What Is Attribution?
Who gets the credit for a sale, and why the answer is usually wrong.
In 2019 I ran two stores from a basement in Stockholm and spent about $50k a month on ads. One Monday I opened three tabs. Meta said we had made $41k from ads that week. Google said $22k. Shopify said we had made $38k in total. From everything. Including my mom.
I did what everyone does: I picked the number I liked, scaled the campaign it pointed at, and lost money for six weeks before I understood what had happened. This page is the explanation I wish someone had given me that Monday.
The one-sentence version
Attribution is the rule you use to decide which ad gets credit for a sale. Every tool has a rule. Most rules are chosen to make the tool look good. The only rule that cannot flatter anyone is the one that starts from your actual orders.
The setup
One customer. One order. Four touches.
Meet Sara. On Tuesday she scrolls past your Instagram reel and keeps scrolling. Wednesday a Meta retargeting ad follows her, she clicks, browses, leaves. Friday she types your brand name into Google and clicks the ad sitting on top of your own organic result. Saturday your 10% email lands and she buys. $120.
One order. Now ask each of your tools who deserves the credit, and watch the answers change.
Try it: one $120 order, four touches
Who gets the credit?
Tue
Instagram ad
$0
Scrolled past a reel. Did not click.
Wed
Meta retargeting click
$120
Clicked, browsed, left.
Fri
Google brand search
$120
Typed the brand name, clicked the ad.
Sat
Email discount
$120
Opened the 10% code, bought. $120.
Credited: $360 for a $120 order
Each platform counts the whole order. Meta claims $120 (the reel and the click are one claim), Google claims $120, Klaviyo claims $120. That is $360 of credit for one $120 order.
Four tools. One order. $360 of credit. That is the whole problem.
Nobody is lying. Meta really did show her an ad within its window. Google really did get the click. The email really was opened before the purchase. Each tool answers honestly inside its own rules, and each set of rules was written by the company that gets paid when the answer is "us".
It is a bit like asking three of your friends who found the restaurant. Everyone remembers suggesting it.
Three words that fool everyone
Click. Window. Model.
Click vs view
Meta counts a sale if Sara only saw the ad within a day of buying. No click needed. That is a view-through conversion, and on retargeting campaigns it can be most of the reported revenue.
The window
7-day click on Meta, 30-day click on Google, 5-day open on email. A single customer can sit inside all three windows at once. That is how one order becomes three.
The model
Last click, first click, linear, data-driven. Fancy names for who wins the tie. Pick a different model and the same order changes owner. None of them can add up to more than the order, unless you run three of them in three tools and sum the reports.
Why it matters
Wrong credit moves real money
Here is how my six weeks went. Brand search had a ROAS of 14, so I gave it more budget. Retargeting had a ROAS of 9, more budget. Prospecting on Meta had a ROAS of 1.8, so I cut it. Three sensible decisions, every one of them backwards.
Brand search and retargeting were harvesting customers that prospecting had found. When I cut the prospecting, there was less to harvest. Brand searches dried up a few weeks later, and the "14 ROAS" campaign quietly became a 6, then a 4, with nobody new coming in.
The data never said prospecting was working. It just credited them to the wrong ad.
The fix
Start from the order, not from the ad platform
Flip the question. Instead of asking each platform "what did you sell?", take each Shopify order and ask "what touched this?" Now the order is the unit. It can only be counted once. Credit can be split, but it cannot exceed $120, because there was only ever $120.
Record every touch server-side
So iOS, Safari and ad blockers do not delete Sara's Wednesday click.
Match touches to the real order
No order, no conversion. Modeled purchases do not exist here.
Ask Sara
A one-question survey catches the Tuesday reel nobody clicked.
Send the truth back
Verified orders go to Meta and Google, so their algorithms learn from real buyers.
This is what TrueROAS does, and I am obviously biased. The principle holds whichever tool you use: if the channel numbers do not add up to your Shopify revenue, you are not looking at attribution. You are looking at four sales pitches.
30-second self-check
Is your attribution lying to you?
Add up last month's revenue as reported by Meta, Google and your email tool. Is it more than your Shopify revenue?
Does your brand-search campaign have a ROAS above 8?
Do Meta's purchases and Shopify's orders differ by more than 20% on a normal week?
Who does not need to care
Honest footnote
If you run one paid channel, no email program, and mostly first-time buyers, last click is roughly right and you can stop reading. The moment you add a second channel or a retargeting campaign, the credit starts overlapping, and this page starts costing you money.
Questions people actually ask
Straight answers
Is attribution the same as tracking?
No. Tracking records what happened (a click, a page view, an order). Attribution decides who gets credit for the order. You can have perfect tracking and terrible attribution: all the touches recorded, and 100% of the credit handed to the last one.
Why do Meta and Google each claim the whole order?
Because each one only sees its own touch and its own rules say a touch within the window counts. Neither knows the other exists. The double counting is not a bug in either tool, it is what happens when you add up two reports that were never designed to be added.
Which attribution model should I use?
For budget decisions, one that starts from your real orders and cannot exceed your real revenue. For diagnosing a single platform's creative, that platform's own reporting is fine. The mistake is using the second for the first.
Does a post-purchase survey fix this?
It fixes part of it. Surveys catch the Instagram reel nobody clicked. They are fuzzy about which campaign and which ad. That is why TrueROAS uses them as one of three signals, not the only one.
See your own numbers, verified against your orders.
Free under $10k/month in revenue. 14-day trial above that. Nothing changes in your ad accounts. You will know within a week which platform has been telling stories.
"I built this because I was making $50k-a-month decisions on numbers I knew were wrong. If TrueROAS does not show you something Meta or Google got wrong in the first two weeks, tell me and I will help you cancel."
Rasmus Arvidsson
Founder, TrueROAS. Answers support himself.
Keep going