The problem
Marketing says "it works". Finance asks "so what?". This service gets both to the same answer, with the uncertainty written next to it.
Incrementality agency for DTC brands that need a budget decision, not a lab
We answer one question for the budget decisions that matter: what would have happened without this spend? We start with the decision, then pick the lightest method that can inform it. A test is one of those methods. It is not the product.
Analytics & measurement
Problems we fix
How we fix it
A result, how sure we are, and what we recommend doing with the budget.
What does an incrementality agency for DTC brands actually do?
Incrementality is the effect that would not exist without the activity. Attribution is different: a rule for splitting credit, not proof of cause. The gap between the two is where budgets leak. More on that in incrementality vs attribution.
Measuring the cause is hard. Economists who ran large advertising experiments found that ad effects are small next to normal swings in sales, so clean measurement needs very large samples1. A mid-size brand has less data than that. So we don't pretend. We match the method to what your data can carry.
What you get
The decision first
Which budget decision is on the table, and what result would change it.
Reconciled reports
Platform figures set against your sales, as the baseline for any method. If the tracking itself is broken, fixing it is separate work.
Platform lift studies where available
Google, for example, offers conversion lift studies based on users or on geography2.
Holdout tests or geo tests
where your volume and regions allow.
The output
a result, how sure we are, and what we recommend doing with the budget.
How it works
- 1
Problem
Name the decision and the threshold. If both plausible results lead to the same budget, we don't test.
- 2
Plan
Check data history, conversion volume, how groups or regions could be split, the expected effect and what a pause would cost you.
- 3
Execution
Run the lightest method that can inform the decision: reconciliation, lift study, holdout, geo test or a model.
- 4
Decision
Read the range of the result against your threshold. Increase, change, stop or keep checking.
Case studies
Rebuilding a growth engine for a recovery-footwear brand
A category-defining recovery-footwear brand had a growing market, a best-in-class product, and a shrinking customer base.
A recovery-footwear brand
- +27.8%
- net-new customers, full year over prior year
- −33%
- cost per new customer, exit quarter vs starting quarter
Weather-based marketing: ads only when the sun is out, conversion rate +76.17% in the UK
For a premium garden and patio brand, ads in DV360 were switched on and off by live weather data and tested against an always-on control in two markets.
Global garden and patio equipment manufacturer
- 76.17%
- increase in conversion rate (United Kingdom)
- 39.32%
- decrease in cost per acquisition (CPA) (United Kingdom)
How we measure our own work
We don't publish client results we can't verify. Instead, we say what we know, what we assume and where the risk sits. We aim to agree how a result will be read before it arrives. "Not enough data to tell yet" is a result too.
Our thinking is public: incrementality for mid-size ecommerce brands, when not to run an incrementality test and MMM vs MTA vs incrementality.
“Their deep knowledge of advanced marketing analytics, especially in the context of tech and SaaS companies, stood out.”
Who it's for, and who it's not for
How you can buy it
On its own
Measurement and incrementality is a separate service. You don't need to buy paid media from us.
With paid media
Alongside our paid media service, so the same team that spends the money checks whether it worked.
Start with a free audit
Our free measurement and tracking audit shows whether your numbers can carry the question at all.
No prices here. Fee terms are agreed before we start.
What we don't promise
Tools
GeoLiftCausalImpactMeridian GeoXConversion lift studies
FAQ
Do we need an incrementality test?
Not always. A test earns its cost when the answer could change a budget and your volume can detect the effect. Otherwise we use what you already have and say what it can't settle.
How many conversions do we need to run a test?
It depends on the effect size and your volume. Published minimums are mostly rules of thumb: one agency guide to Meta lift studies sets its own weekly conversion minimum per test cell6. A research paper on Facebook lift studies notes that standard A/B test formulas don't apply directly and works out power and minimum sample size for them7. We check your minimum detectable effect before promising anything. See also conversion lift study requirements.
What if the test is inconclusive?
Then the data couldn't tell. That is not the same as "the ads don't work". A trade-media overview of lift studies quotes Google saying that a "no lift" reading does not prove the ads failed8. An inconclusive test result is more likely when volume is small. We read the range against your decision threshold.
Can't we just use Meta's or Google's lift studies?
You can, and we do. Trade press has warned that platforms grade their own tests, and that "new to the platform" is not "new to your brand"9. We check the result against your sales.
Do we need marketing-mix modelling (MMM)?
Probably not as a first step. A paper from Meta's team behind the open-source Robyn says a typical dataset holds at least two years of weekly data, though direct-to-consumer brands may manage with a shorter run of daily data10. The same paper treats calibration with experiments as the key safeguard10. An analyst view of Google's Meridian stresses the data, skills and organisational trust it needs11. More on marketing-mix modelling.
Isn't a post-purchase survey enough?
Which tools do you use?
Whatever fits the problem. Our view of the options is in incrementality testing tools and GeoLift vs Meridian GeoX vs CausalImpact. If we ever have a commercial relationship with a tool, we will disclose it.
Who runs this for you?
Growth Strategist, Paid Social
Paweł Skibiński
Eleven years in marketing. About $100M in ad budgets. I test what would have happened if you had switched the channel off.
Marketing Strategist
Stan Heropolitanski
Over ten years in marketing. Budgets up to about $18M a year. I still care more about contribution margin than about any channel's dashboard.
Sources
- Lewis, Rao & Reiley — Measuring the Effects of Advertising: The Digital Frontier (NBER chapter, 2015)Primary source
- About Conversion Lift (Google Ads Help)Primary source
- GeoLift (GitHub, facebookincubator)Primary source
- CausalImpact (dokumentacja pakietu R)Primary source
- Meridian GeoX (Google for Developers)Primary source
- Hunch: conversion lift study on MetaVendor
- Liu, Bettaney, Chamberlain: Designing Experiments to Measure Incrementality on FacebookPrimary source
- PPC Land: Conversion lift (przegląd)Blog or media
- AdExchanger: Beware the platform incrementality trapBlog or media
- Packaging Up Media Mix Modeling: An Introduction to Robyn's Open-Source Approach (arXiv 2403.14674)Primary source
- Is Google's Meridian The Right Open-Source MMM Solution For You? (Forrester)Industry publication
- The post-purchase survey as a marketing attribution tool (Fairing)Vendor
- JumpFly: Realistic Approaches for Small & Mid-Sized AdvertisersVendor