Causal marketing intelligence
Stop paying customers who were already going to buy.
Counterfactual Labs estimates — with causal inference, not correlation — how much of your incentive budget goes to people who would have bought anyway. Then we help you spend only on the genuinely persuadable, and prove the lift against a holdout.
Estimate first, proof second. We label every number for what it is.
Your best-converting campaign might be your most wasteful.
Discounts, 0% APR, rebates — a real share goes to “sure-things”: customers who would have purchased with no incentive at all. Standard analytics can't see this, because they measure who converted, not who was moved. The money looks well spent. It isn't.
- Attribution credits the incentive for sales that would have happened anyway.
- You can't separate persuasion from coincidence without a counterfactual.
- The waste compounds every campaign cycle — invisibly.
Four kinds of customer. You should only pay for one.
Every customer falls into a quadrant by how much marketing actually moves them — their causal uplift (CATE).
Persuadable
High +uplift
Marketing causes the purchase. Spend here.
Sure-thing
~0 uplift, high baseline
They'd have bought anyway. This is your wasted incentive.
Lost-cause
~0 uplift, low baseline
Won't convert either way. Don't bother.
Sleeping-dog
Negative uplift
Contact actively hurts. Exclude them.
From a spreadsheet to a defensible number.
- 01
Connect your data
Bring campaign, incentive and outcome data. Guided onboarding maps it into a causal model — a human confirms every load-bearing definition.
- 02
Estimate the waste
The causal engine estimates uplift per customer and totals the incentive spent on sure-things.
Estimate - 03
Target the persuadable
Get a ranked, exportable audience of the customers marketing actually moves.
- 04
Measure the lift
Hold out a randomized control, run your next cycle, and we measure real lift versus that holdout.
Proof
The number has to be trustworthy. So the science comes first.
Counterfactual Labs is built on a causal-inference engine — uplift meta-learners, propensity modelling, holdout randomization and validity checks (overlap, placebo, calibration) — not dashboard heuristics. The assistant onboards and explains; it never decides the number. The engine owns every causal figure.
- Uplift / CATE per customer
- Holdout-measured lift, not attribution
- Validity & sensitivity checks on every estimate
Built for big-incentive verticals. Starting with automotive.
Automotive runs enormous incentive budgets — 0% APR, rebates, loyalty offers — with a real fraction spent on buyers who were always going to buy. It’s where proving wasted spend pays for itself fastest. We’re partnering with a leading automotive brand as our anchor client.
Find out what you're wasting.
Request a demo and we'll walk you through the diagnostic on your own campaign structure.