Everyone wants to know which marketing is working. Attribution is supposed to provide the answer. In reality, it often creates more questions.
A customer rarely sees one advert, clicks once and buys. They might discover a brand through Instagram, read an editorial article, search on Google, visit the website, leave, receive an email, ask ChatGPT for a recommendation, return through a paid brand search and finally purchase using a cash back partner.
So who gets the credit? Meta? The publisher? Google? CRM? AI discovery? Affiliate?
The answer depends almost entirely on how attribution is being measured. And that matters because attribution doesn’t just report performance. It influences where businesses put their money next.
What is attribution?
Attribution is the process of assigning credit for a conversion to the marketing interactions that contributed to it. The problem is that there are multiple ways of doing this and they can produce very different answers from exactly the same customer journey.
Last-click attribution
100% of the conversion is attributed to the final measurable interaction before purchase. Simple and easy to understand but it can dramatically overvalue channels that are at the bottom of the funnel while undervaluing those that created or influenced it.
First-click attribution
100% of the value goes to the interaction that originally introduced the customer. Useful for understanding discovery but it ignores everything that subsequently moved the customer towards purchase.
Linear attribution
Credit is distributed equally across measurable interactions. More balanced, but it assumes every interaction contributed equally, which is rarely true.
Position-based attribution
More credit is given to selected stages, typically the first and final interactions. It recognises introduction and conversion but still relies on an arbitrary decision about how much each interaction deserves.
Data-driven attribution
Algorithms use available conversion data to estimate the contribution of different interactions. Potentially more sophisticated, but only as useful as the data available to the model.
IMPORTANTLY – Missing data doesn’t magically become accurate because an algorithm is analysing it.
Then every platform has its own version of the truth
This is where attribution becomes particularly difficult. Google has an attribution model. Meta has one. Your analytics platform has another. Your affiliate network has another. Your CRM may have another. Your ecommerce or booking platform records transactions independently. Each platform sees a different part of the customer journey.
A customer might generate one £100 transaction.
Meta reports £100.
Google reports £100.
Affiliate reports £100.
Your CRM attributes £100.
Suddenly your marketing reports appear to have generated £400.
The bank account still only took £100.
This is why adding together platform-reported revenue is one of the fastest ways to create a completely distorted view of marketing performance.
Attribution windows change the answer again
Even within the same platform, changing the attribution window can materially change reported performance. Should an advert receive credit if somebody buys:
Immediately? Within 24 hours? Seven days later? Thirty days later?
What if they viewed an advert but never clicked? What if they clicked several different adverts? What if another channel subsequently converted them?
The longer the attribution window, the more conversions a platform may potentially associate with previous interactions. This can make performance appear stronger without the business generating a single additional transaction.
Click-through versus view-through
This distinction is particularly important in paid social and display.
Click-through attribution
The customer clicked the advertising and subsequently converted. There is a visible interaction.
View-through attribution
The customer was served an advert, didn’t click it, but later converted through another route. The advertising platform may still claim some credit. View-through measurement can capture genuine influence, it can also create a question:
Did the advertising cause the conversion or did it happen to be shown to someone who was going to purchase anyway?
That’s not always easy to answer and it is exactly why platform reported ROAS should not automatically be treated as incremental ROAS.
Tracking isn’t attribution
This is another distinction we think is important.
Tracking asks: “What happened?”
Attribution asks: “What deserves the credit?”
Tracking might tell us that a customer: Clicked an advert. Visited the website. Viewed a product. Joined the mailing list. Returned. Purchased. Attribution attempts to determine which of those interactions contributed to the commercial outcome. If the underlying tracking is broken, incomplete or inconsistent, attribution becomes unreliable before the modelling has even started.
That’s why BlackBx Radar examines data integrity and attribution separately.
First:
Can we trust the data?
Then:
What does the data actually tell us?
The customer journey is becoming harder to see
Modern attribution has another problem. Increasing parts of the journey are invisible – Cookie restrictions, Consent choices, Cross-device behaviour, Privacy changes, Walled gardens, Offline interactions, Word of mouth, Dark social, AI discovery, Customers moving between apps and browsers.
A consumer may discover a product through ChatGPT, send it to their partner on WhatsApp and purchase three days later by typing the brand directly into Google.
Analytics will record:
Organic Search = Conversion
The real journey was considerably more complicated. This means attribution increasingly operates with an incomplete picture.
Absence of measurable evidence doesn’t mean absence of influence.
Attribution versus incrementality
This is perhaps the most important distinction.
Attribution asks:
“Which interaction should receive credit?”
Incrementality asks:
“Would the outcome have happened without the activity?”
They aren’t the same question. A channel could receive thousands of attributed conversions while generating relatively little incremental demand. Another channel could introduce thousands of new customers but receive relatively few last-click conversions.
That is why BlackBx doesn’t look at attribution in isolation. We want to understand causality and commercial contribution, not simply credit allocation.
BLACKBX
WE SEE WHAT OTHERS CAN’T.