Replace broken multi-touch attribution in three steps. Fence it to tactical reads on the trackable slice, stand up incrementality testing to make budget decisions, and add marketing mix modelling for the whole-spend picture. Incrementality and mix modelling need no user-level identity, so privacy cannot degrade them. It is a measurement decision, not a new tracking build.
The obstacle article showed why multi-touch attribution broke: it needed identity coverage above 90%, and below 60% resolution credit defaults to last click (Improvado, 2026). This is how to stop steering budget with that broken instrument, in three steps you can start this quarter.
Step 1: Fence attribution to tactics
Start by demoting the number, not deleting it. Attribution still has one honest job. On the slice of traffic you can still track, it shows relative day-to-day movement between campaigns. That is useful for optimising creative, bids, and audiences inside a channel.
So keep it there and nowhere else. Write a one-line label on every attribution report: “directional signal on trackable traffic, not a basis for budget allocation”. It sounds trivial. It is the whole shift. The damage was never that teams looked at attribution. It was that they let it split the budget.
Do one subtraction at the same time. Remove attribution credit from the slide that decides spend. As long as last-click revenue sits on the budget deck, teams will fund the channels closest to the sale, whatever you say in the meeting.
Owner: the head of marketing or marketing operations, who owns the reporting. What to measure at this step: nothing new yet. This step frees a decision from a broken input. You are clearing the seat before you fill it.
Step 2: Stand up incrementality testing
Now install the engine that makes budget calls: the holdout test. Hold a group back from seeing a campaign, run the campaign to everyone else, then compare the two groups. The difference is the lift the campaign actually caused. Not the sales it happened to sit near. The sales it created.
Start with your largest spend line, because that is where a wrong attribution read costs the most. Pick one channel attribution loves, the branded search or the retargeting that always looks efficient, and test whether it drives incremental sales or just harvests buyers who had already decided. The answer reprices your whole budget more than any dashboard.
Run tests in sequence, not all at once. One clean holdout on one big channel beats ten half-designed experiments. Geographic holdouts work well when you cannot split by user: hold the campaign out of matched regions and compare. This is exactly why incrementality survives privacy. It measures groups, not tracked individuals, so it needs none of the identity coverage that collapsed.
Owner: marketing operations with an analyst who can design a clean test and read it honestly. What to measure: incremental lift per channel, and cost per incremental conversion. Those two numbers do what attribution promised and never delivered. They tell you what your spend caused.
Step 3: Add marketing mix modelling for the whole picture
Incrementality answers “did this specific spend work”. Marketing mix modelling answers the bigger question: how should the total budget be split across every channel, including the ones attribution never saw.
Mix modelling reads aggregate spend against outcomes over time. It sees offline, brand, and channels with no click to track, which is why it also closes the cross-channel and offline measurement gap. It needs no user-level identity, because it works at the level of spend and result, not the individual path. Privacy leaves it untouched.
You do not need a data science department to begin. Start with a simple view of spend versus outcome by channel over time, then add statistical rigour as the practice matures. The discipline of reading aggregate contribution matters more than the sophistication of the first model. Validate the model against your Step 2 holdouts: where incrementality and mix modelling agree, you have a number you can bank.
Here is the rule that governs all three steps. Never optimise against attribution output you cannot validate. If a channel’s case rests only on attribution credit, with no holdout and no mix-model contribution behind it, treat that case as unproven and do not let it move the budget.
Owner: a marketing analytics lead, supported by finance, who owns the model and reviews it each planning cycle. What to measure: modelled contribution and marginal return by channel, reconciled against incrementality results.
How to train your team to hold the fix
A measurement change fails if the team’s instincts do not change with it. Three moves make it stick.
Retrain the meeting, not just the model. The reflex to ask “what does attribution say” is deep. Replace it with a better question the team asks by habit: “what caused this, and can we validate it”. Put that question on the wall of the planning room if you have to.
Protect the transition through one uncomfortable quarter. The first incrementality result will probably show a favourite channel is less incremental than its attribution credit claimed. Someone will want to bury it and go back to the reassuring number. Hold the line. That result is the payoff, not the problem.
Build the honesty into the incentive. If a channel owner’s target still rests on attributed revenue, they will defend the broken number. Move targets onto incremental contribution. People optimise for what they are measured on, so measure them on cause.
Where Morphy helps
This is a strategic bet, not a weekend fix, but the first proof point comes fast. We run it as a focused four-to-eight-week engagement.
Weeks one and two, we audit how much of your budget is currently allocated on last-click or attribution credit you cannot validate, and quantify the exposure. Weeks three to five, we design and run your first incrementality holdout on your largest spend line, and read the result against the attribution number it replaces. Weeks six to eight, we stand up a starter marketing mix model from your existing spend and outcome data, and hand your team the decision rule for using all three instruments together.
The defined metric: your next budget cycle is allocated on validated incremental contribution, with at least one major channel repriced against a holdout you can trust rather than attribution credit you cannot.
Go deeper on customer data maximization
Three ways forward. Pick the one that fits where you are.
- Get the playbook. Practical notes on turning the customer data you already own into revenue, straight to your inbox. Join the newsletter at the foot of this page.
- Take the assessment. Score your customer data maximization in four minutes and see your top revenue blockers. Start the assessment →
- Book a meeting. Bring your data problem. Leave with a prioritised fix, not a platform pitch. Book a call →
The playbook companion to Why has multi-touch attribution broken, and what replaces it?. Post 7 of 25 in the Customer Data Maximization series.
Frequently asked questions
How long before incrementality testing pays back?
Your first holdout test runs in a few weeks and answers one clear question: did this channel cause sales or sit near them. You do not need to test everything at once. Start with your largest spend line, because that is where a wrong attribution read costs the most money.
Do you need to track individual users for this fix?
No, and that is the point. Incrementality testing uses group holdouts and marketing mix modelling reads aggregate spend over time. Neither needs the user-level identity that collapsed under multi-touch attribution (Improvado, 2026). You measure cause and contribution without rebuilding tracking privacy took away.
Can a mid-market team run mix modelling without a data science department?
Yes, at a starter level. You do not need a full econometric model on day one. Begin with a simple spend-versus-outcome view by channel over time, then add rigour as you go. The discipline of reading aggregate contribution matters more than the sophistication of the first model.
What do you tell the board while you transition?
Tell them the truth. Attribution credit was degrading under privacy, so budget now moves on validated cause, not trackable proximity. Show the first incrementality result next to the old attribution number. The gap between them is the case for the change, made in the board's own language of spend and return.