You will get a Bayesian Marketing Mix Model to optimize your ad budget
Project details
I build Bayesian Marketing Mix Models that tell you what your marketing is actually worth and where your next dollar should go. Most measurement is guesswork: last-click takes the credit and the ad platforms mark their own homework. An MMM fixes that by measuring each channel's true, incremental contribution to your sales.
What sets my work apart is that I don't just fit a model and hand it over. I've built hierarchical Bayesian MMMs in PyMC-Marketing in production, managing over $50M in annual media spend, and I calibrate every model against real incrementality experiments so the numbers are grounded in causality, not just correlation. I'm also a PhD candidate in Data Science researching causal inference, so the methodology is rigorous and the uncertainty is handled honestly.
You get more than a model: clear channel ROI, marginal return curves, and a budget recommendation you can act on, explained so a non-technical team can trust it.
What sets my work apart is that I don't just fit a model and hand it over. I've built hierarchical Bayesian MMMs in PyMC-Marketing in production, managing over $50M in annual media spend, and I calibrate every model against real incrementality experiments so the numbers are grounded in causality, not just correlation. I'm also a PhD candidate in Data Science researching causal inference, so the methodology is rigorous and the uncertainty is handled honestly.
You get more than a model: clear channel ROI, marginal return curves, and a budget recommendation you can act on, explained so a non-technical team can trust it.
Machine Learning Tools
NumPy, pandas, PyMC, Python, Python Scikit-Learn, R, scikit-learn, SciPy, SQLWhat's included
| Service Tiers |
Starter
$1,200
|
Standard
$3,500
|
Advanced
$6,500
|
|---|---|---|---|
| Delivery Time | 10 days | 21 days | 30 days |
Number of Revisions | 1 | 2 | 3 |
Number of Model Variations | 1 | 2 | 3 |
Number of Scenarios | 1 | 3 | 5 |
Number of Graphs/Charts | 5 | 10 | 15 |
Model Validation/Testing | - | ||
Model Documentation | - | ||
Data Source Connectivity | - | ||
Source Code | - | - |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$300 - $1,200
Additional Revision
+$150
Additional Model Variation
(+ 4 Days)
+$600
Additional Scenario
(+ 1 Day)
+$250
Additional Graph/Chart
(+ 1 Day)
+$60
Model Validation/Testing
(+ 3 Days)
+$500
Model Documentation
(+ 2 Days)
+$300
Data Source Connectivity
(+ 2 Days)
+$400
Source Code
(+ 1 Day)
+$500Frequently asked questions
About Diogo
Senior Data Scientist | MMM, Attribution & Marketing Measurement
Helsinki, Finland - 2:24 pm local time
Most teams are measuring marketing by guessing. Last-click takes the credit, the platforms mark their own homework, and nobody can say what a channel is truly worth. I fix that with proper marketing measurement: Marketing Mix Modeling, multi-touch attribution, and incrementality testing that tells you what is causing sales, not just what is correlated with them.
I work with founders, growth leads, CMOs and marketing teams who spend real money on paid media and want to trust their numbers before they scale them.
What I build:
- Marketing Mix Models (MMM), hierarchical Bayesian, to measure each channel's true contribution and reallocate budget;
- Multi-touch attribution (Markov chains, Shapley value) across messy, non-linear customer journeys;
- Incrementality and causal inference: geo-lift, geo-holdout, difference-in-differences, synthetic control, A/B and experimentation programmes;
- Predictive models: churn, LTV, lead scoring, propensity, next best action;
- The analytics stack underneath it all: GA4, Google Tag Manager, BigQuery, Snowflake, dbt, Looker dashboards.
A few results from past work:
- 1.5M EUR unlocked through MMM-driven budget reallocation;
- 5M EUR+ in incremental revenue from a structured A/B testing and experimentation roadmap;
- 32% lift in projected lead value by moving paid acquisition to value-based bidding;
- 27% lower CPA and 15% higher conversion rate through better measurement and targeting.
How I work: I treat marketing measurement as causal. I never trust a single model, so I triangulate MMM against attribution and incrementality experiments to make sure the numbers are real. And I translate all of it into one clear recommendation a non-technical founder can act on, because a model nobody uses is worthless.
You are a strong fit if:
- You spend on paid media (Google Ads, Meta, TikTok) and cannot prove what is working;
- You need MMM, attribution, or incrementality testing done properly;
- You have data but no clear path from data to decision;
- You want senior judgement, not a junior running a template.
Not a fit if:
- You want last-click attribution to keep telling you what you want to hear;
- You are looking for the cheapest possible setup over a correct one
My stack: Python (pandas, scikit-learn, PyMC), R, SQL, BigQuery, Snowflake, dbt, Vertex AI, Looker, GA4, Google Tag Manager, Google Ads, Meta Ads.
I am a PhD candidate in Data Science at NOVA-IMS researching causal inference, so you get academic rigour combined with in-house growth-team pragmatism.
Expertise: Marketing Mix Modeling, MMM, Media Mix Modeling, Marketing Measurement, Multi-Touch Attribution, Attribution Modeling, Incrementality Testing, Geo-Lift, Causal Inference, Bayesian Statistics, Marketing Analytics, Predictive Modeling, Churn Prediction, Customer Lifetime Value, LTV, Propensity Modeling, A/B Testing, Experimentation, GA4, Google Tag Manager, Conversion Tracking, BigQuery, Python, Data Science.
Send me a brief and I will reply within a few hours.
Steps for completing your project
After purchasing the project, send requirements so Diogo can start the project.
Delivery time starts when Diogo receives requirements from you.
Diogo works on your project following the steps below.
Revisions may occur after the delivery date.
Kickoff and data audit
We align on your goals and channels on a short call, then I review your spend and outcome data for coverage, quality, and any gaps before modeling begins.
Data preparation
I clean and structure the data into a modeling-ready dataset, aligning spend and outcomes by date and channel and engineering the variables the model needs.