Last-touch attribution gives 100% of the credit for a conversion to the final marketing interaction before it. Everything earlier in the journey gets nothing. It’s the default in most analytics tools, it’s the reason your channel report looks the way it does, and it’s simultaneously the most used and most criticised model in marketing measurement.
Both the use and the criticism are justified, for different accounts. Here’s how it works, when it’s fine, and how to tell which situation you’re in.
How the credit is assigned
A customer’s path to purchase might look like this:
- Sees an Instagram ad → doesn’t click, but remembers the brand
- Clicks a YouTube ad three days later → browses, leaves
- Reads a review article the following week → clicks through, leaves
- Searches the brand name on Google → clicks the branded search ad → buys
Under last-touch, the branded search ad gets 100% of the revenue. The Instagram ad, the YouTube ad and the review get zero. Not “a small share” — zero.
Two variants you’ll see named:
- Last click. Credit to the last interaction the user clicked. This is what most platforms mean when they say last touch.
- Last non-direct click. Direct traffic is skipped and credit passes to the last identifiable marketing source before it. This is what Universal Analytics used as its default, and it exists because giving credit to “someone typed the URL” tells you nothing actionable.
The mirror image is first-touch attribution, which gives all the credit to the first interaction. It has the opposite bias — it over-credits discovery and ignores everything that closed the sale — and it’s wrong in the same structural way.
Why it’s the default
Three genuine advantages, which is why it survives all the criticism:
It’s unambiguous. There’s no model, no weighting, no assumption. One interaction, one rule. Two people looking at the same data get the same answer, and nobody has to defend a methodology.
It’s cheap. It needs no identity resolution, no cross-device stitching, no modelling. Every platform can produce it from data it already has.
It’s directionally fine for short journeys. If most of your customers see one ad and buy the same day, last touch and reality are close enough that the difference doesn’t change decisions.
Where it systematically misleads
The failure isn’t random noise — it’s a consistent bias in one direction, and that’s what makes it dangerous.
It over-credits the bottom of the funnel. Branded search is the classic case. Someone who already decided to buy searches your brand name, clicks the ad, converts. Last touch hands that campaign a spectacular ROAS. The campaign didn’t create the demand; it charged you for demand something else created. Cut the upper-funnel spend that created it, and the branded search numbers stay beautiful for a few weeks while the pipeline quietly empties.
It under-credits everything that can’t be clicked. Video views, podcast mentions, paid social impressions, the review article, the colleague’s recommendation. If it doesn’t produce a click immediately before the conversion, it doesn’t exist in the report.
It breaks entirely on long sales cycles. Which brings us to the case where this question gets asked most.
Last-touch attribution in B2B SaaS
B2B SaaS is where last touch stops being a simplification and becomes actively wrong. Four reasons stack up:
The cycle outlives the window. A ninety-day evaluation with eleven touchpoints, measured through a thirty-day click window, means most of the journey is invisible before you start.
The buying unit isn’t a user. A champion researches, a manager evaluates, a VP approves, procurement signs. Four humans, four devices, often four different acquisition paths — and your analytics sees four unrelated sessions, one of which happened to convert.
The conversion event isn’t the revenue event. A demo request is not a closed-won deal. Last touch tells you which channel produced form fills. It cannot tell you which channel produced the form fills that became contracts, and those are frequently different channels. The one that produces cheap demos is often the one producing demos that never close.
The last touch is almost always branded or direct. By the time a B2B buyer converts, they know your name. So the model credits the channel that captured an already-decided buyer and ignores the eighteen months of content, events and paid social that made them decide.
The practical consequence is that B2B SaaS teams optimising to last-touch CPA systematically underfund demand creation and overfund demand capture, then wonder why capture gets more expensive every quarter. There’s nothing left to capture.
When last touch is genuinely fine
It’s not always wrong. Keep it when:
- The journey is short. Impulse ecommerce, low-consideration purchases, single-session conversions. If your median time-to-conversion is under a day and the median path length is one or two touches, a multi-touch model will give you the same answer with more machinery.
- You’re comparing like with like inside one platform. Deciding which of two Google Ads campaigns gets more budget is a fair last-touch question, because both are measured the same way.
- You have one channel. The model can only get it wrong when there’s something to take credit from.
- You’re being honest about what it is. A directional signal for in-platform decisions, not a statement about what caused the revenue.
What to use instead
No single model replaces it. A realistic stack, in order of effort:
Blended efficiency as the judge. Total revenue divided by total media spend — MER. It can’t be inflated by attribution because neither number comes from an ad platform. Watch it month over month as you shift budget. Covered in how to calculate ROAS.
Incrementality tests. Geo holdouts, or turning a channel off in a controlled way and measuring what happens to total conversions. This is the only method that answers “what did this channel actually cause.” It costs you some revenue to run, and it’s worth it.
Data-driven or multi-touch attribution for allocation within paid. Better than last touch for splitting credit across channels you can observe. Still blind to everything unclickable. Details in why last-click lies.
Revenue-connected reporting for B2B. Push closed-won data back into your analytics and your ad platforms so the optimisation target is the deal, not the form fill. This usually means offline conversion imports and server-side tracking, and for deeper analysis, GA4’s BigQuery export.
Marketing mix modelling, once spend is large enough to justify it. Google’s open-source Meridian reached version 2.0 this month, adding brand-equity signals like branded query volume — useful precisely because branded search volume is the thing last touch mistakes for a channel’s performance.
The short version
Last-touch attribution assigns all credit to the final click. It’s unambiguous, cheap and roughly fine for short journeys and in-platform decisions. It has a consistent bias toward the bottom of the funnel, and on long or multi-stakeholder journeys — B2B SaaS above all — that bias will lead you to defund the thing that creates your pipeline.
You don’t need to replace it with a perfect model. You need one number it can’t distort (blended efficiency) and one method that establishes causation (incrementality tests). With those two in place, last touch becomes what it should always have been: a useful in-platform signal that nobody mistakes for the truth.
If your channel report says one thing and your bank account says another, measurement is usually the first thing I audit.