Your Amazon ACOS from yesterday is not final. Neither is last week’s Sponsored Products. Amazon PPC attribution runs on a delay. The platform keeps updating sales and order data against old dates for up to 28 days after the fact. This means every report you pull is a snapshot of a number still moving. Sellers who don’t understand how Amazon PPC attribution actually works cut winning campaigns. They panic over Monday mornings. And they chase ROAS targets that were never going to hold still long enough to hit.

TL;DR

• Amazon’s attribution window runs from 7 days to 14 days. The 7-day window is for Sponsored Products in Seller Central. The 14-day window is for Sponsored Brands, Sponsored Display, DSP, and Amazon Attribution. So “today’s” ACOS is provisional until that window closes.
• Amazon restates conversion data at 1, 7, and 28 days after each sale. A report you pull this week will look different if you pull it again next month.
• Amazon’s own documentation states a 10 to 20 percent discrepancy is normal. This is the discrepancy between Attribution data and outside analytics. It is not a tracking failure.
• Unified Reporting went generally available on June 8, 2026. And the legacy Sponsored Ads and DSP reports pages retire on December 31, 2026.
• Attribution tells you which ad got credit for a sale. It does not tell you whether that sale would have happened without the ad. Mixing up the two is the most expensive mistake in PPC reporting. And it’s the one this guide spends the most time on.

What Is Amazon PPC Attribution?

Amazon PPC attribution is the rule set Amazon uses. This rule set decides which ad interaction gets credit. The credit is for a sale, an add to cart, or a detail page view. For Sponsored Products, Sponsored Brands, and Sponsored Display, that rule is last touch. This means whichever qualifying ad a shopper clicked most recently before buying gets the credit. This happens as long as the purchase happens inside that ad type’s attribution window.

That’s the whole mechanism. It sounds simple. And the mechanics genuinely are simple. What trips sellers up is one assumption. They assume “attributed” means “caused.” It doesn’t. A shopper can search your brand name. That shopper can click a Sponsored Products ad out of habit. That shopper would have bought anyway. Still, it counts as an attributed sale. We come back to that distinction later in this guide. Because it changes how you should read every ACOS number you look at.

Amazon PPC Attribution Windows by Ad Type

This is the table most sellers actually need and can never find in one place. Amazon publishes the windows separately across different help pages, by ad type, by account type, so most PPC managers end up with half the picture.

Two things in that table matter more than they look at first glance. First, Sponsored Products runs a shorter window for sellers than it does for vendors. If you sell through Seller Central and also manage a Vendor Central account for the same brand, you are not comparing the same math. Or if you’re benchmarking your own performance against a vendor’s published ACOS, you’re also not comparing the same attribution math. Second, Amazon Attribution’s 14-day window sits outside the Sponsored Ads console entirely. It measures off-Amazon marketing. It does not measure your Sponsored Products campaigns. Which brings us to the next confusion.

Amazon PPC Attribution vs. Amazon Attribution

These are two different products with almost the same name, and a lot of published content treats them as interchangeable. They’re not.

Amazon PPC attribution, the subject of this guide, is how Amazon credits sales inside Sponsored Products, Sponsored Brands, and Sponsored Display. It happens automatically. You never set anything up.

Amazon Attribution is a separate, free measurement tool. It tracks what happens on Amazon after a shopper clicks an ad you ran somewhere else. That somewhere else can be Google, Meta, TikTok, an influencer link, or email. You generate a tag. You place it in the external ad’s destination URL. And Amazon reports the resulting clicks, detail page views, and purchases back to you. It runs on its own 14-day, click-based, last-touch model. This model is separate from your Sponsored Ads reporting.

If you’re already sending traffic off Amazon and tagging it, this matters. Eligible brands can earn an average of roughly 10 percent back. This is on qualifying sales driven by that traffic. You earn it through the Brand Referral Bonus program. Amazon enrolls this program directly through Amazon Attribution. Most sellers running external traffic have never turned it on.

Why Yesterday’s ACOS Is Fiction

Here’s the Amazon PPC attribution mechanic behind the number that looks wrong every single morning.

Amazon logs a sale against the date of the ad click, not the date of the purchase. If a shopper clicks your ad on a Tuesday and buys the following Sunday, that sale shows up under Tuesday’s report once the order clears, not Sunday’s. Pull Tuesday’s report on Wednesday and you’ll see a fraction of what Tuesday actually earned. Pull it again the following week and the number will have grown, because more of the attribution window has closed.

We break down that full restatement timeline, with a worked example of an ACOS shifting three separate times over the same 28 days, in the Amazon ACOS attribution window guide. The short version: click and impression counts can also shift for up to 3 days after the fact, while Amazon filters out bot traffic and accidental double clicks. None of this is a bug. It’s how Amazon’s own reporting documentation describes the process. Practically, this means you base any decision on data from the last few days on incomplete information; the newer the data, the less complete it is.

The Monday Mistake: Reading Weekend Data

Monday morning is the worst time to look at your dashboard, and it’s also when most sellers open it first.

Weekend orders take longer to settle than weekday orders. Amazon’s 72-hour window for canceled or failed payments hasn’t closed yet by Monday morning for a Friday or Saturday sale. And the 3-day click validation hasn’t finished running against Sunday’s traffic either. So the weekend numbers you’re staring at on Monday are the least settled data of the entire week. They arrive at the exact moment most PPC managers do their weekly review.

Our rule with clients: never react to weekend data before Wednesday. Judge a new exact-match keyword only after it has accumulated enough clicks to mean something. And give ranking-focused campaigns weeks, not days, before deciding whether they worked. A bid cut made Monday morning off a weekend that hasn’t finished reporting is usually a bid cut made on bad information. We go deeper into the exact day-by-day pattern in the Amazon PPC weekend data guide.

View-Through vs. Click-Through: What Your Report Quietly Misses

Not every ad interaction that leads to a sale is a click.

When Amazon Ads attribution decides which ad gets credit for a purchase, it runs through a hierarchy. Clicks on the exact product you advertised come first. Then clicks on a closely related product from the same brand. Then ad views, if the campaign qualifies for view-based credit. There are broader brand-halo views further down the list. A shopper who sees your Sponsored Brands ad, doesn’t click, browses elsewhere, and buys the product an hour later can still generate a view-through conversion, depending on the campaign type and whether it bills on a viewable-impression basis.

That hierarchy changed on January 1, 2026. Amazon retired the fixed 14-day view-through window for in-store placements. This applied to Sponsored Brands, Sponsored Display vCPM campaigns, and DSP. And it replaced it with what Amazon’s own announcement calls a shopping-signal-enhanced last-touch model. Click attribution across every ad type is unchanged. What changed is how conservatively Amazon now credits a view that never turned into a click. Amazon still exposes the old-style figures as “all views” metrics inside Unified Reporting. So you can compare against 2025 benchmarks. But your default ROAS on view-heavy campaigns may look tighter this year for reasons that have nothing to do with your account. We walk through a worked dollar example of a missed view-through sale. And what to check before assuming a ROAS drop is a campaign problem. That is in the view-through attribution guide.

The Amazon Halo Effect

Here’s a sale your ACOS report will never show you credit for, and it happens constantly.

A shopper clicks your Sponsored Products ad for one item, leaves without buying, comes back two days later through organic search, and buys the product directly from the listing. No ad interaction happened on the return visit. That sale gets zero ACOS credit, even though your ad is the reason the shopper found the product in the first place.

Run this the other direction and you get the halo effect properly. A shopper clicks your ad for one product. Then they buy a different item from your catalog instead, or in addition. Amazon does credit that one to the ad, under its brand-halo rules. This is part of why your account-level ACOS can look better than any single campaign’s ACOS on paper. The Purchased Product Report is the one place that shows you what customers actually bought after clicking. It does not just show what you advertised. And it’s the most underused report in the entire Sponsored Products console.

Put the two effects together and the pattern is clear. Accounts that manage purely to the reported ACOS number are systematically undervaluing their highest-performing spend, because the sales that ad is quietly generating through return visits and cross-catalog halo never show up in the campaign that earned them. It’s one of the more common Amazon PPC data accuracy blind spots we see in new client accounts. We break down how to actually mine the Purchased Product Report for this in the Amazon PPC halo effect guide.

How Long to Wait Before Judging a PPC Change

There isn’t one universal answer, because the honest answer depends on what changed and what type of campaign you changed it in.

A new exact-match keyword needs enough clicks in your Amazon PPC conversion tracking to be statistically meaningful before you touch its bid. It does not need a fixed number of days. A ranking-focused campaign needs weeks of data. This is the kind you’re deliberately running at a higher ACOS to build organic position. The payoff shows up in organic rank, not in the campaign’s own ACOS. A straightforward bid adjustment on an established, high-volume keyword can usually be judged inside a single attribution window once it closes. This is because the sample size is already there. Treating all three the same way is how good decisions get reversed for bad reasons. And checking all three on the same Monday-morning schedule is also how good decisions get reversed. We built a full framework by campaign type and change type in the Amazon PPC change timing framework.

The January 2026 Attribution Model Change

Worth restating plainly, because a lot of PPC content published before 2026 is already out of date on this point.

Amazon retired the fixed 14-day view-through attribution window for in-store ad placements on January 1, 2026. And it replaced it with the shopping-signal-enhanced last-touch model described above. It applies to Sponsored Brands, Sponsored Display vCPM campaigns, and DSP in-store placements. Click attribution was not touched, for any ad type. Industry estimates on the scale of the change vary. And we’re not going to repeat a precise percentage here without Amazon publishing one itself. What we will say plainly: if your Sponsored Brands or vCPM ROAS looks worse in 2026 than it did in 2025, check one thing first. Check whether you’re comparing the new default metric against last year’s view-inclusive figure before assuming the campaign got worse. The “all views” columns inside Unified Reporting exist specifically so you can make that comparison correctly instead of guessing.

Amazon’s Own Discrepancy Benchmark for PPC Reporting Accuracy

If you’ve ever compared Amazon Attribution’s click count to what your publisher or ad platform reports, and the numbers didn’t match, you’re not doing anything wrong.

That range is the benchmark. A gap inside 10 to 20 percent is normal measurement variance, not a broken tag. A gap that shows up suddenly, jumps well past 20 percent, or grows steadily over weeks is the actual signal to investigate, and the first places to check are your date ranges, whether every ad variant has its own tag, and whether the same tag got duplicated across two campaigns by accident. Amazon also recommends waiting 1 to 2 days after launch before validating new tags, since reporting takes time to populate.

What Amazon PPC Reporting Accuracy Actually Means

“Accurate” is doing a lot of work in this guide, so it’s worth being specific about what it actually covers, because Amazon PPC reporting accuracy is not one thing. It’s five separate questions, and a report can pass some of them while failing others.

• Metric accuracy asks whether Amazon recorded the click, the impression, and the spend correctly.

• Attribution accuracy asks whether credit was assigned according to Amazon’s published rules, the windows and hierarchy covered earlier in this guide.

• Temporal accuracy asks whether the sale landed in the right reporting period, which is the entire problem behind the restatement timeline above.

• Scope accuracy asks whether you’re actually looking at the right marketplace, campaign, and account type, not a blended mess of Seller Central and Vendor Central data.

• Causal accuracy, the hardest one, asks whether the ad caused the sale at all, which is the attribution-versus-incrementality question this guide covers in detail below.

Most Amazon PPC data accuracy complaints are really a mismatch on one of the first four. A seller who says “my reporting is wrong” usually means the numbers haven’t settled yet, or they’re comparing two different scopes, not that Amazon recorded the wrong click.

Genuine causal accuracy questions, whether the spend is actually creating sales rather than just claiming credit for them, are rarer and need a different kind of test entirely, which is why that section gets its own worked example later in this guide.

ACOS vs. ROAS vs. TACoS

Three metrics, three different questions, and most reporting problems come from using the wrong one to answer the wrong question.

None of these three is a profitability number. A 40 percent ACOS product with strong margin and a real organic flywheel behind it can produce more profit than a 15 percent ACOS product that never ranks and never builds momentum on its own. If you only ever look at the ratio and never at the dollars underneath it, you’ll make decisions that look disciplined on the dashboard and cost you money in the bank account. That’s a deeper topic than reporting accuracy alone, but it’s the reason we tell clients to report profit dollars alongside ACOS, every week, not the ratio by itself.

TACoS deserves a specific caution here, because it’s the metric most affected by everything covered earlier in this guide. TACoS pulls from total account sales, which includes organic, so it dilutes the restatement lag and attribution window noise that make weekly ACOS swing around. That makes it a more stable trend line, but only if you’re reading it monthly. Checking TACoS week to week reintroduces the same settling problems you were trying to avoid by using it in the first place.

Attribution Is Not the Same as Incrementality

This is the single most expensive misunderstanding in Amazon PPC reporting, and almost nothing published on the topic gives it more than a sentence.

Attribution answers “which ad gets credit for this sale.” Incrementality answers a completely different question: “would this sale have happened without the ad.” Those are not the same thing, and the gap between them can be enormous.

Take a hundred shoppers who were already planning to buy your product this month, because they’d used it before, saw a friend’s recommendation, or found you through organic search. Fifty of them also happen to see your branded Sponsored Products ad along the way. Ten click it. Eight of those ten buy. Amazon will attribute all eight of those purchases to the ad. But most of those eight were buying anyway. The ad probably influenced one or two of them, at most, to buy sooner or choose your listing over a competitor’s in that specific moment. The other six to seven attributed sales would have closed regardless.

That’s not a criticism of Amazon’s attribution model. It’s doing exactly what it’s designed to do, crediting the last qualifying interaction. The mistake is treating attributed sales as proof the ad caused the revenue. It’s proof the ad was present, which is a different claim. For most day-to-day bid and budget decisions, attribution is a fine enough proxy, and running a full incrementality test on every keyword isn’t realistic. But if you’re deciding whether to keep funding a branded defense campaign that’s already capturing demand you’d get organically anyway, that’s exactly the decision where the attribution number alone will mislead you, and it’s worth a controlled holdout test before you commit real budget to the answer.

Amazon Marketing Cloud and Multi-Touch Attribution

For accounts with enough spend and complexity to justify it, two tools go beyond what the standard console reports.

Amazon Marketing Cloud is a privacy-safe clean room where advertisers run SQL queries against pseudonymized, event-level data across Sponsored Ads, DSP, and Amazon Attribution. It used to require a DSP relationship to access. Amazon opened it to sponsored-ads-only advertisers in 2024, then expanded eligibility further in 2025, so a much larger set of sellers can now use it, mainly for path-to-purchase analysis and audience overlap questions that a standard campaign report simply can’t answer.

Multi-touch attribution is newer and narrower in scope. Rather than crediting the single last touchpoint, Amazon’s published MTA research describes a model that distributes credit across multiple Amazon Ads touchpoints in a customer’s path, using a combination of randomized controlled trials and machine learning to weight each touchpoint’s actual contribution. It’s still early. If you’re running a handful of campaigns on a modest budget, standard last-touch reporting and a solid understanding of its limits will serve you better than chasing a beta feature built for portfolios far larger than yours.

Unified Reporting in 2026

This one has a hard deadline attached to it, and it’s easy to miss if you haven’t logged into Measurement and Reporting recently.

Unified Reporting reached general availability on June 8, 2026, consolidating Sponsored Products, Sponsored Brands, Sponsored Display, and DSP data into a single report format, across accounts, countries, and up to six years of historical data depending on the reporting grain. If your team, or an outside tool, still points at the old Sponsored Ads reports page, this is the year to migrate, not the week before the deadline. Amazon’s own support documentation walks through the console path, and an automated migration tool for scheduled reports is expected but hasn’t shipped with a firm date as of this writing.

How to Audit Amazon PPC Reporting Accuracy

Before you act on any number that looks off, run it through this Amazon PPC reporting accuracy checklist in order.

1. Confirm you’re using one source of truth. Compare Amazon’s own Ads reporting against Amazon’s own Ads reporting, not against Seller Central’s Business Reports or a third-party dashboard, until you’ve isolated whether the discrepancy is even real.
2. Freeze the date range on both sides. Comparing Amazon’s January 1 to 31 against an agency dashboard running January 1 to 30 will manufacture a discrepancy that isn’t there.
3. Check the marketplace and timezone. US, UK, and EU marketplace data can get blended by accident in a multi-market account, and timezone mismatches shift late-night orders into the wrong day.
4. Confirm the attribution methodology on both sides. Know the window, the model, and the report definition for each number before you compare them. A 7-day SP number and a 14-day SB number are never going to reconcile cleanly against each other.
5. Compare raw traffic before you compare ROAS. Look at impressions, then clicks, then spend, in that order, before you get to orders and sales. If clicks already don’t match, there’s no point investigating the sales gap yet.
6. Give the attribution window time to close before final judgment. Anything inside the last 28 days is still settling. Treat it as directional, not final.

Amazon PPC Reports for Conversion Tracking

You don’t need every report Amazon offers. You need the handful that actually drive decisions.

• Search Term Report shows the exact words shoppers typed before clicking your ad, and it’s the foundation for both keyword harvesting and negative keyword cleanup.

• Targeting Report shows how your chosen keywords, ASINs, or categories performed, which is a different question than what the search term report answers, since a single broad keyword can match dozens of different search terms.

• Advertised Product Report shows performance for the specific ASIN featured in the ad, so you can tell which listings your spend is actually moving versus which ones just have an active campaign attached to them.

• Placement Report breaks performance down by where the ad actually served, top of search, product pages, or rest of search- and most accounts leave placement bid modifiers untouched despite this being one of the more underused, high-impact settings available.

• Purchased Product Report, covered above, is the one place you can see the halo effect and cross-sell patterns your campaign-level numbers hide entirely.

Search Term Report vs Search Query Performance

The Search Term Report is not the same thing as Search Query Performance data, and the two get confused constantly. The Search Term Report lives inside Sponsored Ads and shows only the queries that triggered a paid click on your campaigns. Search Query Performance, available to brand-registered sellers through Brand Analytics, shows the entire query across the category, paid and organic combined, along with your share of impressions, clicks, and purchases against every other seller competing for that same search. The Search Term Report answers “how did my ad do on this term.” Search Query Performance answers “how much of this term’s total demand am I actually capturing.” Pulling bid decisions from SQP data, which includes organic traffic your ads never touched, is a quiet way to overbid on terms where your organic listing was already doing the work.

Reconciling Amazon PPC Attribution Windows and Halo Sales

Reconciling Amazon PPC attribution windows, halo sales, and restated data across every campaign in an account is most of what a good PPC manager actually does week to week. If that’s more than your team has time to run properly, Selouse manages Amazon PPC reporting and reconciliation for 600+ brands, and our Amazon PPC management service handles exactly this layer of the work.

FAQ

What is Amazon PPC attribution? 

It’s the rule set Amazon uses to decide which ad gets credit for a sale, an add to cart, or a detail page view. For Sponsored Products, Sponsored Brands, and Sponsored Display, it works on a last-touch basis within an attribution window that ranges from 7 to 14 days depending on the ad type and account type.

Why did my ACOS change after I already pulled the report? 

Amazon restates conversion data at 1, 7, and 28 days after each sale, and logs sales against the click date rather than the purchase date. A number you pulled last week is still catching up to purchases that happened inside its attribution window.

Is a mismatch between Amazon Attribution and my own analytics a problem? 

Not automatically. Amazon’s own documentation states a 10 to 20 percent gap is normal, caused by different click-counting methods between platforms. A gap well past that range, or one that appears suddenly, is worth investigating.

Will my old Sponsored Ads reports stop working? 

Yes, on a fixed date. The legacy Sponsored Ads reports and Amazon DSP reports pages retire December 31, 2026, and any scheduled reports built there need to move to Unified Reporting before that deadline.

What’s the difference between an attributed sale and an incremental sale? 

An attributed sale is one Amazon’s model assigned to your ad, whether or not the shopper would have bought anyway. An incremental sale is one that happened specifically because of the ad. Most attributed sales are not fully incremental, especially on branded terms.

Does the January 2026 attribution change affect my Sponsored Products ACOS? 

No. That change applies to view-based credit on Sponsored Brands, Sponsored Display vCPM, and DSP in-store placements. Click attribution, which is how Sponsored Products has always been credited, is unchanged.

What does Amazon PPC reporting accuracy actually mean? 

It covers five separate questions: whether the click and spend data was recorded correctly, whether credit was assigned by the right rules, whether the sale landed in the right reporting period, whether you’re comparing the right scope, and whether the ad actually caused the sale. A report can be accurate on the first four and still mislead you on the fifth.

How does Amazon PPC conversion tracking actually work? 

Amazon tracks a click, then watches for a qualifying action, a detail page view, an add to cart, or a purchase, inside that ad type’s attribution window. If the action happens inside the window and no other ad interaction qualifies for credit first, Amazon logs it against the original click.

Is Amazon Ads attribution the same as Amazon PPC attribution? 

They’re often used interchangeably, and inside the Sponsored Ads console they mean the same thing: the last-touch credit model applied to Sponsored Products, Sponsored Brands, and Sponsored Display. Amazon Attribution, capitalized as a product name, is a separate tool for off-Amazon traffic, covered earlier in this guide.

How do I check Amazon PPC data accuracy before trusting a report? 

Run it through the audit checklist above: one source of truth, a frozen date range, a confirmed marketplace and timezone, a known attribution methodology on both sides, raw traffic compared before ROAS, and enough time for the attribution window to close.