- Open-source MMM
- Incrementality tests
- You own the models
AI, data and automation · MMM, incrementality and attribution
A marketing mix modelling agency that shows which euro worked
We are a marketing mix modelling agency in Palma de Mallorca, and we answer what no platform dashboard can: how much of your revenue exists because of your marketing. Attribution, MMM and incrementality tests become one source of truth that you own.
Personal reply, No obligation, English, Español
What it is
Why attribution alone is no longer enough
Attribution alone no longer answers the budget question: it credits the clicks it can see, but cannot show which sales would have happened anyway. Marketing mix modelling (MMM, spelt 'modeling' in the US) and incrementality tests add what is missing: each channel's real effect.
Every platform counts conversions with its own windows and models, so what Google, Meta and the rest report easily adds up to more than you sold. We never simply add those figures up; we check them against your actual sales.
From Palma de Mallorca we work with companies across Spain and abroad, on the data foundation built under analytics, first-party data and server-side tracking. Our marketing mix modelling guide explains the concepts.
In depth
Attribution, MMM and incrementality: which question each answers
Three methods answer three questions, and none is enough alone: attribution describes the journey, MMM estimates each channel's contribution, and an incrementality test demonstrates cause and effect.
So we triangulate, as the IAB's 'Modernizing MMM' guide (December 2025) recommends. In our method, tests calibrate the model, the model gives the overall picture and attribution steers daily optimisation in Google Ads or paid social.
- Attribution. Answers: which ads preceded the purchase. Needs: user-level tracking and consent. Pace: near-daily. Limit: blind to what cannot be tracked.
- MMM. Answers: what each channel contributes to sales. Needs: aggregated spend and sales data, no personal data. Pace: monthly or quarterly. Limit: demands history and variation.
- Incrementality. Answers: whether a campaign caused sales that would not otherwise exist. Needs: a control group or region. Pace: weeks per test. Limit: one question per test.
In depth
Why Google and Meta both claim the same sale
Because each platform sees only its part of the journey and counts it by its own rules. If someone sees an Instagram ad and clicks a search ad days later, both can record the full conversion. Each figure is consistent on its own; adding them is the mistake.
Signal loss compounds it. It comes not from Chrome removing third-party cookies (it kept them) but from consent rules and Safari and Firefox restrictions. Since March 2024, European Economic Area traffic without consent signals cannot feed Google audiences or remarketing.
The foundations (Consent Mode v2, enhanced conversions, the Google tag gateway and a Meta Conversions API set-up beyond the one-click option) belong to analytics and tracking; here we model that data.
In depth
Open-source MMM and incrementality tests
Mid-sized companies can now use methods once reserved for the largest advertisers. We work with open tools such as Google's Meridian and Meta's Robyn, plus geo tests and holdout groups: no proprietary licences, no model you cannot audit.
Google announced on 29 January 2025 that Meridian was open to everyone, and says its incrementality tests have accepted lower budgets since November 2025; Meta offers an incremental attribution setting. These are vendor claims, so we treat the tools as instruments, not referees.
A geo test compares regions where a campaign runs with comparable regions where it is paused; a holdout keeps a control group without ads. Results calibrate the model and feed budget scenarios, each with its range of uncertainty.
In depth
The data you need, and when not to run an MMM
An MMM needs a long, unbroken, usually weekly history of channel spend and sales, with enough variation to tell effects apart, plus whatever else moves sales: season, prices, promotions or public holidays. That record is essential for seasonal businesses, common in the Balearics.
When an MMM does not fit, we say so first and propose a shorter route: well-configured attribution, simple tests and tidy data for modelling later. It does not fit if:
- Your advertising history is short or has gaps.
- Almost all your budget sits in one or two channels.
- The budget is so small that its effect is lost in the noise.
In depth
One source of truth, and it is yours
The outcome is a triangulated dashboard your leadership can read unaided: actual sales, contribution by channel, test results and attribution, each figure with its source and margin of error.
AI visibility counts as a brand indicator: since 31 August 2026, Search Console has shown every site its impressions in AI Overviews and AI Mode, without clicks or CTR, and Bing Webmaster Tools shows Copilot citations. We read both alongside our GEO work.
Models, code, dashboards and data are yours from day one: code in your repository, data in your accounts, every assumption written down, so your team or another provider can carry on. Decisions then feed your digital strategy and conversion rate optimisation.
What's included
What our marketing measurement includes
Six deliverables, scaled to your data, channels and decisions.
Measurement plan and KPI tree
Which decisions need making, which method answers each and which indicators everyone shares.
Data review
Spend, sales and external factors: what exists, what is missing and whether it supports a model.
Test calendar
Geo and holdout tests, ordered by the decision they unlock.
Marketing mix model and readouts
An MMM in Meridian or Robyn where data allows, with readouts of what it can and cannot show.
Budget scenarios
Simulated splits across channels, given as ranges rather than single figures.
Assumptions and handover
Code, dashboards and documented assumptions in your accounts, ready for your team.
How we work
How we approach measurement, step by step
A cycle that starts from a business decision, not from a tool.
Phase 01 / 04
Business questions
Which investment decisions lie ahead and what you must know to make them.
Phase 02 / 04
Data and design
We review history, consent and sales reconciliation, then agree the plan, KPI tree and test calendar.
Phase 03 / 04
Tests and model
We run the experiments and fit the model, with every assumption written down.
Phase 04 / 04
Readout and review
Results with their uncertainty, budget options and a regular model refresh.
The SERMAN approach
Why trust us with your measurement
Method, not promises
A transparent process that states only what the data supports.
Three methods that check each other
Attribution, model and experiments are compared; no figure stands without a second source.
No black box
Open tools, code in your repository and written assumptions: you can audit everything.
What it is not
What this service is not
- No guarantee of ROAS or sales: better measurement improves decisions; it secures no outcome.
- Not a vanity dashboard: if an indicator changes no decision, it stays out.
- No invented benchmarks: we compare you with your own history and tests.
What we measure
The indicators that guide your budget
Each indicator comes with its range of uncertainty and its method.
Incremental sales
Sales or leads that would not exist without the campaign, per tests and model.
Incremental return by channel
Incremental revenue per euro spent, set against the ROAS each platform reports.
Channel contribution
Sales the model assigns to each channel, and sales that would arrive anyway.
Marginal return
What the next euro adds in each channel, and where saturation begins.
AI visibility
Impressions in AI Overviews and AI Mode, plus Copilot citations, as a brand indicator.
Tools & platforms
Tools we work with
- Google Meridian
- Robyn (Meta)
- Google Analytics 4
- BigQuery
- Looker Studio
- Google Ads
- Meta Ads Manager
- Google Search Console
- Bing Webmaster Tools
- Python and R
Who it's for
A particularly good fit if…
- You invest in several channels and every platform claims the same sales.
- You have to justify the marketing budget to leadership or finance.
- Your business is seasonal and you want to separate the calendar from marketing's effect.
- You are about to raise, cut or reallocate spend and want sound grounds.
Compliance
Privacy and rules we respect
Applied by default; not a substitute for legal advice.
Consent and cookies
User-level measurement follows the AEPD cookie guide and Consent Mode v2, with no techniques to identify people who decline tracking.
GDPR and aggregated data
MMM and geo tests use aggregated data, not personal data. CRM data comes with data minimisation and a data processing agreement.
AI with human review
AI helps prepare data and code; people review models and conclusions. Your data trains no models unless agreed by contract.
Industries
Where we apply it
Tech, SaaS & Startups
Go-to-market strategy, AI answer visibility, LinkedIn, lifecycle automation and pipeline measurement for startups, SaaS and tech companies.
FAQ
Frequently asked questions
What is marketing mix modelling, and does my business need it?
A statistical model that estimates what each channel contributes to your sales from aggregated data on spend, sales and external factors. It pays off with steady spend across several channels and enough history; otherwise, sound attribution and incrementality tests are enough.
What is the difference between attribution and incrementality?
Attribution shares the credit for a sale among the ads the buyer interacted with. Incrementality tests whether that sale would have happened anyway, using a control group or region. One describes the journey; the other proves cause and effect.
How much data do you need for an MMM?
A long, unbroken history of channel spend and sales, usually weekly, with real changes in spend. The more channels you separate, the more history you need; our data review shows whether your case supports a reliable model.
What is Google's Meridian?
Google's open-source marketing mix modelling framework, opened to everyone on 29 January 2025. It lets you build an auditable MMM of your own; we use it alongside Robyn, Meta's open alternative.
How do you know whether a campaign generated sales that would not have happened anyway?
With an experiment: a control group or comparable regions go without the campaign, and the sales difference against those who see it is measured. That difference, with its margin of error, is the incremental effect.
Why do Google Ads and Meta both claim the same sale?
Because each platform records conversions with its own windows and models, and sees only its own ads. One purchase can appear in full in both dashboards. That is why, as a marketing mix modelling agency, we check those figures against your real sales rather than adding them up.
Related services
Pairs well with
Analytics, First-Party Data & Server-Side Tracking
GA4, first-party data, server-side tracking and Consent Mode v2 set up properly: reliable, lawful data you own, the base for campaigns and reports.
Google Ads Management
Google Ads for the AI Max era: Search, Performance Max and Demand Gen on reliable tracking, in an account you own.
Paid Social Advertising
Meta, TikTok, LinkedIn and Pinterest campaigns where creative does the targeting, built on clean signals and ad accounts you own.
Digital Strategy & Fractional CMO
Digital marketing strategy and fractional CMO leadership: a diagnosis, a channel and budget plan, and a growth system you can measure.
Conversion Rate Optimisation (CRO)
Research, A/B testing and UX fixes that help more visitors become customers, with no dark patterns and accessibility as a conversion lever.
Insights
Go deeper
1 Oct 2026 · 16 min read · Strategy, brand & social
The future of digital marketing: 10 shifts defining 2026-2030
Ten shifts that will define digital marketing between 2026 and 2030: what is already live in Spain, what comes next and what to do in the next 90 days.
1 Oct 2026 · 12 min read · Paid media & retail media
ChatGPT Ads in Spain: a practical guide to advertising in AI assistants (2026)
ChatGPT has shown ads in Spain since 24 August 2026. Who sees them, how access works, what your website needs and which AI assistants carry no advertising at all.
1 Oct 2026 · 11 min read · AI, data & automation
Marketing mix modeling explained: incrementality and measurement in the AI era
What marketing mix modeling is, what incrementality adds and how to combine both with attribution to set budgets when consent and AI leave gaps in the data.
Contact
Start with a review of your data
We tell you what your data can measure today and which method fits your budget decisions.
Personal reply, No obligation, English, Español
Start with SERMAN
Tell us about your project
First conversation
Diagnosis and proposal