learn-marketing-attribution-with-phoebe / Leader session 1 of 6
Learn Marketing Attribution with Phoebe · Leader session 1 of 6

The credit problem: who gets the sale?

Five channels touched one customer. One bought a $92 order. Which channel gets the credit - and who decides? This session is the whole course in miniature: one real journey, eight ways to split the credit, eight different answers, and the reason your last-click dashboard has been quietly lying to you. No math, no code - just the question every marketing dollar depends on.

🟢 Start here Leaders · CMO / CDAIO No tech required 45 min live + self-study
0-3 · Welcome 3-16 · The credit problem 16-38 · Split it yourself 38-45 · Q&A
Part 0

Why this course exists

Every marketing team on earth reports a number that says "this channel drove the sale." Almost none of them can defend it. Attribution is the discipline of assigning credit for a conversion across the touchpoints that led to it - and it turns out that innocent-looking word, "credit", is one of the most consequential and most misunderstood decisions in a marketing budget. This leader track teaches you to read those numbers, trust the right ones, and reallocate a budget without getting fooled. Six sessions, one real brand, zero code.

Live - presented in session Self-study - read after class ★ Work-along - everyone does it The brand: Lumen Skincare
★ What you walk out with today A gut-level feel for why the same customer journey produces wildly different "who won" answers depending only on the rule you pick - and the confidence to ask your team "which model is this, and why that one?" the next time a channel report lands on your desk.
Part 1 · the running project

Meet Lumen, and one real journey 5 min live

Both tracks of this course work on one brand: Lumen Skincare, an $18M/year direct-to-consumer skincare label with nine marketing channels and a $4M media budget its CMO needs to defend. We start with a single customer - one journey we will carry through every session and every model.

📺 Display impression · day -14 📱 Paid social click · day -7 ✉️ Email open · day -3 🔎 Organic click · day -1 💳 Paid search click · day 0 🛒 CONVERT · $92 Fourteen days, five touches, one purchase. Now: which of the five earned the $92? Hold that question - it has no single right answer.
🔍 Click to zoom - the canonical Lumen journey (we reuse this exact path all course long)
LiveWhat "attribution" actually means3 min

Attribution is the act of distributing credit for one conversion across the touchpoints that preceded it. That's it. The customer above saw a display ad, clicked a paid-social post, opened an email, searched, and searched again - then bought. Attribution decides how much of that $92 each touch "earned". Every marketing report you have ever read made this decision for you, silently, using a rule someone chose - often without telling you which one.

  • It is a credit-splitting rule, not a fact. There is no laboratory that can tell you the "true" contribution of the email open. Attribution is a modeling choice dressed up as a measurement.
  • The rule you pick moves money. If your model hands all the credit to the last click, you will pour budget into the channels that show up last - and starve the ones that created the demand in the first place.
  • Leaders own the rule, not the math. You don't need to build the model. You need to know which one is running, what it flatters, and what it hides.
Real world

A DTC brand cut its entire display budget because the last-click dashboard showed display "drove" almost no sales. Revenue dropped 18% over the next quarter - because display was the channel that started most journeys. The dashboard wasn't wrong; it was answering a different question than the one the CFO thought it was answering. That gap is this whole course.

LiveThe nine channels and the shape of a journey2 min

Lumen spends across nine channels. Some create demand (people who didn't know they wanted you), some capture demand (people already looking). Attribution is really a fight over how much each side gets paid.

Mostly creates demandMostly captures demand
Display, Paid social, Influencer, CTVPaid search, Organic search, Direct
Show up early, rarely get the last clickShow up late, hog the last click

Email and affiliate sit in the middle. Hold this create-vs-capture split in your head - it is the reason last-click attribution systematically over-pays search and under-pays social, and why the "hero channel" in your dashboard may just be the channel standing closest to the finish line.

Part 2 · the payoff

One journey, eight models, eight answers 6 min live

Here is the moment the whole field clicks. We take that exact same $92 journey and run it through eight attribution models. Nothing about the customer changes. Only the rule changes. Watch where the $92 lands.

Display P.social Email Organic P.search First-touch display $92 Last-touch paid search $92 Linear $18.40 each Time-decay recency wins Position (U) 40% 40% ends Last non-direct paid search $92 (no direct in path) Markov* learned from all journeys Shapley* rewards channels that lift others The heuristic models (rows 1-6) use a FIXED rule - no data learning. The last two (*) learn credit from the whole dataset. Same customer, same $92, same five touches. The ONLY thing that changed between rows is the rule. That is the credit problem. Notice first-touch and last-touch are perfect opposites - and most dashboards quietly run last-touch.
🔍 Click to zoom - the same journey, eight models, eight answers (Markov + Shapley bars are illustrative of learned spread)
LiveRead the chart like a leader3 min

You do not need to compute any of these. You need to see the pattern:

  • First-touch and last-touch are mirror images. One crowns the opener (display), the other crowns the closer (paid search). Both are single-touch: they throw away 80% of the journey. Most ad platforms default to last-touch - which is why every platform thinks it deserves the credit.
  • The "fair-looking" middle models (linear, time-decay, position) are still guesses. Linear says every touch is equal - is an impression really worth as much as the final click? Time-decay says recent is better - says who? Position says the first and last matter most - a reasonable story, but still a story you asserted, not learned.
  • The learned models (Markov, Shapley) look at all your journeys at once and ask a sharper question: what actually changes when this channel is present versus absent? That's Session 3 - and it's where credit stops being a guess.
The one question to carry into every meeting When someone shows you channel performance, ask: "Which attribution model is this, and what would it look like under last-touch versus data-driven?" The answer tells you whether you're looking at a fact or a flattering rule.
Self-studyThe eight models in one table3 min read
ModelRule in one lineOver-credits
First-touch100% to the first touchAwareness channels
Last-touch100% to the final touchSearch & retargeting
Last non-directLast touch, but skip "direct"Same, minus self-navigation
LinearEqual split across all touchesLow-value impressions
Time-decayMore credit to recent touchesBottom-of-funnel
Position (U)40% first + 40% last, 20% middleEndpoints over nurture
MarkovCredit from "removal effect" across all pathsLearned - fewer blind spots
ShapleyAverage marginal contribution across coalitionsLearned - rewards synergy

Session 2 goes deep on the six heuristics; Session 3 decodes Markov and Shapley without the math. For now, just internalize: the top six are rules you assert, the bottom two are patterns you learn.

Part 3 · the map

Three eras, stacked on top of each other 4 min live

Attribution didn't replace itself over time - it accumulated. Three eras of thinking now sit in every marketing stack at once, and a modern leader needs to know which one is answering which question.

Era 1 · Heuristics Fixed rules: first, last, linear, time-decay, position Arbitrary but simple. Still the tool default. Era 2 · Data-driven Learned credit: Markov, Shapley, ML, GA4 DDA Smarter, but still correlational - not proof. Era 3 · MMM + lift Mix modeling + geo incrementality experiments The 2024-26 pivot - privacy-proof, causal. Only Era 3's experiments are truly causal. Eras 1 and 2 answer "who correlates with the sale", not "who caused it". You will meet all three. The skill is matching the era to the decision: heuristics for a quick read, data-driven for tactics, MMM + lift for the budget.
🔍 Click to zoom - three eras of attribution, all live in your stack at once
LiveWhere this course takes you2 min

Your leader track walks these three eras in order: heuristics (Session 2), data-driven decoded (Session 3), then the big three-way fight - MMM vs multi-touch vs incrementality (Session 4), the 2026 privacy break that reshuffled everything (Session 5), and finally turning all of it into a defensible budget move (Session 6). Your builders, on the parallel track, actually build each model on Lumen's data.

Real world

The most expensive attribution mistake is using the wrong era for the job: running a last-touch heuristic (Era 1) to make a strategic budget reallocation that needed a mix model and a holdout test (Era 3). It feels rigorous - there's a dashboard! - but you're using a bathroom scale to weigh a truck.

Work-along 1 of 2

Interrogate a last-click report ★ 10 min · everyone

We put Lumen's actual channel report on screen - the one the last-click dashboard produces - and pull it apart together. Bring your own report next week; today we use Lumen's.

Look at Lumen's last-click summary: paid search 41% of "attributed" revenue, paid social 12%, display 3%. On paper, search is the hero and display is a waste.

Ask the create-vs-capture question out loud: which of these channels do customers find you through, and which do they come back through? Search is mostly capture - people already looking for "Lumen vitamin C serum".

Now ask the killer follow-up: "How many of those paid-search conversions had a display or social touch earlier in the journey?" On Lumen's data, 68% did. Last-click gave those upstream channels zero.

Write the one-sentence verdict a leader should say: "This report tells me who closed, not who created - I will not cut display on this number alone."

The move that earns respect You don't need to know the right answer in the room. You need to name which question the number answers. "Who closed" and "who mattered" are different questions - saying so out loud is what separates a leader who owns attribution from one who is owned by it.
Work-along 2 of 2

Split the $92 yourself ★ 10 min · pen and paper

No spreadsheet. Take the five-touch Lumen journey and split the $92 by hand under four models. Feeling the arbitrariness is the point.

First-touch: all $92 to display. Write it down. Notice how it ignores the four touches that actually moved the person toward buying.

Last-touch: all $92 to paid search. The opposite verdict, same customer. If your dashboard runs this, display looks worthless.

Linear: $92 / 5 = $18.40 to each of the five touches. Fairer-feeling - but is an impression really worth the same as the final click?

Position (U-shaped): 40% to display ($36.80), 40% to paid search ($36.80), and the remaining 20% ($18.40) split across the three middle touches (~$6.13 each). A story about "create and close matter most".

Look at your four rows. Same journey. Display earned anywhere from $0 to $92 depending only on the rule. That range is the credit problem - and no amount of cleaner data makes it go away. Only a better model does.

Real world

Run this exercise with your own leadership team and watch the argument start - the paid-search owner loves last-touch, the brand owner loves first-touch, and everyone is "right" under their preferred rule. That argument, made explicit, is worth more than any dashboard: it reveals that your reporting has been quietly picking a winner all along.

Before Session 2

This week ◐ 25 min total

Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · Attribution is best described as...

There is no lab that reveals the "true" credit. Attribution is a choice dressed as a measurement - which is why the rule you pick moves real money.

2 · Why does last-touch attribution systematically over-pay paid search?

Create-vs-capture: search captures people already looking, so it sits near the finish line. Last-touch hands the whole prize to whoever's standing there.

3 · A colleague wants to reallocate the annual media budget using the last-click dashboard. Your leader move?

Match the era to the decision. Last-touch is a quick tactical read; a budget reallocation needs a mix model and an incrementality test - Sessions 4 through 6.

Source material

What this session covers

This session distills the opening of the leading attribution courses - the "what is attribution and why do the models disagree" foundation - into a leader-first framing. Certificates and full video courses stay with their official sources.

Marketing Attribution and Mix Modeling (LinkedIn Learning) - ch.1what attribution is, the model families - Parts 1-2
Marketing Measurement: Attribution to Incrementality (Udemy) - introthe three eras framing - Part 3
Google Analytics attribution models (Skillshop)the eight standard models - Part 2 table
The heuristic models in depthfull treatment in Leader Session 2
Markov & Shapley (data-driven)decoded without math in Leader Session 3

Leader Session 1 cheat sheet · pin this

Attribution =Splitting credit for one conversion across the touchpoints before it. A rule you choose, not a fact you measure.
The credit problemSame journey, different rule, different winner. Display earned $0 to $92 on one journey - only the model changed.
Create vs captureDisplay/social/influencer create demand (show up early); search/direct capture it (show up last). Last-touch over-pays capture.
Three eras1 Heuristics (fixed rules) · 2 Data-driven (learned, still correlational) · 3 MMM + incrementality (causal, privacy-proof).
The meeting question"Which model is this, and what would it look like under last-touch vs data-driven?" Fact or flattering rule?
Match era to decisionHeuristic = quick read. Data-driven = tactics. MMM + holdout test = budget. Don't weigh a truck on a bathroom scale.