions. And hold the scorekeeping until well after the event, since conversion data can keep moving for up to 14 days.
Then adjust for how your shoppers buy. Deal windows push people to decide now, so during Cyber 5 more orders usually land close to the click than in an ordinary week, which makes next-morning reads more trustworthy than usual for impulse-priced items. The opposite holds for considered purchases. If orders normally trail clicks by several days on your products, that gap does not vanish because it is Black Friday, but it should shrink. Ultimately, the data is relative: if a product is selling well at 2 p.m. but not at 2 a.m., and enough time sits between you and the 2 a.m. spend, the gap is not likely an attribution problem.
Widening the scope, your own history says which brand you are: the bigger the gap between the two pulls in the lag test above, the longer your hours stay provisional, and the more your event-night calls should lean on spend, clicks and the six-cause check instead of young ACoS.
Automation reacts to the same illusion
Rules that tighten bids when ACoS rises react to provisional numbers too. On a recent customer call, anonymised, an operator described automation over-reacting to a temporarily inflated ACoS after an event and continuing to behave defensively once the numbers had recovered. If a rule fires on same-day ACoS during Cyber 5, give it a lookback that includes settled hours, or pause it for the weekend and make the calls by hand. The Peak-Event Playbook covers how to set skills that act on conditions instead of clock times.
Read Every Hour in the Marketplace's Time Zone
A team in one time zone reading a marketplace in another will misplace the quiet hour. 11 p.m. in New York is 8 p.m. in Los Angeles, and the dip you blame on late-night browsing may be the dinner-hour lull in the region that buys most. Amazon's budget guide says ads stop when the daily budget is used up and return when the budget resets at midnight, so the hour straddling that reset behaves differently from the hours on either side of it.
Before comparing any two hours, confirm which time zone the report uses, convert once, and note the offset you applied at the top of your notes. Qore's Sales by Hour report shows hours in the marketplace's time zone, which removes the conversion step for Amazon, Walmart and Google accounts alike.
A Quiet Hour Has Six Usual Causes, and Only One Is Demand
When impressions or orders drop in an hour, work through the causes in order of how cheap they are to check. Demand is last because it is the only one you cannot fix tonight.
| Cause | What it looks like | Where to check | What to do |
|---|---|---|---|
| Budget ran out | Spend flat to zero after a fast early burn | Campaign status and budget report | Raise the budget if settled hours justify it |
| Buy Box moved | Impressions and orders fall on one product while others hold | Buy Box report, current state and reason lost | Fix the cause; do not raise bids into a lost Buy Box |
| Variation out of stock or off | Parent still spending, one child dead | Inventory and listing status | Shift spend to in-stock variations, restore the listing |
| A rule paused it | Clean stop at a round hour, no error | Automation and change history | Review the trigger, resume if it fired on provisional data |
| A deal started, ended or was suppressed | A spike then a cliff at the deal's edges, or a sudden quiet with ads still serving | Deals dashboard status and schedule | Fix eligibility if suppressed; compare deal hours only with deal hours |
| Demand dropped | Every check above is clean and the whole category is softer | Same event day last year, other products | Hold, and wait for settled data before cutting |
The budget ran out
Amazon says budgets are spent as quickly as possible, so a small budget can be gone within minutes during high shopper engagement. Spend can also run above the daily figure on some days: Amazon averages the daily budget over a calendar month, and a campaign can spend up to 25% more than its average daily budget on a given day. A campaign that shows "out of budget" at 2 p.m. is a sign of demand you were already winning, and it is the most common reason impressions fall to zero with no error message. If you searched for why your Amazon campaign is out of budget on Black Friday, this is the first place to look.
The Buy Box moved
Sponsored Products ads for a product generally stop serving when that product does not hold the featured offer, so a lost Buy Box shows up in your ad report as a quiet product. Raising the bid does nothing until the cause is fixed. A competitor's price move, a pricing rule that ran too low or too high, or a fulfillment problem can each take it. Check current state first, then the reason.
A variation went out of stock or was switched off
On a parent with several children, one dead variation can look like a weak campaign. Check coverage per child, not per parent. The out-of-stock guide covers how a stockout pulls ads, pricing and rank together.
A rule paused it
Automations that pause, cut or cap leave no error. The signature is a clean stop at a round hour. Open the change history before you open the campaign.
A deal started, ended or was suppressed
Deal windows reshape the hour around them. A Lightning Deal's start pulls a spike and its end leaves a cliff, and neither says anything about your bids: that is the schedule, not performance. A deal suppressed mid-event over a price requirement or stock goes quiet with no error in the ad console, because the ads keep serving into a listing that lost its badge. So check the Deals dashboard before the campaign. The same logic applies to reading the numbers: conversion during a badge window sets a baseline the rest of the weekend will not match, so judge deal hours against deal hours and full-price hours against full-price hours, this year and last.
Demand dropped
This is the diagnosis of exclusion. If the budget, the Buy Box, the stock, the rules and the deal schedule are all clean, and the same event day last year was softer at this hour too, demand moved. Cutting bids then is a defensible call, but make it on settled hours, not the last three.
Compare the Hour With the Same Event Day Last Year, Not With Yesterday
Yesterday is the wrong baseline on Cyber 5, because the weekend has its own shape: Friday and Monday rarely look like Thursday or Tuesday. The fair comparison is the same event day last year, matched hour for hour in the marketplace's time zone, so Black Friday 2026 is read against Black Friday 2025 and not against the same calendar date.
Two cautions apply. Last year's hours have fully settled and this year's have not, so compare spend and clicks first and sales second. And if your catalog, prices or budgets changed materially since last year, say so in your notes before reading the gap as good or bad. For deeper anomaly checks outside event days, see the anomaly detection guide.
A Cyber 5 Decision Table: Add Budget, Hold or Pull Back
The table below is a default for the weekend. The thresholds are yours: your target ACoS, your margin, your inventory position. Nothing here names a number because the right one depends on the account.
| Call | Signals that support it | Signals that argue against it |
|---|---|---|
| Add budget | Budget ran out early; older, settled hours at or better than your target; a live deal converting with stock to cover it; Buy Box held | The evidence is only the last two or three hours; stock is thin on the products that would absorb the spend |
| Hold | Recent hours look weak but sales are still landing; spend is pacing; no cause found in the six-cause check | Older hours are also over target after the same event day last year is factored in |
| Pull back | Settled hours stay over target with a confirmed conversion drop; the Buy Box is lost or stock is nearly gone on that product | The cause is fixable tonight, in which case fix it and keep the bid |
Pulling back on one product is not the same as pulling back on the campaign. Cut where the cause is confirmed and leave the rest running.
If It Is Not Yet Thanksgiving: An Hour of Prep Replaces the 2 A.M. Guess
Everything above gets faster if four things exist before the window opens.
- The baseline, exported and converted. Pull last year's same-event-day hours now, convert them to the marketplace's time zone, and note at the top of the sheet which time zone the hours are now in. At 11 p.m. you want to read a comparison, not rebuild one.
- Your account's reporting lag, measured. Run the two-pull test from the section above this week, so "too young to trust" is a number and not a feeling.
- Triggers, not timers. A timer wakes you at 2 a.m. to find nothing wrong. A trigger fires only when a condition you wrote in advance is met: budget 80% consumed before evening, a Buy Box lost on a deal ASIN, two consecutive hours of zero impressions. Set alerts on conditions and let quiet nights stay quiet. The Peak-Event Playbook shows how to build these as conditions rather than clock times.
- The pre-mortem, run. Our tentpole pre-mortem covers the regrets sellers report after event weekends; deciding the thresholds now is what makes the decision table above usable at midnight.
Dayparting, Marketing Stream and AMC: Where Hourly Data Goes After the Weekend
Three adjacent tools come up whenever hourly data does, and only one belongs in the event-night loop.
Dayparting is the strategic payoff of hourly history, and the event weekend is the wrong week to start it. Amazon's console offers budget rules that can schedule increases for event days, but no native hour-by-hour bid scheduling; dayparting runs through the API or tools built on hourly data. Build hour-of-day rules from weeks of settled history, and switch them off for Cyber 5 itself, because the weekend's traffic shape matches no ordinary week. The settled hours you collect this weekend become next year's dayparting input.
That raises the fair question of what guards the account hour by hour during the event, if not a daypart. Three things. Budget headroom first: lift budgets so the clock is not what decides when you go dark, and schedule event-day increases with Amazon's budget rules instead of editing at midnight. Condition triggers second: the alerts from the prep section bring a person in when a line is crossed, which is what a daypart was approximating anyway. And one clock still matters: know when the midnight budget reset lands in marketplace time, and read the hours on either side of it as different things. Bids stay set to your margin all weekend; the hour-by-hour throttle during Cyber 5 is budget and attention, not a bid schedule.
Amazon Marketing Stream is the hourly source most tooling is built on: it pushes campaign metrics in near real time and then restates them as conversions attribute, which is the same lag this whole post is built around. If your dashboard is Stream-fed, the freshness is real and the provisionality is too.
Amazon Marketing Cloud holds event-level data with hourly timestamps, which makes it the place for the post-weekend read: how deal hours converted against full-price hours, which hours drove new-to-brand orders, what the path to purchase looked like across the window. It is a lookback tool, not a live one, so leave it out of the 11 p.m. loop and book it for the week after.
Ask the Account Instead of Staring at It
The six-cause check is six lookups: hourly sales, budget status, Buy Box state, stock by variation, automation history and deal status, each in a different screen and compared with last year. By 11 p.m. on Friday it is also the step that gets skipped, because it takes longer than the decision feels like it deserves.
Qore from Trellis is where you turn a review you do by hand into a workflow that runs itself, with Q, its assistant, helping you shape your own standard. Four pieces apply to this weekend:
- Sales by Hour report. Hourly results in the marketplace's time zone, in the Reports hub that is available on every Amazon, Walmart and Google Ads account.
- Buy Box report. Each product's current state, with a Why You Lost It filter that gives the reason.
- Q in Slack. Ask the account a question in a channel, and approve or reject a proposed change in the thread.
- Q by text. The same question from a phone, without opening a laptop.
Actions default to manual approval, so a staged budget change waits for your yes, and autonomy is available once you trust a skill. If you run dynamic pricing with our platform Qinetix, pricing and ads work as parallel mechanisms with shared visibility, which lets you see a price move next to a Buy Box change. Each runs on its own logic.
| Question | By hand | With Qore and Q |
|---|---|---|
| What did this hour do, in marketplace time? | Export, convert time zones, compare in a sheet | Sales by Hour report in the marketplace's time zone |
| Did a product lose the Buy Box, and why? | Check each listing in Seller Central | Buy Box report with current state and Why You Lost It filter |
| Approve a budget change from your phone | Log in and edit the campaign | Approve or reject in a Slack thread, or ask by text |
| Decide whether to chase volume or protect margin | Your call | Still your call; Q can lay out the data but will not set your strategy |
| Be ready by Friday night | Nothing to set up | Q in Slack and by text need setup by a user first, so do it before the weekend |
The reports show you the state and the reason, and Q can stage a change, but the add, hold or pull-back call is still yours. Qore is in open beta, and you can sign up here. To see how the reports and Q fit together, start with Q and dashboards and reporting.
Conclusion
An hourly number at 11 p.m. is a provisional number. Read spend as final and sales as provisional, put every hour in the marketplace's time zone, run the six-cause check before you touch a bid, and compare against the same event day last year, deal hours against deal hours. Add budget where settled hours support it, hold where the evidence is only recent hours, and pull back only where a cause is confirmed. Setting up your baseline, your triggers and your Q access before Thanksgiving is the cheapest preparation you can do, and the peak season guide covers the planning around it. Because the sales window runs longer than four days, repeat the same checks each night rather than only on the headline days.
Frequently Asked Questions
Spend posts quickly and sales post later, so the newest hours are missing sales that have not been attributed yet. Reopen the same hours later in the day before you cut anything. Optmyzr's study found that sales and conversions settle more slowly than clicks and cost.
Read hourly results in the marketplace's time zone. Qore's Sales by Hour report shows them that way. In any other tool, confirm the report's time zone setting before you compare hours, because Amazon's budget guide does not state which time zone the midnight reset follows.
The usual causes are a spent daily budget, a lost Buy Box, an out-of-stock or switched-off variation, a paused automation rule and a deal that ended or was suppressed. Check them in that order before assuming demand fell.
A drop to zero with no error most often means the budget ran out, the product lost the Buy Box or a rule paused the campaign. Amazon notes that budgets can be exhausted within minutes during high shopper engagement.
Raise budget before you raise bids if the campaign is running out early and settled hours are at or better than your target. Raising bids into a lost Buy Box or thin stock spends money without earning sales.
Amazon publishes no fixed number of hours. Optmyzr's study, citing Amazon, says conversion and sales data can take up to 14 days to settle, with clicks and cost settling much faster. Measure your own account's lag by re-pulling the same day's hourly report a day later, and make no cut on hours younger than your next-morning reopen.
Compare each hour with the same event day last year, matched in the marketplace's time zone, and compare spend and clicks before sales. If your catalog, prices or budgets changed since last year, account for that before reading the gap.
