Independent marketing measurement science

Know what
worked.
Know what
to do next.

Independent marketing measurement that separates genuine marketing impact from noise, and turns it into better budget decisions.

  • Marketing Mix Modelling
  • Incrementality
  • Experimentation
Talk about your measurement challenge →
Fig. 00 · IllustrativeMarketing signal, with its uncertainty kept in view.

Every platform
wants the credit.

Platform reporting tells you what happened inside the platform. It does not tell you what would have happened without the marketing, which is the question your budget decision actually needs answered.

Platform reported ROAS

4.2×

What the platform attributes

Every platform attributes conversions. Not all of those conversions would have been lost without the campaign.

Incremental marketing impact

1.8×

What the marketing actually drove

Incremental measurement isolates the effect that would disappear if you stopped the marketing.

01More data.
02More attribution.
03Still not always a better answer.

Five questions
measurement
should answer.

Most measurement doesn't answer all of them. Good measurement tries to.

01

What did marketing actually contribute?

Separate the marketing effect from everything that would have happened anyway: seasonal trends, price changes, broader category growth.

02

Which channels created incremental value?

Not which channels were present when sales happened, but which channels caused sales that would not have occurred without them.

03

How confident should we be in the result?

A number without its uncertainty range is not evidence. Knowing how wide the estimate is matters as much as knowing the estimate itself.

04

Where should the next budget go?

Which channels still have headroom? Which are close to saturation? What reallocation improves return without requiring extra budget?

05

What should we test next?

Measurement improves when it is validated. Experiments close the gap between modelled estimates and real-world evidence.

Don't just model it.
Challenge it.

A model can fit the data extremely well and still lead to the wrong marketing decision. Before we make recommendations, we test whether the conclusions survive reasonable changes to the model. This is one of the most important things we do, and one of the least common.

01

Change the assumptions.

Does the channel ROI conclusion stay the same when reasonable modelling assumptions change? If the answer shifts materially, the decision is not yet reliable.

Prior sensitivity
02

Challenge the baseline.

Are we accurately separating underlying demand from the contribution of marketing? An overly flexible baseline can quietly absorb media effects.

Baseline flexibility
03

Separate correlated channels.

Can the model genuinely distinguish which channel drove the outcome, or are correlated spend patterns making that difficult? Identifiability matters.

Confounding & identifiability
04

Test unseen periods.

Does the model still perform on data it was not trained on? Out-of-sample accuracy tells you whether the model has captured real patterns or fitted noise.

Holdout validation
05

Compare reasonable models.

Would another defensible model specification lead to the same budget recommendation? Robustness across specifications builds confidence in the decision.

Specification robustness
06

Know when not to trust the number.

If the data cannot support a reliable estimate for a channel, we say so. False precision is a more serious problem than acknowledged uncertainty.

Same data.
Same strong fit.
Different decision.

Three defensible model specifications applied to the same dataset. All produce an excellent predictive fit. All produce materially different channel ROI estimates, and different budget recommendations.

Model A

Standard baseline

4.2×

Meta / Paid Social · Illustrative ROI

↑ Increase investment

Model B

Tighter priors

2.1×

Meta / Paid Social · Illustrative ROI

→ Hold investment

Model C

Flexible baseline

1.3×

Meta / Paid Social · Illustrative ROI

↓ Test first

Model A fit0.94
Model B fit0.95
Model C fit0.94
R² 0.94
≠ reliable

High predictive fit does not automatically mean reliable attribution. RightMeasure tests whether the business decision, not just the model, is stable before recommending action.

From data
to decision.

↻ Each engagement is a loop. Experiments feed the next model.

01

Data

Media, business drivers and commercial outcomes: understood, cleaned and ready.

02

Diagnose

Data quality, demand patterns and measurement risks, identified before any model runs.

03

Model

Incremental contribution estimated using robust Marketing Mix Modelling and sound prior assumptions.

04

Challenge

Assumptions tested, channels interrogated, alternative specifications compared. The step most providers skip.

05

Decide

Robust findings, not fragile ones, translated into investment scenarios and practical recommendations.

06

Test

Experiments validate the model. Evidence feeds back into the next iteration. Measurement improves over time.

What we do.

Three disciplines. One objective: understand what marketing actually drives, and where the next budget should go.

01

Marketing Mix Modelling

Understand the incremental contribution of marketing across channels: online, offline, brand and performance, beyond what platform attribution can tell you.

  • Channel contribution & ROI
  • Marginal ROI and saturation
  • Response curves
  • Budget scenarios
02

Incrementality

Use controlled experiments to determine whether marketing caused the outcome, and calibrate your MMM against real-world evidence rather than modelled assumptions alone.

  • Geo experiments
  • Holdout design
  • Lift measurement
  • Experiment-informed MMM calibration
03

Media optimisation

Turn validated measurement into a practical plan: where to invest, where to hold, where to test, and how to make the next budget work harder.

  • Budget allocation
  • Scenario planning
  • Saturation analysis
  • Experiment roadmap

Outputs you can
interrogate.

Illustrative examples. Each engagement produces outputs specific to your data, channels and decisions.

Channel contribution

Illustrative

Share of incremental outcome by channel, with the 90% range shown. The width of the estimate is as important as the number itself.

TV & Video
31%
Paid Search
24%
Paid Social
14%
Affiliates
12%
Display
7%
Point estimate90% range

Marginal ROI

Illustrative

Where the next unit of budget sits on the response curve, and how close a channel is to saturation.

CURRENT · mROI 1.3×SATURATIONMEDIA SPEND →RESPONSE →

Budget scenarios

Illustrative

Compare reallocations before committing budget.

Modelled revenue lift+8.4%
Media budgetReallocated
Search 30%Social 14%TV 40%Display 16%

Specification stability: Paid Social ROI

Illustrative

The same channel estimated across 28 model specifications. A single spec reads 0.9×. The distribution tells the full story, which is why we run more than one.

MEDIAN 1.5×0×1×2×3×4×ONE SPEC → 0.9×

Rigorous
underneath.

RightMeasure uses Google's Meridian Bayesian MMM framework together with systematic model validation, sensitivity testing and causal diagnostics. For those who want to look under the surface.

01Google MeridianOpen-source Bayesian MMM framework from Google.
02Bayesian MMMProbabilistic modelling with uncertainty quantification built in.
03Prior sensitivitySystematically varying prior assumptions to test conclusion stability.
04Baseline / knot testingStress-testing how the time-varying baseline absorbs or separates media effects.
05Confounding diagnosticsExamining whether channel spend patterns allow reliable attribution.
06Holdout validationOut-of-sample accuracy testing to check for genuine signal vs. overfitting.
07Incrementality experimentsGeo and holdout experiments to validate and calibrate modelled estimates.
08Budget optimisationMarginal-ROI and saturation-aware scenario planning across channels.

Independent.
Focused.

RightMeasure is an independent marketing measurement practice focused on rigorous MMM, incrementality and evidence-based media decisions.

The work combines marketing science, causal reasoning and modern statistical modelling, with a strong emphasis on understanding when a result is genuinely reliable enough to act on.

We are small and independent by design. You work directly with the people shaping the analysis, challenging the assumptions and translating the evidence into recommendations. No account managers between you and the work.

The differentiation is straightforward: most MMM providers fit a model and report the result. RightMeasure actively challenges the model before recommending that a business acts on it.

Focus
Marketing science & causal measurement
Methods
Bayesian MMM · Geo experiments · Holdouts
Approach
Independent · Evidence-first · Direct
Based
United Kingdom · Working remotely

Better decisions
start with
better evidence.

If you are trying to understand what your marketing actually drove, or whether your existing measurement is reliable enough to act on, let's talk.