Last-click attribution gives 100% of the credit to a customer’s final click before they convert. Multi-touch attribution (MTA) splits credit across every touchpoint in the journey, impressions, clicks, and cross-device visits included. The two produce different numbers because last-click simply ignores everything that happened before that final click.
Pick the wrong one, and you’re defending a budget with numbers that are quietly lying to you. It changes which channels look profitable, which get funded, and which get quietly cut. This guide covers what each model actually does, what last-click structurally can’t see, when it’s genuinely fine to keep using it, and what to test and ask when you’re evaluating attribution software.
What is last-click attribution
Last-click attribution is a rule that assigns 100% of the credit for a conversion to the final touchpoint a customer interacted with before converting, usually the last ad clicked, or the last search that led to the site. Every touchpoint before that gets nothing.
It’s the default in most free web analytics platforms, and for good reason: it’s the simplest model to implement and the easiest to explain. There’s no weighting logic to defend, no model to audit, no data science required. If a customer clicked a paid search ad and bought thirty seconds later, paid search gets 100% of the credit, full stop. For a business with short, single-channel purchase paths, someone searches, clicks, buys, that simplicity isn’t really a limitation. It’s an accurate enough picture of what happened.
The problem is that most customer journeys aren’t that simple, and last-click doesn’t know the difference. It applies the same one-touchpoint rule whether the buyer’s path took ten seconds or ten weeks.
What is multi-touch attribution
Multi-touch attribution (MTA) is a measurement methodology that distributes conversion credit across multiple touchpoints in a customer’s journey, rather than assigning it all to one. “MTA” describes that methodology, not a single method and the models inside it vary enormously in how they decide who gets what credit.
At the simple end are rules-based models: first-click (100% to the first touchpoint), linear (equal credit to every touchpoint), time-decay (more credit to touchpoints closer to conversion), and position-based or U-shaped models (a fixed split, commonly 40% first touch, 40% last touch, 20% to everything in between). These are more informative than last-click, but the weighting is still a fixed assumption applied to every journey, whether or not it fits.
| Dimension | How credit is assigned | Main limitation vs. data-driven |
|---|---|---|
| Last-click | 100% to the final touchpoint | Ignores the entire journey, overvalues branded search |
| First-click | 100% to the first touchpoint | Overvalues top-of-funnel, ignores conversion drivers |
| Linear | Equal credit to all touchpoints | No distinction between high and low-impact channels |
| Time-decay | More credit to touchpoints closer to conversion | Systematically undervalues awareness channels |
| U-shaped | 40% first, 40% last, 20% middle | Middle touchpoints underweighted regardless of actual impact |
At the other end is data-driven attribution (DDA), sometimes called algorithmic or AI-driven attribution. Instead of applying a fixed rule, a data-driven model is trained on actual conversion patterns and estimates each touchpoint’s real contribution to the probability of converting, a video view on one channel might raise conversion probability by half a percent, a follow-up click on another by two percent, and so on down the journey. It’s worth being precise here: “MTA” and “AI attribution” are not interchangeable terms. Data-driven attribution is one type of MTA model, generally the most accurate when it has enough conversion volume and complete-enough data to train on, not a synonym for the whole methodology.
That last qualifier matters more than it sounds. A data-driven model is only as good as what it’s trained on. If the underlying data only captures clicks, no impressions, no cross-device journeys, the model will still learn from an incomplete picture, no matter how sophisticated the algorithm is. Attribution methodology and measurement capability are two different problems, and solving one doesn’t automatically solve the other. We’ll come back to that distinction throughout this guide, because it’s the single most common gap buyers miss when evaluating MTA software.
Last-click vs. multi-touch attribution: side-by-side comparison
| Attribute | Last-click attribution | Multi-touch attribution |
|---|---|---|
| Visibility | Tracks only the session that ended in conversion, typically via one cookie or click ID. Walled-garden impressions and any activity before the last click aren’t counted. | Can include impressions, earlier clicks, and cross-device sessions, but only if the platform’s underlying measurement actually collects that data. Not every MTA tool captures impressions or resolves cross-device identity by default. |
| Credit logic | 100% of credit to the final touchpoint, every time, regardless of channel or journey length. | Ranges from fixed heuristics (first-click, linear, time-decay, position-based) to data-driven models that learn credit weights from real conversion patterns. |
| Auditability | Trivial to audit, one rule, no calculation, but structurally biased toward whatever channel tends to close. | Rules-based models are equally easy to audit but the weighting is still arbitrary. Data-driven models need the vendor to expose methodology, confidence levels, and adjustable parameters, or they function as a black box. |
| Typical accuracy caveats | Systematically overcredits bottom-funnel and direct traffic, and undercredits upper-funnel and awareness spend (see below). | Bounded by the completeness of the underlying data. A data-driven model trained on click-only data will still miss impressions and cross-device journeys. Better methodology doesn’t fix a data gap by itself. |
| Implementation effort | Effectively none. Most free web analytics tools default to it already. | Ranges from a moderate lift (connecting ad platforms and order/CRM data) to a heavier one if impression-level or cross-device tracking has to be built from scratch. Timeframes vary widely by vendor; ask directly (see the checklist below). |
What last-click actually misses
Last-click’s blind spots are structural: every touchpoint before the final click gets zero credit, and that gap widens as the journey spans more platforms, devices, and sessions.
Walled gardens. Platforms like Meta, TikTok, YouTube, Snap, and Pinterest don’t expose impression-level data to third-party tracking. If a customer sees a video ad on one of these platforms and doesn’t click, that impression is invisible to last-click measurement entirely, even though it may have started the journey. This is also why platform self-reported numbers and independent last-click numbers for the same channel routinely disagree by a wide margin: each platform is scoring against its own incomplete view.
One measured example: Roivenue’s Synthetic Impressions methodology found Meta’s real credit share came out to 272%, compared with 95% under a standard data-driven attribution model applied to click-only data, and 673% under Meta’s own self-reported view-through numbers. Roivenue’s Synthetic Impressions estimate impression-level contribution from walled gardens using post-view conversion matching, without third-party pixels. The three numbers disagree because they’re measuring three different slices of the same journey, click-only data misses the impressions that led to the click, and platform self-reporting has every incentive to claim generously.
Cross-device journeys. A customer researching on their phone during a commute and completing a purchase on a laptop that evening looks, to a cookie- or click-based tool, like two unrelated visitors. Without a way to probabilistically or deterministically link those sessions, last-click credits whatever touchpoint happened to close the second, unrelated-looking visit and drops the research phase entirely.
Cookie loss and consent gaps. Safari and Firefox already block third-party cookies by default, and GDPR- and CCPA-style consent requirements mean a meaningful share of visits never get a trackable cookie in the first place, regardless of browser. Every visit that falls into that gap is a visit last-click can’t connect to anything. It just disappears from the reported journey, taking whatever channel touched it with it.
Put together, these gaps explain a pattern that shows up consistently in attribution data: across Roivenue’s client base, roughly 70% of conversion journeys involve two or more touchpoints, and about 10% span more than ten. Under a last-click model, all of that upstream activity, impressions, earlier clicks, other devices, is credited to nothing.
When last-click is good enough
It’s tempting for anyone selling MTA software to imply every advertiser needs it immediately. That’s not honest, and it’s not this guide’s position.
Last-click is a reasonable, low-cost choice when:
- You run on one or two channels, and the customer journey genuinely tends to be one click and done, a single paid search campaign driving direct-response purchases, for instance.
- Purchases happen on one device, so there’s no cross-device journey to lose in the first place.
- The sales cycle is short, hours or days, not weeks, so there’s little upper-funnel activity happening far enough in advance to matter.
- Budget and team size don’t support acting on more complexity. A more granular attribution model is only useful if someone has the time to actually change budget allocation based on what it shows. If nobody’s going to act on the extra data, the extra data isn’t paying for itself yet.
The signal that it’s time to move past last-click isn’t a specific revenue threshold. It’s growing channel count, growing markets, and growing uncertainty about where budget should actually go. If you’re running impressions on walled-garden platforms, marketing across multiple devices, or increasingly guessing at whether upper-funnel spend is working, last-click’s blind spots are actively costing you decision quality, not just measurement precision.
Frequently Asked Questions
What’s the difference between last-click and multi-touch attribution?
Last-click attribution gives 100% of conversion credit to the final touchpoint before a purchase. Multi-touch attribution distributes credit across every touchpoint in the journey, which can include impressions, earlier clicks, and different devices, depending on what the underlying platform actually measures. The two will report different numbers for the same campaign whenever a customer’s path involved more than one touchpoint.
Is multi-touch attribution worth it for a small marketing team?
It depends more on channel and journey complexity than team size. A small team running one or two channels with short, single-device purchase paths may get an accurate enough picture from last-click. A small team running paid social, search, and video across a multi-week consideration cycle is likely underinvesting in the channels last-click can’t see, regardless of headcount.
Does multi-touch attribution replace Google Analytics?
Not automatically, and not by itself. Standard web analytics tools measure on-site behavior using session and click data; they generally don’t capture impressions from walled-garden platforms or resolve cross-device identity without additional tracking infrastructure. A full-featured MTA platform typically replaces the analytics layer for attribution and journey reporting specifically, rather than sitting on top of a click-based tool and inheriting its blind spots.
How accurate is AI-driven attribution?
It depends entirely on the data it’s trained on. A data-driven attribution model applied to complete, high-volume conversion data, including impressions and cross-device journeys, will produce meaningfully more accurate credit assignment than a fixed rule like last-click. Applied to incomplete data, the same model will still produce a more nuanced-looking number, but one still built on the same gaps. Ask any vendor what data feeds their model before evaluating the model itself.
How long does it take to implement multi-touch attribution?
This varies significantly by vendor and by how much of your data infrastructure (ad platform connections, order data, margin data) already exists in a usable form. Ask any vendor for a week-by-week implementation timeline rather than a single number.
45 minutes on relevant demo data, with a summary for Finance.
