The only session that moves money
Every session so far taught you to read attribution. This one teaches you to act on it - carefully. The whole point of measuring credit was never the dashboard; it was to reallocate a budget with confidence. But a reallocation is also the most dangerous thing a marketing leader does: move the wrong money and you starve the channel that was quietly creating all your demand. So the capstone skill is not "find the true numbers" - there are none - it is a discipline for moving budget in a way you can defend when a skeptical board asks "how do you know?" Six sessions, one real brand, zero code. Let's land the plane.
The reallocation decision framework 6 min live
Never move budget on one model, and never bet the farm. The whole capstone reduces to a four-step loop you run every time you want to shift money: triangulate, move at the margin, hold out, measure - then round the loop again. It is deliberately boring, and that is what makes it defensible.
LiveThe four rules, said plainly3 min▶
Each step exists to stop a specific way leaders get burned:
- Triangulate - never move on one model. Any single method can be fooled. Only act where your MMM, your data-driven attribution, and a lift test point the same direction. Agreement across independent methods is the closest thing to truth you get.
- Move at the margin - shift 10 to 20%, then measure. A reallocation is a hypothesis, not a verdict. Move a slice, watch what happens, and you can course-correct. Move everything and a mistake is catastrophic and irreversible.
- Protect demand creation. The channels that create demand - social, display, influencer, CTV - rarely get the last click, so every naive model under-credits them. Do not starve the top of the funnel just because the closer got the applause.
- Tie every move to a hypothesis and a holdout. "We believe growing social will lift incremental sales; a geo-holdout will confirm it within 30 days." That single structure turns a budget gamble into an experiment.
Self-studyWhy "move at the margin" beats being right2 min read▶
Leaders love to be right, and attribution punishes that instinct. Because there is no true credit split, you will never have certainty before a move - so the winning strategy is not "be right", it is "be reversible". Shifting 15% of a budget and reading the result gives you two things a big confident bet never does: a real-world measurement of whether your thesis held, and the option to undo it cheaply if it did not. Over a year of marginal, tested moves you converge on a genuinely better budget - and every step of the journey is defensible. The leader who bets the whole budget on one model's say-so is not bold; they are one bad model away from a very public mistake.
Lumen's $4M budget on trial 5 min live
Here is Lumen's actual media split, and here is the tension it creates. Last-touch crowns paid search as the hero. But the data-driven models and the MMM tell a different story: paid social, display, and email are under-credited demand-creators doing work the last click never sees. Below, the current split and a proposed marginal reallocation, side by side.
LiveReading the trial like a leader3 min▶
Notice what the proposal does and does not do:
- It grows the under-credited demand creators - modestly. Paid social 32 to 36, display 10 to 13, email 5 to 6. These are the channels the MMM and data-driven models say do more than the last click reveals.
- It trims the last-click darling and the weakest tail. Paid search 24 to 19 (it will still capture the demand that already exists; we are just no longer over-funding a harvester), affiliate 4 to 2.
- It holds the untrackable and the free. CTV and organic stay put - not because they do not matter, but because the case to move them is not yet triangulated. You do not touch what you cannot yet defend.
- The whole move is about 15% of the budget. That is deliberate. It is a testable hypothesis, not a reinvention.
The seductive mistake is to look at an MMM that says "social is under-funded" and swing 40% of the budget overnight. The disciplined version - grow social 4 points, trim search 5, and run a holdout to confirm - captures most of the upside with almost none of the risk. If the thesis is wrong, you are out a marginal shift for a quarter, not a year of misallocated millions.
Self-studyWho creates, who captures - the Lumen map2 min read▶
The reallocation only makes sense if you hold the create-vs-capture split in your head. For Lumen: paid social, display, influencer, and CTV create demand - they reach people who were not yet looking, show up early, and rarely earn the last click. Paid search and direct capture demand - they meet people already searching for "Lumen vitamin C serum", show up last, and hog the closing credit. Email and affiliate sit in the middle. Last-touch systematically over-pays the capture side and under-pays the create side, which is precisely why the proposal shifts money from search toward social and display. You are not saying search is worthless - you are correcting a known, structural bias in how it gets credited.
Defend it to the board: the one-slide story 5 min live
A reallocation is only as good as your ability to defend it in the room. The whole case fits on one slide with four beats: who creates, who captures, where the models agree, and the test that will prove it. Learn to walk a board through those four and you can get almost any disciplined move approved.
LiveWhat a leader actually says in the room3 min▶
The board does not want your model. It wants your judgment, expressed as a story it can trust. Walk the four beats:
- Name who creates and who captures. "Our last-click report crowns search, but search mostly captures demand our upstream channels created." This preempts the "but search has the best ROAS" objection before it lands.
- Show where independent methods agree. "Our MMM and our data-driven attribution both say social and display are under-funded relative to their real contribution." Agreement across methods is your credibility.
- Make it marginal and testable. "We are moving 15%, not betting the plan, and a geo-holdout will confirm the lift is causal within 30 days." This is what turns a "no, too risky" into a "yes, let's see".
- Commit to reading the test. "If the holdout does not show the lift, we roll it back." Nothing earns board trust faster than a leader who has pre-committed to being proven wrong.
Boards approve disciplined, reversible, tested moves and reject confident all-in bets - even when the all-in bet has a prettier dashboard. The leader who says "three methods agree, we'll move 15%, and here's the test that confirms it in a month" almost always gets the yes. The one who says "the MMM says social is under-funded so I want to move 40%" almost always gets a hard round of questions and a no.
Reallocate Lumen's $4M ★ 10 min · everyone
Build the move yourself. Using Lumen's current split, pick two channels to grow, two to trim, and name the single holdout test that de-risks the whole reallocation.
Pick two to grow. Look for under-credited demand creators the MMM and data-driven models favor. Paid social (32 to ~36%) and display (10 to ~13%) are the defensible picks - both create demand and both are under-paid by last-touch.
Pick two to trim. The last-click darling that is being over-funded (paid search, 24 to ~19% - it still captures the demand, just with less over-investment) and the weakest tail (affiliate, 4 to ~2%). Keep it marginal.
Check the totals. Grows and trims should roughly net out - you are reallocating within $4M, not asking for more. A ~15% shift keeps it a testable hypothesis, not a teardown.
Name the one holdout test. A four-week geo-lift on paid social: hold it back in a set of regions, measure the incremental sales gap. That single test tells you whether the growth thesis is causally real before you scale it further.
Write the board sentence ★ 10 min · pen and paper
The capstone deliverable is one sentence a board would approve. Fill in the template with your Lumen numbers, then rewrite it for a reallocation in your own organization.
Start from the template: "We are shifting $X from Y to Z because three independent methods agree - and a geo-test will confirm it within 30 days." Every word earns its place.
Fill in the Lumen version. "We are shifting ~$200K from paid search to paid social and display because our MMM and data-driven attribution both show them under-funded - and a four-week geo-holdout will confirm the lift is causal before we scale."
Pressure-test it. Does it name the source (three methods, not one)? Is the move marginal (a slice, not the farm)? Does it promise causal proof (the geo-test)? If any answer is no, the sentence is not board-ready yet.
Write your own. Take a real reallocation you are considering and force it into the same sentence. If you cannot fill in "three methods agree" and "a test will confirm", you have found the work you still need to do before you move the money.
The sentence is a forcing function. If you cannot write it honestly, you are not ready to reallocate - and that is useful information, not a failure. A leader who walks into a board meeting with that one sentence, backed by a chart of where the methods agree and a test date on the calendar, is operating at a level almost no marketing team reaches. That is the whole point of this course, in a single line.
Put it to work ◐ 30 min total
- Draft a real reallocation. Take your own budget, pick two channels to grow and two to trim at the margin, and write the create-vs-capture logic behind each choice.
- Write your board sentence. Force your move into the "$X from Y to Z, three methods agree, a test confirms in 30 days" template. Notice exactly where it breaks - that gap is your homework.
- Schedule one incrementality test. Pick your biggest or most-contested channel and put a geo-holdout on the calendar. One real experiment teaches more than a quarter of dashboard-watching.
- Optional - go build. If you want to see how these models are actually constructed on Lumen's data, the builder track (b1 onward) builds every one of them in SQL and Python. You now have the leader's map; the builders have the machine.
Three questions before you graduate 🎯 ◐ 90 seconds
1 · Before moving budget, the first discipline is to...
No single method is proof. Acting only where independent methods point the same direction is the closest thing to truth a marketing leader gets - so triangulation comes first.
2 · Your MMM says paid social is badly under-funded. The disciplined move is to...
A reallocation is a hypothesis, not a verdict. Moving a slice keeps the decision reversible and testable; betting the farm on one model turns a possible mistake into a catastrophic one.
3 · What finally makes a reallocation defensible to a board?
Correlational models can be wrong. Committing every move to a causal experiment - and to rolling it back if the test fails - is what turns a budget gamble into a decision a board will sign.
What this session covers
The capstone is a synthesis - it does not map to one course chapter but pulls the whole leader track together into a single defensible decision. Certificates and full video courses stay with their official sources.