MMMAudit
Independent · expert-led · vendor-agnostic

Independent MMM validation

MMM = Marketing Mix Modelling

An honest second opinion on the model that steers your media budget, from specialists in marketing modelling and effectiveness. For brand teams, consultancies, and agencies who need a verdict they can trust.

See packages Book a 20-min call

New here? What is an MMM audit →

Plain English

What is Marketing Mix Modelling?

Marketing Mix Modelling (MMM) estimates how media and other drivers contribute to sales or business outcomes. It is the model many organisations use to decide where next year's budget goes. When that model is wrong, or quietly fragile, misallocation compounds.

The gap

AI and automated MMM promise certainty. Business reality is messier.

One-click MMM, auto-tuned pipelines, and “AI-powered” dashboards make it easy to produce a model. They do not guarantee the model answers the right business question, survives scrutiny, or agrees with your experiments.

Speed is useful. An independent audit checks whether the story on the dashboard is one you can stake budget on.

Defaults are not decisions

Priors, transforms, and optimiser settings often stay at tooling defaults, and quietly drive the ROI ranking.

Pretty charts, weak identification

A polished contribution chart can still rest on a specification that cannot separate media from seasonality or promo.

Automation skips judgement

Business context, data quirks, and experiment evidence still need human review. That is what an audit is for.

Why an independent review

Models that guide large budgets deserve a second opinion

Open-source stacks like Meridian, Robyn, and PyMC-Marketing, and vendor platforms, have made sophisticated MMM accessible. The difference between a well-specified model and one with hidden weaknesses is rarely visible from the dashboard alone.

Stakes are high

An MMM informs where significant budget goes. A small misspecification can compound into serious misallocation over time.

Built-in blind spots

The team that built the model knows it intimately, and may not see where it's weak. Independence brings a fresh pair of eyes.

Assumptions add up

Choices in model assumptions, adstock specification, data cuts, and validation strategy all shape the story. An audit tests whether those choices are defensible.

Who's auditing

Three specialists, one independent check

A credibility-led partnership combining deep technical expertise in Bayesian and causal inference with in-house marketing effectiveness leadership, including two IPA Effectiveness Awards technical judges.

Joe Wilkinson
Analytics & in-housing

Joe Wilkinson

15+ years in marketing analytics across FMCG, automotive, retail, and e-commerce. Led large-scale measurement projects for Fortune 500 brands. Founder of Tyedal and IPA Effectiveness Awards technical judge (2026).

Tyedal LinkedIn IPA judge
Technical lead

Dr Ben Vincent

Bayesian data scientist & causal inference specialist. Director of Inference Works, lead developer of CausalPy, and contributor to PyMC-Marketing. Former faculty applying Bayesian methods to decision-making research.

Inference Works Portfolio LinkedIn GitHub
Founder, Nous Analytics

Nadya Ochirova

Nearly a decade at the intersection of marketing measurement and commercial decisions: first as an econometrician, later accountable for the budget. Built effectiveness teams from scratch at Sage and The Economist Group, and presented investment decisions to CEOs and CFOs. Founded Nous Analytics in 2026 for strategy-led effectiveness programmes. IPA technical judge (2024).

LinkedIn IPA judge
What you get

A verdict you can act on

A written report with a plain-English verdict, issues ranked by severity, and specific fixes. We inspect the parts of the model that actually change decisions.

See how we audit Meridian, Robyn & PyMC →

  • Specification: is the model structured to answer the business question, and is it identified?
  • Assumptions: are they defensible, or quietly driving the results?
  • Data: coverage, leakage, treatment of seasonality, pricing and promotions.
  • Adstock & saturation: are the transforms sensible and the parameters plausible?
  • Validation: does the model hold up out-of-sample and under sensitivity checks?
  • Calibration (optional): does it agree with your geo-lift / incrementality experiments?
How it works

Four steps, three weeks

A focused engagement designed to give you a clear verdict without disrupting your measurement programme.

1

Intro + scope

Free 20-min fit call, then optional fixed-fee scoping (£750, credited to the audit).

2

You share

Model code or spec, the data, and any experiment results, under NDA.

3

We audit

Independent review against the checklist; reproduction if scoped.

4

Report + walkthrough

Written report and a call to talk through findings and fixes.

Standard turnaround three weeks; two-week rush available.

Beyond the audit

Three ways we can help

The audit is the core offer, but not every prospect needs one. Here's what else we provide.

Free MMM guide

A best-practice guide covering specification, model assumptions, validation, and calibration, in plain English. Ask for it on the scoping call.

Topic guides

Deep-dive guides on specific gaps: causal modelling, sensitivity analysis, adstock specification. Browse the guides library.

Remediation & improvement

Every audit comes with a prioritised list of issues and concrete fixes. We can help implement improvements, or define requirements if a rebuild is the right call.

Clear packages

Three audit levels, plus a fixed-fee scope

Pick the shape of your modelling estate. Indicative from-prices; final quote after scoping. Optional add-ons: calibration, reproduction, rush.

I · Core

One model / national

from £3,500 excl. VAT

A single national (or single-market) MMM that guides budget decisions.

  • Full validation checklist
  • Written verdict + ranked fixes
  • Findings walkthrough
Package details
II · Multi

Multiple models / regional

from £6,500 excl. VAT

Several markets, brands, or regional models with shared methods.

  • Cross-model consistency review
  • Prioritised issues across the set
  • Written verdict + walkthrough
Package details
III · Complex

Nested / multi-level

from £12,000 excl. VAT

Hierarchical or nested structures: brand → region → channel, multi-brand suites.

  • Hierarchy & pooling scrutiny
  • Decision-risk stress test
  • Written verdict + walkthrough
Package details

Fixed-fee initial scoping: £750

Structured readiness review under NDA: what we need, risks we can already see, and a firm audit quote. Fully credited if you proceed to an audit within 90 days.

£750
Book intro call
FAQ

The questions everyone asks

Is my data confidential?
Yes. Everything runs under NDA, and we don't retain your data after the engagement closes.
Do you audit models your own team built?
No. We never audit a model any of us built or contributed to. If an audit reveals a model that needs a full rebuild, we'll point you to the right people: us in a separate engagement, or someone else entirely.
What do you need from me?
The model code or specification, the data it was fit on, and any experiment results you want it calibrated against. We confirm exactly what's needed on the scoping call.
What if you find serious problems?
You get a prioritised list of issues ranked by severity, with specific fixes for each. Most issues can be resolved within the current model: adjusting assumptions, refining the specification, or improving the data inputs. If a full rebuild is the right call, we'll recommend specialists.
Do you work with my vendor's model?
We work with the company that commissions the MMM, not directly with the vendor. We audit the vendor's outputs and give you a constructive set of improvement points to take to them.

More pricing & FAQ →

Get started

Book a free 20-minute intro call, or jump to packages and the fixed-fee scope.

Book an intro call See pricing