The last-click trap: who really earned the sale?

1. Before you start

Attribution is how you decide which marketing touch gets the credit when a customer finally buys. A customer almost never sees one ad and checks out; they meet your brand many times — a creator’s video, a scroll-past ad, an email, a search for your name — and only one of those touches is the last one before the order lands. An attribution model is the rule that splits the credit across those touches. A tiny example: someone watches a creator’s video on Monday, forgets about it, then on Friday searches your brand name and buys. Last-click attribution gives 100% of the credit to Friday’s brand search and 0% to the video that started it. First-click flips it: 100% to the video, 0% to the search. Same sale, two completely different stories about what earned it — and marketing budgets get moved on the strength of those stories.

Three honest statements before you start:

This is a Decide course. You read a situation, learn the concept, and make a call. You do not write or run any code, and there is no calculator here — the reasoning is the work.
The companies in the cases, Fernwood & Co. and Brindle Coffee Club, are composite — invented direct-to-consumer brands built from ordinary, realistic dynamics. Every channel number you see is an in-course illustrative figure chosen for clean teaching, not a measurement of any real company.
This is not a certification. It proves, to you, that you can read an attribution report without being fooled by it and defend where the budget should go.

If you have ever looked at a marketing dashboard and wondered “did that channel really do that?”, you are in the right place.

2. The Situation

Fernwood & Co. is a direct-to-consumer home-fragrance brand — soy candles and reed diffusers sold only through its own website — and its growth lead has just been handed a blunt instruction: “the creator and prospecting spend converts almost nothing, so cut it and put the money into the channels that actually sell.” That instruction comes straight off the last-click report, where two top-of-funnel channels show a rounding-error number of orders while brand search and retargeting look like the whole business. The trap is that the report is telling the literal truth about last clicks and a dangerous lie about what earned the sale — and if the growth lead acts on it, demand will quietly dry up a month later.

3. What you’ll be able to do

After this course you will be able to:

Read a channel report and say which attribution model produced it, and therefore which touches it systematically over-credits and which it starves.
Explain, to a finance partner who only trusts the last-click number, why last-click credit pools at the bottom of the funnel by construction — not because top-of-funnel channels are weak.
Decide whether to cut, keep, or test a channel that a last-click report undervalues, and name the one piece of evidence that would settle it.

4. Prerequisites & time box

Difficulty: 5 / 8 — a manager-level decision: several factors move at once and the naive reading points the wrong way, so you have to reason past it.

Time box: about 22 minutes of reading (measured), plus your own thinking time on the call in section 7. That is under the 25-minute cap for a concept course.

Free-tier honesty: nothing to sign up for; requires_gpu: false. Everything is in this page.

5. The case & where the numbers come from

Fernwood & Co. (the taught case, section 6) and Brindle Coffee Club (the transfer, section 7) are composite companies: invented direct-to-consumer retailers assembled from typical channel dynamics. Every order count, budget, and channel figure below is an in-course illustrative assumption, chosen so the direction of the bias is easy to see. No number here is drawn from, or claimed about, any real brand.

What is cited, in section 11, is the standard definition of each concept the course teaches — what marketing attribution is, what last-click, first-click, and multi-touch attribution mean, the shape of the purchase funnel, what behavioral retargeting is, and the alternative measurement approaches (incrementality testing and marketing-mix modeling) — because those are established ideas with public references, not invented for this course.

Here is Fernwood’s last-click channel report for one month. Every downstream point in the course reads off this table (illustrative figures):

Channel	Where it sits in the funnel	Orders credited (last-click)
Brand search (people searching “Fernwood”)	Bottom	380
Retargeting (ads to people who already visited)	Bottom	260
Email	Bottom / middle	150
Non-brand search (“soy candle gift”)	Middle	110
Paid social prospecting (cold Meta ads)	Top	60
Creator / influencer content	Top	40
Total new-customer orders		1,000

6. The Concepts

What an attribution model is

A customer’s path to a purchase is a sequence of touches across channels. Attribution is, in the standard definition, “the identification of a set of user actions (‘events’ or ‘touchpoints’) that contribute to a desired outcome, and then the assignment of a value to each of these events.” The model is the rule for that assignment. It answers one question: when the order finally lands, how much credit does each touch get?

This matters because credit drives money. Marketers rank channels by the credit the model hands them, then move budget toward the top of the ranking and away from the bottom. So the choice of model is not a reporting detail — it silently decides which channels get funded next quarter. Every model is an assumption about causation dressed up as a measurement, and the trap is forgetting the assumption and treating the output as simple fact.

Last-click attribution

Last-click gives 100% of the credit to the final touch before the purchase — the last ad, link, or search the customer clicked. It is the default in most ad platforms and analytics tools because it is simple, deterministic, and needs no modelling: whoever the customer touched last wins the whole order.

Its assumption is that the last touch is what caused the sale. Its limit is that it ignores every touch before the last one. Watch it work on a single Fernwood customer, Maya (illustrative journey):

Day 1 — Maya watches a creator’s video about Fernwood candles. (top of funnel)
Day 4 — she scrolls past a cold Meta prospecting ad. (top of funnel)
Day 9 — she searches “Fernwood” and clicks the brand-search result. (bottom)
Day 10 — a retargeting ad reminds her; she clicks and buys. (bottom)

Under last-click, retargeting gets 100% of the credit and the creator video, the prospecting ad, and the brand search all get zero. The channel that closed the sale takes everything the four channels earned together. Do that across thousands of journeys and you get exactly Fernwood’s report: the bottom-funnel closers (brand search, retargeting) look like the whole business, and the top-funnel channels that started the journeys look worthless.

First-click attribution

First-click (also called first-touch) is last-click’s mirror image: it gives 100% of the credit to the first touch and zero to everything after. On Maya’s journey, first-click hands the entire order to the Day-1 creator video and gives nothing to the brand search or retargeting that were present when she actually bought.

Its assumption is that the touch that first brought the customer in is what caused the sale. Its limit is the opposite of last-click’s: it over-credits openers (creators, prospecting, awareness) and starves closers (retargeting, brand search, email). This is the key point most people miss: first-click is not the “fix” for last-click. Swapping one single-touch model for another does not remove the bias — it just moves the distortion to the other end of the funnel. Both models are wrong in a known direction, which is at least useful to know.

Multi-touch attribution

Multi-touch attribution (MTA) spreads the credit across several touches in the path instead of dumping it all on one. Common rules:

Linear — split the credit evenly. Maya’s four touches each get 25%.
Position-based (often 40/20/40) — give the first and last touches 40% each and split the remaining 20% across the middle. On Maya’s path the creator video gets 40%, retargeting gets 40%, and the two middle touches share 20% (about 10% each).
Time-decay — touches closer to the purchase get more credit than earlier ones.
Data-driven — an algorithm assigns credit from patterns across many converting and non-converting paths.

Its assumption is that every touch appearing in a converting path contributed something, so each deserves a share. That is closer to how buying actually works — but MTA has real limits too. It is still built on correlation, not causation: it distributes credit among the touches it saw, and it cannot see touches it never tracked (offline word of mouth, a friend’s recommendation, an ad viewed but not clicked). And it can be fooled by selection bias — handing credit to a channel that merely appears in high-converting journeys without causing them. This is why the standard reference warns that attribution models “frequently diverge from lift measurements obtained through controlled experiments.” Even multi-touch is a better story, not ground truth. The only way to know a channel’s causal contribution is to turn it off for part of your audience and watch what happens to total sales — an incrementality or holdout test.

Why last-click starves the top of the funnel

Now put the pieces together, because this is the whole course. Journeys flow top to bottom: awareness first (creators, prospecting, non-brand search), then consideration, then the closing touches (brand search, retargeting, email). By the time someone buys, the last touch is almost always a bottom-funnel channel — that is what “bottom of the funnel” means. So last-click credit pools at the bottom by construction, no matter what actually created the demand. It is not measuring that retargeting is a great channel; it is measuring that retargeting is near the exit, where the last click tends to happen.

Here is the failure mode, step by step:

The last-click report shows top-of-funnel channels converting almost nothing (Fernwood: creator 40, prospecting 60) and bottom-of-funnel channels converting everything (brand search 380, retargeting 260).
Budget follows credit, so the “underperforming” top of funnel gets cut.
Four to eight weeks later, fewer new people have entered the funnel — so demand dries up.
Now even the bottom-funnel closers shrink, because brand search and retargeting can only harvest demand that something upstream created. There is less to harvest. Retargeting needs site visitors to retarget; brand search needs people who already know the brand name to search it. Cut the channels that produce those people and the “heroes” fade too.

That is the last-click trap: a report that is literally accurate about last clicks leads straight to cutting the demand engine and then wondering why the reliable channels also declined. The bias is systematic and directional — it always over-credits the bottom and starves the top — so once you know the shape of it, you can correct for it every time you read one of these reports.

7. Your Call

You have read Fernwood’s report and learned where last-click bias points. Now a different company, a different pressure, and a real budget decision land on your desk.

Brindle Coffee Club is a direct-to-consumer coffee-subscription brand (a different company in a different DTC category from Fernwood’s home fragrance). Finance has frozen the monthly ad budget and is demanding a $40,000 cut — from $200,000 down to $160,000 — and wants the growth lead to name exactly one channel to cut. To make it harder, a browser cookie change last month degraded conversion tracking, so retargeting’s tracked orders dropped from 250 to 150 month-over-month. Here is Brindle’s last-click report (illustrative figures):

Channel	Funnel position	Subscriptions (last-click)
Brand search	Bottom	500
Retargeting	Bottom	150 (was 250 before the cookie change)
Email	Bottom / middle	120
Non-brand search	Middle	80
Paid social prospecting	Top	30
Podcast sponsorships	Top	20

How this differs from the taught case (the transfer): this is a different company and DTC sector (a coffee-subscription brand, not Fernwood’s candles), the figures are different so you must reason over new data, it asks a different kind of decision (a fixed-budget reallocation with a mandated single cut, not a read of a single report), and it adds a new constraint — a frozen budget and degraded tracking from the cookie change. The core concept is the same: last-click bias and which touches an attribution model over- or under-credits.

8. Self-check

Before you write the memo, make sure you can say each of these in one sentence, without an answer key:

What is your call on Brindle’s cut, and what single piece of evidence — a holdout test result — would change it?
Why does last-click credit pool at the bottom of the funnel by construction, rather than because bottom-funnel channels are genuinely the best?
What does each rejected option cost you: cutting the lowest last-click channel, switching to first-click, and reading the cookie-driven retargeting drop as a real performance change?

If any of these is fuzzy, reread the last two headings in section 6 — the direction of the bias and the funnel failure loop are the heart of the course.

9. Stretch

Push the thinking further on your own:

Brindle’s brand-search row is its biggest (500 subs). Brand search mostly catches people who already decided to look for the brand. What upstream channels most likely created those searchers — and if you cut them, what happens to the 500 a quarter later?
Suppose a holdout test shows retargeting’s true incremental lift is only a third of its last-click credit — most of those buyers would have converted anyway. Does that change which channel you cut, and how would you explain “incremental versus credited” to finance?
The genuinely hard one: design the cheapest honest test Brindle could run next month to rank its six channels by incremental contribution, given a frozen budget and degraded cookie tracking. What would you turn off, for whom, and what would you measure — and what can this test still not tell you?

10. Ship it — your decision memo

Write a one-page memo to Brindle’s finance partner. State the call (do not cut podcasts or prospecting on the strength of their last-click count; the defensible move is to protect top-of-funnel demand and cut the channel a holdout test shows adds the least incremental volume — and to run that test before committing). Show the reasoning: last-click credit pools at the bottom of the funnel by construction, so the lowest-credited rows are the demand engine, not the waste. Name what you rejected and why (cutting the lowest last-click row; switching to first-click; reading the cookie-driven retargeting drop as a real decline). Name the one thing that would change your mind (a clean incrementality test showing a top-funnel channel truly adds no new subscribers). Keep it to a single page a finance partner grasps in two minutes. This memo is your own argued claim — not a credential.

11. Sources

Fernwood & Co. and Brindle Coffee Club, and every channel figure attached to them, are composite and illustrative — invented from ordinary DTC-retail dynamics for clean teaching, not drawn from or claimed about any real company. What is cited below are the standard definitions of the concepts the course teaches.

Concept / claim	Source (publisher)	URL	Accessed
Marketing attribution; last-click less accurate; multi-touch attribution	Wikipedia — Attribution (marketing)	https://en.wikipedia.org/wiki/Attribution_modeling	2026-07-19
Purchase funnel: awareness-to-action stages, top vs bottom of funnel	Wikipedia — Purchase funnel	https://en.wikipedia.org/wiki/Purchase_funnel	2026-07-19
Customer journey across multiple touchpoints	Wikipedia — Customer journey	https://en.wikipedia.org/wiki/Customer_journey	2026-07-19
Behavioral retargeting: ads to prior site visitors	Wikipedia — Behavioral retargeting	https://en.wikipedia.org/wiki/Behavioral_retargeting	2026-07-19
Marketing-mix modeling as an alternative measurement to attribution	Wikipedia — Marketing mix modeling	https://en.wikipedia.org/wiki/Marketing_mix_modeling	2026-07-19
Brand awareness as a top-of-funnel objective	Wikipedia — Brand awareness	https://en.wikipedia.org/wiki/Brand_awareness	2026-07-19

Next up

Finished this call? Continue the Marketing track:

Brand or performance? Splitting the budget when the dashboard only sees one → · Browse all courses