Last click ROAS doesn't tell you where margin is hiding.

Last-click ROAS lies, and it lies most loudly about your most profitable channel.

Most operators are running with a model that rewards whichever channel was closest to the purchase. Meta usually wins. Google sometimes. Affiliate networks frequently do not, even when they're the only reason the customer existed in the first place.

The lie compounds. You spend more on the channel that looks efficient. You starve the channel that's actually delivering high-margin, high-LTV customers. The math gets worse every quarter, and last-click ROAS keeps telling you it's getting better.

What we actually do

We rebuild attribution around three numbers that matter:

• Contribution margin per acquisition. Not revenue. Not first-order value. The actual margin that lands on the P&L after COGS and channel costs.
• 12-month LTV by acquisition channel. Some channels bring repeat customers. Some bring one-and-dones. Treating them the same is how you over-invest in the wrong place.
• Segment-level breakdown. Same channel, different cohorts. The Meta lookalike against your top-decile customers behaves nothing like the broad-targeting Meta campaign, and rolling them up hides which one is paying for itself.

The unglamorous part

Most of this work is plumbing. Stitching warehouse data with ad-platform data with order data. Building a model that doesn't break when one channel's API changes. Running back-tests against historical cohorts to verify the model isn't lying in a different way than the old one.

None of it is novel. All of it is missing from most operators' setups.

What you get

A reallocation recommendation that's defensible — backed by margin math, not pixel data. A quarterly model refresh as new cohort data lands. The ability to say no to a channel that looks profitable on the dashboard but isn't on the P&L.