Your dashboards say every channel is a winner. Your CFO has questions. LiftProof runs the same synthetic-control experiments the $100K/yr tools run — to tell you which dollars are actually working, and which are just along for the ride.
Every channel claims credit. We'll tell you which ones earn it. Switch between common DTC incrementality tests below.
Meta's self-reported ROAS says 7×. Your CFO says prove it. Geo-holdout tests show the actual incremental effect — because iOS 14 broke attribution and Meta still thinks every purchase is theirs.
“Found out 60% of our 'Meta-attributed' revenue would've happened anyway. Saved $1.2M/yr.”— Growth lead, Tier-1 DTC
No data science team. No R scripts. No billable strategist. Upload your sales data, click a few buttons, and get the same answers a $100K tool would give you.
Daily orders or revenue by region. We auto-detect your geo format (ZIP, DMA, state, country), validate everything, and flag data issues before you commit. Shopify, Amazon, anything with a column of numbers — we'll take it.
Shopify · Amazon · Any CSVOur Claude-powered copilot picks treatment and control regions, runs power analysis automatically, and recommends the experiment design with the lowest minimum detectable effect. It's the statistician you couldn't afford.
Power analysis · Market selectionMulti-model synthetic control ensemble spits out lift, confidence intervals, p-values, and iROAS. Export a PDF your CFO won't squint at. Share the link with your team. Ship the decision.
Lift · iROAS · CPIA · CIYou don't pre-pick markets. You don't hand-pick a window. Dump your full US sales table — we figure out the rest, and surface every assumption so you can override it.
| date | geo | orders | revenue |
|---|---|---|---|
| 2025-09-01 | CA | 1,284 | $158,330 |
| 2025-09-01 | TX | 1,102 | $132,410 |
| 2025-09-01 | NY | 968 | $121,022 |
| 2025-09-02 | CA | 1,310 | $162,800 |
| 2025-09-02 | TX | 1,075 | $128,990 |
| … | … | 0 | $0 |
We parse locally first — nothing leaves your browser until you commit. We auto-detect date format, geo granularity (ZIP / DMA / state / country), and the KPI columns. Anything weird gets flagged as an error or warning before you save.
Mixed granularity? We normalize. Inconsistent state codes ("CA" vs "California")? Reconciled. Sparse geos with too few orders to be useful? Excluded with a note so you know what was dropped and why.
Given your KPI volume, market count, and pre-period length, we compute the minimum detectable effect at 80% power. If your test window is too short or your markets too noisy, we say so before you spend a dollar.
Our copilot picks candidate treatment markets that balance volume and pre-period correlation against the donor pool — the splits that fit a synthetic twin tightly. You can accept the recommendation, drag-edit the split, or pick markets manually.
Augmented synthetic control + Fisher permutation runs in the background. You get the lift, 95% CI, p-value, iROAS, and a plain-English caveat in under a minute. Export a CFO-ready PDF or share a read-only link.
It’s the econometrics trick Nobel-adjacent economists use to measure causality when you can’t run a true A/B test. Click through the stages to watch it work on a real Meta Ads experiment.
A trimmed-down version of the actual designer inside LiftProof. Tune the channel, spend, market count, and window — and watch the projected lift, confidence interval, and power update in real time.
The same synthetic-control methods the $100K/yr tools run — with an AI copilot in place of the billable human strategist.
| LiftProof | Paid services | DIY (GeoLift R) | |
|---|---|---|---|
| Price | Free → $149/test | $50–100K/yr | Free (R skills required) |
| Synthetic-control methods | Multi-model ensemble | Proprietary ensemble | Augmented SCM |
| Power analysis | Automated | Automated | Manual in R |
| AI copilot | Claude-powered | Human strategist ($$) | — |
| Setup time | Minutes | Days to weeks | Days of coding |
| Methodology transparency | Open methods | Proprietary | Open source |
One free experiment per month, forever. Need more? Pay $149 per test, or subscribe when you’re running multiple in parallel.
One full experiment, every month. Run it, export the PDF, take it to your CFO.
One-off readout. No subscription, no commitment, full feature set.
For growth teams running tests in parallel across channels and quarters.
Multi-brand workspaces for agencies and holding-co growth teams.
All tiers include the full SCM + DiD ensemble, Fisher permutation testing, and the Claude copilot. No credit card to start.
You get one full experiment per month, free, forever. No credit card. Need more? Pay-as-you-go is $149 per additional test — no commitment. Teams running multiple concurrent tests usually move to Pro ($399/mo, 3 active tests) or Agency ($999/mo, unlimited).
Nope. If you can export a CSV and click a button, you can run a LiftProof experiment. Our Claude-powered copilot handles market selection, power analysis, and model specification for you. You get the verdict in plain English.
MMM correlates spend and sales across time — it's observational. Synthetic control experiments are causal: we build a statistical twin of your treatment group and measure what would've happened without the ads. Different tool, different question, way more defensible.
One CSV with three columns: date (daily or weekly), geo (ZIP, DMA, state, or country), and a KPI (orders, revenue, units, or new customers). Minimum 8 weeks of pre-period across ~20+ geos; 12+ weeks gives tighter confidence intervals. You can dump your full US sales table — we'll detect the geo level and propose a treatment/control split. No attribution data, no pixel data, no cookies, no creepy third-party joins.
We use the same methodology Meta published in its own incrementality papers — Augmented Synthetic Control Methods plus Fisher permutation testing for significance. The PDF export cites the model, the donors, the power, and the p-value. We've had CFOs write us unsolicited thank-you notes.
Only for as long as it takes to run the experiment and serve the result. Geo-aggregated data only — no customer PII ever touches our system. SOC 2 report available for teams that ask.
One free experiment per month. Pay-as-you-go beyond. No credit card to start. No sales call. No data science degree.