An attribution model is the rule that decides which marketing touchpoints get credit when a conversion happens. It’s how you answer the only question that matters at budget time: of all the ads, emails, searches, and clicks that touched a buyer, which ones actually moved the needle — and how much? Pick the wrong model and you’ll defund the channels that fed your pipeline while pouring money into the ones that just closed the deal.
Attribution Model
An attribution model is a rule or algorithm that assigns credit for a conversion to one or more touchpoints in the customer journey, used to evaluate channel performance and guide budget decisions.
How attribution actually works
A customer rarely converts on first contact. They see a paid social ad, forget it, run an organic search a week later, click a retargeting ad, read an email, and finally convert. That ordered sequence of interactions is the conversion path, and every interaction on it is a touchpoint.
An attribution model takes that path and splits a single conversion (or its revenue) into fractions of credit across the touchpoints. The model is just the splitting rule. Last-click gives 100% to the final touch; linear splits it evenly; data-driven uses your own historical conversion data to estimate each touch’s real lift.
The model you choose doesn’t change reality — it changes the story your reports tell about reality. That’s why two analysts looking at the same data can rank the same channels in opposite order. We see this constantly: a conversion funnel that looks broken under last-click looks healthy the moment you switch to a multi-touch view.
Attribution isn’t a truth machine. It’s a lens. Every model encodes an assumption about how buying decisions get made — and that assumption is usually wrong at the edges. The job is picking the lens that’s least wrong for your sales cycle.
The model types, compared
There are two families. Single-touch models hand all the credit to one interaction. Multi-touch models spread it. Here’s how the common ones distribute a single conversion across a four-touchpoint path:
| Model | Credit rule | Best for | Blind spot |
|---|---|---|---|
| Last-click | 100% to final touch | Short cycles, direct-response | Erases everything upper-funnel |
| First-click | 100% to first touch | Awareness, demand-gen measurement | Ignores what closed the deal |
| Linear | Equal split across all touches | Long, education-heavy cycles | Treats a trivial touch like a decisive one |
| Time-decay | More credit to recent touches | Promotions, fast B2C | Still under-credits early discovery |
| Position-based (U-shaped) | 40% first / 40% last / 20% middle | Balancing acquisition + conversion | The 20% middle is an arbitrary guess |
| Data-driven (DDA) | Algorithmic, from your conversion data | High-volume accounts with clean data | Needs scale; it’s a black box |
Last-click is still the default in most ad platforms, and it’s the most dangerous one. It’s why paid social and top-of-funnel content perennially look “unprofitable” — they rarely get the closing click, so last-click never sees them. Last-click attribution made sense when journeys were short and tracking was simple. Today it systematically defunds your demand creation.
Multi-touch attribution fixes the credit-hoarding problem by distributing it. Linear, time-decay, and position-based are rule-based multi-touch models. Data-driven is the algorithmic one — Google Analytics 4 made DDA its default and retired the rule-based models in its standard reports. DDA only works with enough conversion volume to find patterns, so small accounts fall back to position-based or hand-built rules. See multi-touch attribution for the full breakdown.
Why the privacy era broke attribution
Most attribution writing pretends it’s still 2018. It isn’t. The plumbing that fed these models has been quietly ripped out, and any practitioner-grade answer has to account for it.
- Third-party cookie deprecation. Safari and Firefox have blocked third-party cookies for years; Chrome has spent the privacy-sandbox era throttling them. Cross-site, multi-session path tracking — the thing multi-touch needs — is no longer reliable.
- iOS App Tracking Transparency (ATT). Apple’s opt-in prompt gutted app and cross-app measurement. A large share of users simply vanish from deterministic tracking.
- Consent Mode. Under GDPR-era rules, you can’t fire measurement tags until a user consents. Google’s Consent Mode v2 models the conversions you can’t observe — meaning a chunk of your attribution is now statistically inferred, not counted.
The net effect: attribution has shifted from deterministic (we tracked this exact user across these exact touches) to probabilistic and modeled (we estimated this path from partial signals). DDA, conversion modeling, and consent-aware setups aren’t optional extras anymore — they’re load-bearing. Anyone selling you pixel-perfect cross-channel attribution in 2026 is selling dashboard theater.
This is also why AI Overviews and zero-click search matter here. When a user gets their answer inside an AI Overview without clicking, that influence never becomes a tracked touchpoint at all. Brand discovery increasingly happens in places attribution physically cannot instrument — which is an argument for triangulating with digital marketing analytics and incrementality testing rather than trusting any single model.
How to actually choose one
Stop hunting for the “correct” model. There isn’t one. Match the model to your reality:
- Map your real sales cycle. A 3-day impulse purchase and a 9-month B2B deal need different lenses. Short cycle → time-decay or last-click. Long, multi-touch cycle → position-based or DDA.
- Check your data volume. DDA needs scale. Below a few hundred conversions a month, its patterns are noise — use position-based instead.
- Decide what you’re trying to fund. If you’re justifying upper-funnel spend, never report it under last-click. Use first-click or a multi-touch view so demand creation gets visible credit.
- Pick one model as your source of truth, then sanity-check against incrementality. The honest gut-check is a geo holdout or a spend-pause test: turn a channel off and watch what actually happens to conversions. Attribution tells you correlation; an incrementality test tells you cause.
For SEO-heavy programs specifically, attribution almost always under-credits organic, because organic shows up early in the path and as untracked branded search. We account for that explicitly in our growth program so organic doesn’t get starved by a last-click report.
Frequently Asked Questions
What is an attribution model in simple terms?
An attribution model is the rule that decides which marketing touchpoints get credit for a conversion. If a customer saw an ad, searched, then clicked an email before buying, the model determines how much of that sale each step earns — which shapes how you judge channels and spend your budget.
What is the best attribution model?
There’s no universally best model. Last-click suits short, direct-response cycles; data-driven suits high-volume accounts with clean data; position-based works for long, multi-touch journeys. Match the model to your sales cycle and data volume, then validate it against incrementality tests rather than trusting it blindly.
What is the difference between single-touch and multi-touch attribution?
Single-touch models (last-click, first-click) award all credit to one interaction. Multi-touch models (linear, time-decay, position-based, data-driven) distribute credit across every touchpoint in the path. Multi-touch gives a fairer view of how channels work together but needs more data and cleaner tracking to be reliable.
Why has attribution become less accurate?
Third-party cookie deprecation, Apple’s iOS App Tracking Transparency, and consent requirements have removed much of the user-level tracking attribution relied on. Modern measurement is increasingly probabilistic and modeled rather than counted, so attribution is now a directional lens best paired with incrementality testing.
Does Google Analytics 4 still offer last-click attribution?
GA4 made data-driven attribution its default and removed most rule-based models from standard reports, though last-click remains available in some settings. Because GA4 relies on Consent Mode and conversion modeling, treat its attribution as a modeled estimate, not an exact account of every user’s path.