Most B2B teams between $50M and $500M ARR will tell you their measurement program is “modern.” Pressed for what that means, the answer is usually a mix of platform-reported ROAS, a partial-funnel attribution model, and one incrementality holdout someone ran on LinkedIn last quarter. None of those tell you whether your media budget is creating demand or watching it happen. Geo-lift testing does — and it has quietly become the gold-standard incrementality method for B2B because it survives everything that broke the old playbook.
TL;DR
- Channel-level holdouts test whether a single channel is incremental. Geo-lift tests whether a marketing market is incremental. Both are useful. Only the second answers budget allocation questions.
- Geo-lift works without person-level identifiers. It splits markets into treatment and synthetic-control regions, applies time-series modeling, and produces a credible counterfactual — making it the rare measurement method that the iOS privacy stack and the cookieless web have not damaged.
- The minimum credible cadence is quarterly, on the largest two or three media investments. Anything less frequent becomes a slide for the board rather than a budget input.
- The diagnostic that a team is doing geo-lift as discipline rather than as performance: the results actually change next quarter’s plan. If the answer never moves spend, the test was decoration.
What Geo-Lift Actually Does That Holdouts Don’t
A channel holdout asks: if I stop spending on LinkedIn in Audience A for six weeks, how do conversions change versus a comparable Audience B? It’s an honest test, and a useful one. The limit is that it only tells you about that channel, in that audience, in that period — and only on conversions the channel can claim.
A geo-lift test asks something larger: if I run a full media program in Markets 1, 4, 7 and 9 at a higher intensity than in Markets 2, 3, 5, 6, 8 and 10, what happens to all demand signals in the treated markets? Pipeline, website traffic, branded search, hand-raises, demo requests, closed-won revenue. The lift is measured across the whole demand system, not against one channel’s attribution credit.
The two methods answer different questions. Channel holdouts inform tactical optimization. Geo-lift informs strategic allocation. The reason most B2B teams default to holdouts is that holdouts are cheaper and faster to run, not that they’re more useful at the budget table.
Why It Took This Long to Become Practical
Two things changed in 2025–2026 that pushed geo-lift from “MMM-adjacent specialist tool” to “operating discipline.”
The first is synthetic control modeling. Older geo-experiments required you to find naturally matched markets — Indianapolis vs. Columbus, Calgary vs. Edmonton — and trust that they would have moved identically without the campaign. Synthetic control builds a composite counterfactual from a weighted basket of donor markets, dramatically reducing the false-positive rate and making it possible to run experiments on smaller media budgets than the old methodology required.
The second is the privacy collapse. Person-level attribution has become unreliable enough — between iOS App Tracking Transparency, cookie deprecation, the LinkedIn measurement ceiling, and walled-garden modeling — that the rest of the measurement stack now has to carry weight it wasn’t designed to carry. Geo-lift doesn’t need to identify a person to work. It needs aggregate market signals, which still exist.
What an Operating Geo-Lift Discipline Looks Like
The version of this that actually informs decisions has four features, none of them exotic.
A quarterly cadence on the largest two or three media investments. Most B2B teams have one or two channels that absorb 60%+ of working media — typically LinkedIn paid, content syndication, or trade media. Those are the channels where attribution overstatement is most expensive. Run a geo-lift on each at least once a year, ideally twice.
A minimum cell size that makes the statistics work. Below roughly $30K of media spend per market per test, the noise overwhelms the signal. For most $50–500M ARR B2B teams in North America, this means working with state-level or DMA-level cells rather than metros — the cells need enough demand volume that a 10–15% lift is detectable against the underlying baseline variance.
A finance reviewer in the room when results are read out. Geo-lift exposes channels that look great in the platform dashboard and produce no measurable demand lift. The internal politics of acting on that result are nontrivial. Having the CFO or VP Finance hear the result first-hand turns an analytical finding into a budget input.
A rolling library of past results. One geo-lift in isolation is a data point. Six geo-lifts across two years is a measurement program. The library is what lets you separate channel decay from a bad test, and what makes the conversation with the board boring instead of dramatic.
Where It Exposes the Stack
The channels where geo-lift most often disagrees with platform-reported performance are predictable. Branded paid search frequently shows zero incremental lift — the conversions were going to happen via organic anyway. Retargeting against site visitors shows lift well below the attribution claim, usually 20–40% of reported. Account-based display, especially against accounts already in pipeline, often shows no measurable market-level lift at all.
These results don’t mean those channels are worthless. Branded search still defends the front door. Retargeting still serves a real conversion role. ABM display still provides air cover. What they mean is that platform-reported ROAS is not a credible input to allocation decisions, and incremental media dollars should not flow toward whichever channel claims the highest attribution number.
The channels that hold up under geo-lift tend to be the ones attribution underweights: brand campaigns on LinkedIn and CTV, mid-funnel content distribution, trade publications, podcast sponsorships. The conversion is offline or delayed, so the attribution dashboard misses it. The market-level lift study finds it.
The Cost and Where It Lives in the Org
A respectable geo-lift program runs $40–120K per year in vendor cost — platforms like Haus, LiftLab, Sellforte, Recast and Measured all have credible offerings — plus internal time from a measurement lead to coordinate test design, market split, and reading the results back to the team.
The internal owner question matters more than the vendor choice. Teams that run geo-lift as a marketing-ops side project produce slides. Teams that run it inside a marketing science function, with a charter that lets the science team challenge planning decisions, produce budget moves. The difference shows up within a year — usually in a meaningfully shifted channel mix and a board narrative that has actually changed.
The Bottom Line
The honest test of whether a measurement program is doing the job is whether it changes the allocation. Geo-lift, run quarterly on the biggest channels, with a synthetic-control method, in a room where finance is present, will change it. Channel holdouts alone will not. Attribution alone definitely will not. If the budget conversation looks identical in 2027 to how it looked in 2024, the measurement program isn’t producing decisions — it’s producing decoration. Geo-lift is the cheapest operationally credible way to fix that.
Additional Resources
From the Zaitz Marketing Knowledge Library:
- What is Incrementality in Marketing? — The methodological foundation under geo-lift
- Why Your Attribution Model Is Lying to You — Why platform-reported performance is the wrong input to allocation
- Marketing Mix Modeling Explained — How MMM and geo-lift fit together in a unified measurement layer
External Reading:
- Haus on geo-lift testing methodology — Practitioner-grade explainers from one of the lighter-weight geo-lift vendors
- Sellforte capability comparison — How synthetic control changes the experimental design
Want a second read on your measurement setup?
Start with a Growth Architecture Review. We will map your channel mix, audit your attribution, and show you where the real leverage is.
Book a Conversation →