CPA
What it means
CPA is your Amazon Ad Spend divided by the orders Amazon credited to those ads: spend €2,000, get 200 credited orders, and each one cost €10 to win. It is a euro cost per order, and it counts only the orders credited inside the attribution window — not your total Amazon orders.
It prices the same Ad Spend that ACOSACOSabove that, those sales lose money.Under 20% is healthy does, measured against orders rather than the revenue those orders brought in. Meta reports a figure under the same label from its own credited purchases; the two price different orders on different platforms and cannot be averaged into one acquisition cost.
Show the math
Formula and a worked example
Ad Spend is everything Amazon charged you for advertising in the period, across every campaign type. Purchases, the denominator, is the count of orders Amazon credited to those ads inside its attribution window — not the orders your Amazon account took.
Worked example. You spend €3,000 and Amazon credits 250 orders, so CPA = 3,000 ÷ 250 = €12. If a typical order leaves you €15 after product cost, fees and shipping, you keep €3 on each one. Let CPA drift to €16 and every ad-driven order costs you a euro, while revenue keeps growing.
Up is bad here. A rising CPA means each order is getting more expensive to buy, whatever else is going up alongside it.
It answers the question
What does it cost to buy one more Amazon order? If that is more than the order leaves behind after costs, growth is losing you money.
Why it matters
CPA is quoted in the same currency as the margin an order leaves you, so you can act on it without a spreadsheet. A ratio tells you something changed; this tells you whether the next order is worth buying.
The catch is basket size. Two SKUs both sitting at €12 CPA look identical until one sells a €20 accessory and the other a €90 bundle — same cost to acquire, opposite verdicts, which is why CPA and ACOS can point different ways in the same month.
What good looks like
There is no universal standard, because your ceiling is set by your own economics rather than an industry figure. Work out what a typical order leaves after product cost, Amazon fees and shipping, then treat a comfortable fraction of that as your limit.
Beyond that, watch your own trend per SKU. A CPA rising month after month on the same products means the auction is getting dearer or the listing is converting worse, and the SKU table is where you find out which.
How to improve it
| Lever | What you do | Expect | How long | Watch out for |
|---|---|---|---|---|
| Fast Cut the terms that spend and never convert | Add negatives for search terms with 30+ clicks and no orders over 30 days | CPA down within a fortnight | 1–2 weeks | You lose discovery. Some terms convert slowly on a long consideration cycle, and a negative you added in March is easy to forget in September. |
| Fast Stop paying top-of-search prices for mid-page value | Trim placement bid modifiers where the premium slot converts no better than the rest | CPC falls and CPA follows | 1–2 weeks | You slide down the page, Impressions drop, and the competitor who takes the slot usually charges you more to win it back. |
| Slow Make the listing convert harder | Main image, price, review count and delivery promise on your top ad-driven SKUs | The same Clicks produce more orders, so CPA falls | 4–8 weeks | Review and photography work costs real money that never appears in Ad Spend, and cutting price lowers CPA by shrinking the very margin CPA is meant to protect. |
| Slow Move budget to SKUs that already convert | Shift spend towards products with a proven order rate and a Buy Box you reliably hold | CPA across the account falls without changing a single bid | 1 quarter | Your catalogue narrows. Newer SKUs never get the velocity they need to earn rank, so next year's winners never start. |
Every lever costs something somewhere. The last column is the one to read twice.
Read it with
CPA prices the order. The number beside it says whether that price was worth paying.
Cheaper and bigger
More orders for less each. Either the auction eased or the listing improved — find out which, because only one of them is yours to keep.
Buying the volume
Orders are up because you are paying more for each one. Deliberate during a launch, quietly costly as a standing setting.
Trimmed back
You cut the expensive orders and kept the cheap ones. Efficiency improved by shrinking, which is fine until velocity drags rank down with it.
Worst of both
Fewer orders, each one dearer. This is almost never a bidding problem — look for a stockout, a lost Buy Box or a competitor undercutting you.
One is the cost of buying an order, the other is what the order leaves behind. Side by side, CPA stops being a number to minimise and becomes a limit: €12 is excellent at 40% Contribution Margin on a €90 basket and ruinous at 15% on a €20 one.
Together they tell you where a rising cost per order started. CPA up with CPC up means the auction got more expensive and it is a bidding job. CPA up with CPC flat means you are paying the same for traffic that has stopped buying — the listing, the price or the stock.
Common misreads
The denominator moved. Credited orders arrive late while the attribution window fills in, so the last two weeks of any period always read expensive. Let the period settle before you judge it.
A €4 CPA on a €6 item is a loss; €30 on a €200 bundle is a bargain. The number is meaningless without the basket behind it, which is why two SKUs at the same CPA can deserve opposite decisions.
They agree only while basket size holds still. Sell more of a cheaper SKU and ACOS worsens while CPA improves — same account, same spend, two different stories.
Also called
Cost per acquisition · cost per order · CPO · cost per action
See yoursYour cost per ad-driven Amazon order for the last 30 days, next to Ad Spend.
Open Ads Analytics →CPA
What it means
On the Meta pages, CPA is your Meta ad spend divided by the purchases Meta credited to those ads — a euro cost per order. The app reads it from Meta's own cost-per-purchase figure, so the orders underneath are the website purchases Meta claims, not every order your store took that day.
That is a different denominator from the Amazon figure of the same name, drawn from a different platform's records. Read them side by side if you like; never add or average them.
Show the math
Formula and a worked example
Ad Spend is your Meta SpendSpendspend that does not move total sales is the first thing to cut. for the period. Purchases, the denominator, is the count of website purchases Meta credited to your ads — the same claim that sits behind Purchase ValuePurchase ValueTotal purchase value attributed to this format. Pair it with spend to judge which formats return the most revenue and ROAS on these pages.
Worked example. You spend €4,000 in a month and Meta credits 160 purchases, so CPA = 4,000 ÷ 160 = €25. Split by age, the 25–34 rows spent €1,200 for 80 purchases — €15 each — while the 55+ rows spent the same €1,200 for 25 purchases, or €48 each.
Neither slice pays the €25 on the card. That figure is the blend of every slice, and it is common for no single audience to sit anywhere near it — which is the point of this number on Meta: it is only useful once you cut it by something.
It answers the question
What does one order cost through Meta, and which slice of the spend is paying too much for it? The app carries CPA on every Meta breakdown — age, gender, country, placement, campaign, ad, landing page and creative format — so the answer is a sort, not a guess.
Why it matters
CPA is the one Meta figure quoted in the same units as your margin, which makes it the number that decides where the budget moves next. ROAS ranks your ads; CPA tells you whether the winner is affordable at all.
It is also the fastest read on audience fatigue. A campaign whose creative has stopped landing keeps spending at roughly the same rate while credited purchases thin out, and the cost per order climbs long before anyone calls the ad tired.
What good looks like
There is no published band, and a euro figure could not carry one across categories or markets anyway. Your ceiling is what an order leaves after product cost, shipping and fees — set that number first, take a comfortable fraction of it, and treat the rest as spending you have chosen rather than spending that works.
After that it is your own trend, read per slice. An account CPA that holds while the gap between your best and worst audience widens is a budget problem hiding inside a stable average.
How to improve it
| Lever | What you do | Expect | How long | Watch out for |
|---|---|---|---|---|
| Fast Move budget off the expensive slices | Sort the age, country and placement tables by CPA and pull spend out of the rows sitting above your ceiling | Account CPA falls without touching a single creative | 1–2 weeks | Those rows carried Reach as well as cost. Narrowing to the cheap ones pushes Frequency up on the audience you keep, and fatigue lands sooner. |
| Fast Find out whether it is the click or the page | Compare CPC (Link) against CPA — cheap clicks with an expensive CPA point past the ad, at the landing page | A specific thing to fix instead of a bid to change | Same week | It only tells you where the problem is. The fix that follows is site work, and it lands weeks after the diagnosis. |
| Slow Refresh the creative before frequency forces you to | Ship new hooks and formats on the campaigns whose Frequency is climbing | Cost per order returns towards where it was | 3–6 weeks | Production costs real money that never shows up in Spend, and you trade a known CPA for an unknown one every time you swap. |
| Slow Raise what an order is worth | Bundle, add a second product to the page, or lift price on the SKUs Meta sells most of | The same CPA becomes affordable because the order behind it is bigger | 1 quarter | You changed the product, not the advertising. Compare like with like before crediting the ads, and a higher price usually costs you some of the purchases the CPA was counting. |
Every lever costs something somewhere. The last column is the one to read twice.
Read it with
CPA prices the order. What it does not tell you is where the cost came from, or whether the audience has any room left in it.
CPC (Link) is Spend divided by Link clicksLink clicksCounts only people who reached your site, not likes or post expands. If these climb while orders don't, the ad is working and the page isn't, so it prices the visit rather than the order. Both rising together means Meta is charging more for traffic. CPA rising while CPC (Link) holds means the traffic costs the same and has stopped buying — that is your site or your offer, and no bid change will fix it.
Frequency is the average number of times one person saw your ad, and the app treats anything past 5 as fatigue. CPA drifting up while Frequency climbs is saturation: you are paying more because you keep buying the same people. New audience or new creative, not a new bid.
Common misreads
They price different orders. Meta counts website purchases it credits to its own ads; Amazon counts orders credited inside its attribution window on Amazon. One blended figure would be an average of two things that never met.
CPA prices the order, not the margin on it. A €12 CPA against a €40 basket at thin margin still loses, and the purchases in the denominator are the ones Meta claims — check them against what your store actually took before you scale on the strength of it.
The account figure is an average across slices that can be three times apart. Sort by placement and audience first: it is common for one expensive row to move the whole number while everything else is where you left it.
Also called
Cost per purchase · cost per acquisition · cost per order
See yoursYour Meta cost per credited purchase for the period, next to Spend, ROAS and Frequency.
Open Meta Analytics →