Every marketing team has that meeting. The one where paid search takes credit for a conversion, email takes credit for the same conversion, organic search takes credit for it too, and by the time everyone’s finished presenting their dashboards, the business has apparently acquired one customer four times. Nobody’s lying. Everyone’s reporting correctly. The problem is they’re all using different rules to decide who deserves the credit, and nobody in the room agreed on those rules in advance.
That is marketing attribution in its most natural habitat — not a software problem, not a tracking problem, a definitions problem, and the businesses that figure it out stop making budget decisions based on whichever channel happened to shout loudest.
What Is Marketing Attribution?
Marketing attribution is the practice of assigning credit for a conversion, a purchase, a lead, a sign-up, to the marketing touchpoints that contributed to it. Someone saw an ad on Monday, clicked an email on Wednesday, found the site through organic search on Friday, and converted on Saturday. Attribution is the set of rules that decides which of those four moments gets the credit, how much of it, and why.
Done well, it tells you which channels are genuinely earning their spend. Done badly, it tells you a story that sounds like insight and is actually just whichever model happened to favor the channel the person building the dashboard works on.
The Main Attribution Models and What They Actually Mean
Every attribution model is a different answer to the same question: who gets credit for the sale? None of them are wrong. All of them are incomplete. Here’s what each one is actually measuring:
- Last-click attribution gives 100% of the credit to the final touchpoint before conversion. Clean, simple, and systematically undervalues every channel that did the work of warming the lead up. The channel that closes always looks like a genius. The channel that started the relationship gets nothing.
- First-click attribution gives 100% of the credit to the first touchpoint. Same logic, opposite bias. Great for understanding what’s starting journeys, useless for understanding what’s finishing them.
- Linear attribution splits credit equally across every touchpoint in the journey. Fair in theory. Assumes every interaction contributed equally, which is almost never true.
- Time-decay attribution gives more credit to touchpoints closer to the conversion. Closer in time means more responsible, roughly. Logical for short sales cycles, misleading for longer ones where the touchpoint that started everything genuinely mattered.
- Position-based attribution (also called U-shaped) splits the majority of credit between the first and last touchpoints, distributing the rest among everything in between. A reasonable compromise, still a compromise.
- Data-driven attribution uses machine learning to distribute credit based on which touchpoints actually correlated with conversions across the full dataset, rather than a fixed rule. This is GA4’s default model now, and it produces more defensible numbers than any rule-based model, as long as you have enough conversion volume for the model to learn from, which smaller accounts frequently don’t.
Why Attribution Models Disagree With Each Other
Take one customer journey and run it through each model above, and you’ll get six different answers about which channel deserves credit. This isn’t a bug. It’s the entire point — each model is asking a slightly different question about causality, and real customer journeys don’t have one clean cause. A prospect who saw a display ad, read a blog post, clicked a retargeting ad, and then converted through a branded search was influenced by all four. The model decides whose story gets told.
Where this becomes a real operational problem is when different teams are using different models to report their own channel’s contribution, which is exactly what happens by default in most marketing stacks. Paid search reports in last-click inside Google Ads. Email reports opens and clicks without attribution context. GA4 applies data-driven across everything. Nobody agreed that these should reconcile with each other, and so they don’t, and the resulting revenue tracking conversation is the one described in the opening paragraph of this article.
How Attribution Connects to Paid Ads Performance
Attribution isn’t an analytics problem that lives in a reporting tab — it’s a budget problem that directly changes which campaigns get scaled and which get cut. According to HubSpot’s State of Marketing research, proving ROI and attributing revenue to the right channels consistently ranks among marketers’ top challenges year over year, and the reason isn’t a lack of data, it’s that the same data produces different ROI numbers depending on which model you apply to it.
In practice, this plays out in two specific ways. Running Google Ads under last-click attribution makes brand keywords look brilliant, because they’re always the last click before conversion for anyone who was already going to convert regardless. And running Meta Ads with a 7-day click / 1-day view attribution window while GA4 uses a longer data-driven model means Meta can claim a conversion that GA4 didn’t assign to it, and both numbers are technically correct under their own rules. The Google Ads Help documentation and GA4’s attribution documentation both outline their respective models clearly — the problem isn’t that the rules are hidden, it’s that nobody reads both sets of rules and notices they’re different.
Multi-Touch Attribution vs Single-Touch
Single-touch models, first-click and last-click, are simple and fast but assign all credit to one moment in a journey that had many. Multi-touch models, linear, time-decay, position-based, data-driven, distribute credit across the full journey and produce a more complete picture at the cost of more complexity and more data required.
Which you use should be determined by your sales cycle length and your conversion volume. Short cycle, low volume: a simple model is probably fine, there isn’t enough data to make a sophisticated model meaningful anyway. Long cycle, high volume: a rule-based multi-touch model at minimum, data-driven if you have the conversion volume to feed it. The mistake is applying a simple model to a complex journey because it’s easier to explain, and then making budget decisions as if the simplified picture were the whole truth.
What the Blogs Conveniently Leave Out
The fine print that determines whether attribution actually improves decisions or just produces more confident-looking errors:
- Attribution models don’t fix broken tracking — they amplify it. A sophisticated data-driven model applied to incomplete data produces sophisticated-looking wrong answers. Before arguing about which attribution model to use, confirm that the underlying event data is actually complete, deduplicated, and consistent across platforms. This is exactly the problem covered in Why Your Conversion Tracking Is Broken — attribution is the layer above tracking, and it inherits every flaw from the layer below it.
- View-through attribution is where budgets go to get inflated. Most ad platforms, Meta included, offer view-through attribution — crediting a conversion to an ad the user saw but never clicked. A 1-day view-through window on Meta means any conversion that happens within 24 hours of someone seeing your ad gets credited to that campaign, even if the person never engaged with it at all. This can make awareness campaigns look like conversion machines. Know what window your platforms are reporting under before reading the numbers.
- Data-driven attribution needs a minimum conversion volume to function properly. GA4’s data-driven model requires roughly 400 conversions and 4,000 ad clicks in a 30-day period to work as intended. Below that threshold, the model has too little signal to distinguish meaningful patterns from noise, and it falls back to last-click or produces unreliable outputs without necessarily telling you it has done so.
- Cross-device journeys are mostly invisible to every model. A user who sees an ad on their phone, researches on a tablet, and converts on a desktop is treated by most attribution systems as three separate anonymous users unless they’re logged into a platform that can stitch the sessions together. The resulting attribution gap is real, structural, and almost never acknowledged in attribution model comparisons.
- Attribution windows are settings, not facts. The lookback window, how far back a platform looks for the touchpoint to credit, is adjustable in most platforms and almost universally left at its default. Different defaults across platforms mean a single conversion can be claimed by multiple channels simultaneously, all operating within their own window settings, none of which were chosen to reconcile with each other.
Where This Fits
Marketing attribution isn’t the thing you configure once inside GA4 and then trust. It’s an ongoing decision about which version of the truth you’re using to make budget decisions, and the version that looks the cleanest is almost never the most accurate. The businesses that get this right don’t have better data than everyone else — they have clearer rules about which model they’re using, why, and what its known blind spots are.
This is the foundation underneath the marketing attribution systems OJC Labs builds — proper tracking first, then attribution that actually reflects how customers move through a real journey, not just the last click before the sale.
If your channel dashboards are telling four different stories about the same customers, see how we’ve untangled this in practice.