Marketing Attribution Models Explained: Which One Fits Your Agency
Every attribution model distorts reality in a predictable direction. Here is how last-click, first-click, linear, time-decay, position-based, and data-driven models compare in practice.
Updated 7 min read
Every attribution model is wrong. That is not a hot take; it is a structural fact. A visitor sees a display ad on Monday, clicks an email on Wednesday, searches your client's brand on Friday, and converts. Which touchpoint "caused" the conversion? The honest answer is: all of them, to some degree, and no single number can capture that.
The question is not which model is correct. It is which model is the least misleading for the decisions your client needs to make. Change the model, and the same data tells a completely different story about what is working.
Last-click attribution
Last-click gives 100 percent of the credit to the final touchpoint before conversion. If a visitor clicked a Google Ads result and then converted, Google Ads gets all the credit, even if the visitor first discovered the brand through a podcast sponsorship three weeks earlier.
Why agencies use it: It is the default in most tools, simple to explain, and ties directly to the action that closed the deal. When a client asks "what drove this sale?" a last-click answer is concrete.
Where it breaks down: It is blind to everything that built awareness and consideration. Top-of-funnel channels like organic social, display, and content marketing will always look worthless under last-click because they rarely close. If budget decisions are based on last-click, you will systematically defund the channels that fill the top of the pipeline.
First-click attribution
First-click is the mirror image. It gives 100 percent of the credit to the very first touchpoint. If a visitor first landed on the site through an Instagram ad and converted two months later via a branded search, Instagram gets all the credit.
Why agencies use it: It highlights the channels that create new audiences. For clients focused on growth and acquisition, first-click tells them where new visitors are coming from.
Where it breaks down: It completely ignores the nurture and closing sequence. A visitor who discovers the brand through organic search but only converts after three retargeting ads and a discount email? First-click credits organic search and suggests the retargeting spend was pointless. That is a dangerous conclusion.
Linear attribution
Linear splits credit equally across every touchpoint in the path. Five touches? Each gets 20 percent.
Why agencies use it: It acknowledges that the full journey matters, not just the first or last touch. It is easy to explain: "every channel that touched this visitor gets equal credit."
Where it breaks down: Equal is not the same as fair. A branded search (the visitor already knows the name) gets the same weight as the original display ad that introduced them to the brand. It also inflates the apparent value of mid-funnel touches that may have been incidental rather than influential.
Time-decay attribution
Time-decay assigns more credit to touchpoints closer to the conversion and less to those further back. A click three days before conversion gets more weight than a click three weeks before.
Why agencies use it: It reflects a reasonable intuition: recent interactions probably mattered more than distant ones. For businesses with long sales cycles (B2B SaaS, real estate, financial services), it keeps historical touches in the picture without letting a six-month-old blog visit dominate the attribution.
Where it breaks down: It is harder to explain in a report. Clients want to know "how much credit did this channel get?" and the answer involves exponential decay functions. It also systematically undervalues awareness channels, just less aggressively than last-click does.
Position-based (U-shaped) attribution
Position-based gives 40 percent of the credit to the first touchpoint, 40 percent to the last, and splits the remaining 20 percent across everything in the middle.
Why agencies use it: It is a compromise. Both the first touch (awareness) and the last touch (conversion) get significant credit, while mid-funnel interactions still contribute. For agencies managing both brand-building and performance campaigns, it tells a story that validates both.
Where it breaks down: The 40/40/20 split is arbitrary. There is no empirical reason the first and last touches should each get exactly 40 percent. For short paths with only two touchpoints, it collapses into a 50/50 split that tells you little.
Data-driven attribution
Data-driven attribution uses machine learning to analyze your actual conversion data and assign credit based on what statistically contributes to conversions. It is the default in GA4 and the model Google pushes hardest.
Why agencies use it: In theory, it is the best model because it is calibrated to your real data rather than following a fixed formula. It can detect that email is more influential than display for one client and the reverse for another.
Where it breaks down: It needs volume. Google typically recommends at least 300 conversions per month for stable data-driven attribution. Most agency clients do not hit that threshold. For a local service business doing 30 conversions a month, data-driven attribution is running on fumes. Worse, when volume is too low, GA4 silently falls back to a last-click model without warning. Two clients on the same GA4 setup can be running on completely different models without anyone noticing. See Google's GA4 attribution documentation and the Google Ads attribution model guide for current thresholds.
| Model | Best for | Watch out for |
|---|---|---|
| Last-click | Performance-focused clients, direct response, short sales cycles | Kills awareness channel budgets over time |
| First-click | Growth-stage clients, new audience acquisition | Ignores everything after discovery |
| Linear | Multi-channel campaigns where every touch matters | Inflates incidental mid-funnel touches |
| Time-decay | Long B2B sales cycles, complex journeys | Hard to explain, undervalues awareness |
| Position-based | Agencies managing both brand and performance | Arbitrary 40/40/20 split |
| Data-driven | High-volume e-commerce, mature ad accounts | Needs 300+ monthly conversions; fails silently at low volume |
Cross-channel attribution gaps
Here is the part that makes attribution genuinely hard: each platform runs its own model with its own data, and they will never agree.
GA4 uses a lookback window for data-driven attribution and only sees interactions it can track (mostly website visits and events).
Google Ads uses its own lookback window and counts a conversion if the click happened within that window, regardless of what GA4 says. For more on how these windows work, see our Google Ads attribution windows guide.
Meta Ads uses a 7-day click and 1-day view window by default and counts view-through conversions that GA4 does not see at all because GA4 cannot track ad impressions that did not result in a site visit. Our guide to the Facebook Ads vs GA4 discrepancy covers this in detail.
The same 100 conversions will show up as different numbers in each platform. None of these numbers are wrong. They are measuring different things with different rules.
What to tell clients
Do not promise one "true" number. It does not exist. Instead, frame it this way:
Pick a primary model and use it consistently for trend analysis. Month-over-month comparisons are valid as long as the model stays the same.
Show the range across platforms for any channel where the numbers diverge materially. "Paid search drove between 47 and 52 conversions this month depending on the attribution window" is honest and useful.
Watch for trends, not absolutes. If a channel's attributed conversions are rising month over month, it is probably contributing more, regardless of whether the absolute number is perfectly accurate.
Revisit the model annually. As the client's business changes (new channels, different sales cycle, higher volume), the right model may change too.
SyncingBeam shows the same underlying data through different lenses, so you can present the cross-source range to clients instead of hiding the disagreements. The model you choose shapes the story, and understanding each model's distortion is what separates a defensible report from a misleading one. For a deeper look at how GA4 and Google Ads numbers diverge in practice, see our definitive guide to GA4 vs Google Ads discrepancies.