How we make this list: we compare pricing pages, product documentation and public user reviews. Ratings come from G2, Capterra or Trustpilot, with the date we checked them.
Key takeaways
- Haus and Measured are the specialist picks for serious geo holdout programmes at enterprise budgets.
- WorkMagic is the entry point for Shopify brands, with app store pricing reported from $29 per month against five figure enterprise contracts elsewhere.
- Northbeam prints real starting prices, $1,500 per month for Starter, but incrementality sits on its higher plans as an add on.
- Every result in this category is an estimate with a range; any vendor promising an exact lift number is overselling the method.
Quick comparison
| Tool | Best for | Starting price | Free plan |
|---|
 Haus | Geo holdout experiments at scale | See website | No |
 Measured | Enterprise incrementality programmes | See website | No |
 WorkMagic | First geo tests for Shopify brands | $29 per month | Yes |
 Northbeam | Attribution with incrementality add ons | $1,500 per month | No |
 LiftLab | Controlled experiment design | See website | No |
 Recast | Bayesian MMM tied to experiments | See website | No |
 INCRMNTAL | Always on modelled measurement | See website | No |
 SegmentStream | Expert led geo experiments | See website | No |
Is your product missing or out of date?
We review this list regularly. Vendors can request a listing, send updated pricing or ask about featured placement.
Starting priceSee website
Free planNo
Haus does one job: design, run and read geo experiments. You pick a channel or campaign, Haus builds matched test and control markets, holds ads back in the control group, and reports the lift with its uncertainty range. Because experiments are the entire product, the design help is deeper than in platforms where testing is a side feature.
The honest limits are the method's limits. A test needs enough conversion volume in the test window to detect a difference, small brands may not have it, and the answer covers the tested channel at the tested budget level. Haus pricing is custom against spend and scope; expect an annual conversation, not a checkout page.
What we like- Built only for incrementality, so experiment design is the whole product, geo testing plus lift modelling and an MMM informed by your own test results, automation handles market selection and readouts
Watch out for- Custom pricing with no published figures, measures only what you test, so untested channels stay unknown, results live apart from your attribution stack and need reconciling by hand
Starting priceSee website
Free planNo
Measured is the enterprise incumbent in this category, built for CPG, retail and large advertisers running ongoing experiment programmes rather than one test. It runs geo holdouts and audience level experiments across major media platforms and folds the results into cross channel reporting that media teams use for budget shifts.
That programme shape is both its strength and its cost. Implementation takes time, contracts are enterprise tiered by company size and data volume, and a brand that needs one answer about one channel will find Haus or WorkMagic a faster road. A brand spending eight figures that needs defensible answers for a finance team is exactly who Measured is built for.
What we like- Long running enterprise platform for geo and audience level experiments, works across 70 plus media integrations, combines testing with attribution style reporting, aimed at multi agency, multi brand advertisers
Watch out for- Enterprise pricing with longer implementation cycles, setup complexity is real, sized for large budgets and can be heavy machinery for a single brand test
Starting price$29 per month
Free planYes
WorkMagic is where smaller direct to consumer brands start. It automates the parts of a geo test that used to need an analyst: choosing matched markets, setting test and control groups and reading the result. Alongside testing it offers attribution and light media mix modelling, giving a small team one measurement home.
Be clear eyed about the pricing source. WorkMagic's own site does not list prices; the $29 per month starting figure and free tier come from its Shopify App Store listing as reported in 2026 comparisons, so confirm the live listing before budgeting. Even at several times that figure, it undercuts enterprise contracts by orders of magnitude.
What we like- Automated market selection, test setup and analysis with little manual work, Shopify native, so store data connects without a project, also tracks halo effects across Amazon and retail, the cheapest credible entry here
Watch out for- The vendor site does not print pricing, so the $29 figure comes from its Shopify App Store listing, depth trails the enterprise tools, Shopify centric, so other stacks fit less well
#4Northbeam
Best for: Attribution with incrementality add ons4.5/5 on
G2 (16 reviews, checked Oct 2026)
Visit northbeam.co
Starting price$1,500 per month
Free planNo
Northbeam is an attribution platform first, and a sensible route into testing for brands already using it. Its Starter plan starts at $1,500 per month for brands spending under $1.5 million a year on ads, per pricing read in October 2026, with Professional from $3,500 per month. Incrementality is an optional addition on Professional, and its MMM product sits on Enterprise.
That structure cuts both ways. A brand that only wants experiments overpays for attribution it may not need. A brand that wants one vendor for daily attribution plus occasional causal checks gets an honest ladder, with more published pricing than most of this list manages.
What we like- Published starting prices, rare in this category, multi touch attribution as the base product, incrementality on Professional and media mix modelling on Enterprise, built for ecommerce revenue data
Watch out for- Starter at $1,500 per month is attribution only, so the testing job needs the higher tiers, annual commitments on paid plans, final price moves with pageviews and spend
Starting priceSee website
Free planNo
LiftLab sells controlled experiments as a discipline: design the test, run it cleanly, and turn the readout into a budget decision. It appears consistently in specialist roundups of this category for its design rigour, and it suits brands whose media mix includes channels where simple geo splits are awkward.
As with every tool here, the constraint is your own data. A clean experiment needs enough spend and conversions in the test cells, and LiftLab cannot manufacture statistical power a small budget does not have. Pricing is quoted per engagement and scope; ask what a single channel test costs before talking about a programme.
What we like- Experiment design is the core craft, with statisticians behind the method, built to test media the geo method struggles with, including some offline channels, results feed media planning directly
Watch out for- Quote only pricing, a smaller name than Haus or Measured, experiment cadence still limited by your media calendar and volume
Starting priceSee website
Free planNo
Recast approaches incrementality from the modelling side. Its Bayesian media mix models estimate each channel's contribution continuously, and the company is known for checking those model estimates against real geo experiments, which is the right scientific order: models for everyday decisions, experiments to keep the models honest.
That makes Recast a good fit for a team moving from last click reporting to a measurement system, rather than a team that needs a single test readout next month. Pricing is custom. As with any MMM, garbage history in means garbage estimates out, so expect an early project around data quality.
What we like- Media mix models that stakeholders can rerun as data refreshes, models validated against incrementality tests rather than taken on faith, built for in house teams rather than black box consulting
Watch out for- An MMM first product, so teams wanting only experiments carry extra weight, quote only pricing, model quality still depends on the history and granularity you feed it
Starting priceSee website
Free planNo
INCRMNTAL takes the other road: instead of running holdout experiments, it models incrementality continuously from patterns in spend and outcome data. Nothing gets paused, no region goes dark, and measurement runs all the time. For national campaigns or app marketing where geo splits break down, that may be the only workable method.
Hold the evidence standard clearly in mind. A well built model is useful, and it is not a controlled experiment. Report its outputs as estimates with ranges, cross check them with an occasional true test when a channel allows one, and be wary of any always on number presented to decimal places. Pricing is quoted.
What we like- Always on approach needs no paused campaigns or holdout regions, works from aggregated data without user level tracking, useful where geo splits are impractical, such as national campaigns
Watch out for- Modelled rather than experimental, so results carry wider uncertainty, quote only pricing, a different evidence standard than a true holdout, which finance teams may probe
Starting priceSee website
Free planNo
SegmentStream pairs its geo experiment tooling with people: marketing scientists help design the test and interpret the readout, and the platform extends into attribution and budget allocation so the result lands in a plan rather than a slide. For a team with budget but no in house measurement scientist, that service layer is the product.
Expect an engagement shaped sales process and custom pricing. The right test of any expert led vendor is the design conversation itself: if they push back on an underpowered test or a contaminated control group, that is the expertise you are paying for. If every test looks runnable, walk away.
What we like- Geo experiments with human marketing science support, connects experiment results to budget allocation decisions, also offers attribution and modelling for a combined view
Watch out for- Quote only pricing, expert led means engagements rather than self serve speed, smaller brand recognition in the US than Haus or Measured
How we chose
We required a real incrementality method: controlled holdout experiments, or a measurement model explicitly built and validated for incremental effect. Pure attribution tools that relabel correlation as incrementality were left out. The list spans the honest range of the market, enterprise programmes, a Shopify entry point, and modelled alternatives for cases where experiments cannot run. Prices come from vendor pages and dated October 2026 pricing reads; most vendors quote, and the entries say so rather than borrowing estimates. We did not claim hands-on testing.
How to read an incrementality result without fooling yourself
Ask for the range, not the point. A lift of 12 percent with a range from 4 to 20 is a different decision from 12 percent with a range from 11 to 13, even though the headline matches. A vendor who shows you only the headline is hiding the part that matters. Ask how wide the range is and what would narrow it: a longer test, more markets, or more spend.
Check the power before the test, not after. A test that could only detect a 40 percent swing will return not proven for most real effects, and that result tells you nothing except that the test was too small. Any credible tool or consultant will run a power calculation up front and tell you the smallest effect the design can see. If yours cannot, the design is the problem.
Remember what a test cannot say. It measures the channel at the spend level tested; doubling the budget rarely doubles the lift, because marginal returns fall. It measures the test window; a result from a sale period will not describe a quiet month. And it measures against the control's world, which includes competitors advertising into your dark regions. Plan a retest cadence for your biggest channels, because an eighteen month old result is history, not evidence.
Frequently asked questions
How is incrementality testing different from attribution?
Attribution divides credit for conversions among the ads that touched the buyer, using correlation. Incrementality testing compares a group that saw ads with a matched group that did not, so the difference is a causal estimate of what the ads actually added.
How much does incrementality testing cost?
Most vendors quote against ad spend and scope, and enterprise programmes run to five or six figures a year. WorkMagic lists from $29 per month on the Shopify App Store, and Northbeam prints attribution plans from $1,500 per month, with testing on higher tiers.
How long does a geo test need to run?
Long enough to gather the conversions the design needs, commonly several weeks per test. The binding constraint is statistical power: low volume channels and short windows produce wide ranges that cannot settle a budget argument.
Can a small brand run incrementality tests?
Yes, with the right channel. A focused geo test on one large channel through WorkMagic or Haus can work at modest budgets. Channels with thin conversion volume cannot be tested reliably at any price, and an honest vendor will tell you that first.
Is a modelled incrementality score as good as a holdout test?
No. Models such as INCRMNTAL and MMM estimate effect from patterns and are useful where experiments cannot run, but a controlled holdout is the stronger evidence. Best practice uses models for ongoing decisions and real tests to check them.
Sources we checked (0)
Is your product missing or out of date?
We review this list regularly. Vendors can request a listing, send updated pricing or ask about featured placement.
Shihab Mia · Founder, Rankite
Shihab has 5+ years in B2B and SaaS SEO. He writes Rankite's software buying guides from pricing pages, documentation and public user reviews. More about Shihab.