Case Study · 01 Heineken · Global Analytics & AI Hub

Allocation AI

A Bayesian budget-allocation engine that decides where Heineken's marketing money earns the most — steering €3.2B in annual spend from intuition to evidence.

Role
Product owner / designer
Timeline
2022 – 2024
Domain
Marketing mix & RM
Stack
Bayesian ML · Azure
Allocation AI — one source, many optimally-filled vessels
One source, many vessels — the allocation problem in a single image: how much pressure does each channel deserve?
€600M
Targeted yearly revenue uplift by 2029
€305M
Gross-profit uplift by 2027 (~€16M cost-to-achieve)
2–4%
Net-revenue uplift in Mexico & Brazil pilots
250+
ROI projections evaluated per optimization run
01 — Overview

What Allocation AI does

Every year Heineken's operating companies face the same question: where should the next marketing dollar go? Five million on TV, or on trade promotions? More weight behind Heineken, or a local power brand? Historically these calls leaned on experience and last year's plan.

Allocation AI answers it with evidence. It builds a Bayesian response model for each OpCo, brand and channel, learning how sales actually react to investment. The core output is a set of response curves — spend on one axis, expected volume or ROI on the other — that let teams test budget scenarios before committing a cent.

An optimizer then reads those curves and proposes the allocation that maximizes return within real strategic constraints: brand-role rules, long-term brand-building floors, and short-term sales limits.

02 — The Challenge

A powerful model no one outside data science could use.

Opacity

Response curves and Bayesian priors are trustworthy to statisticians and invisible to the commercial leaders who actually sign off budgets.

Trust

Asking a market director to hand budget authority to a black box is a hard sell. The output has to be legible, arguable, and overridable.

Scale

Whatever we built for one pilot market had to hold up across dozens of OpCos, each with its own brands, channels and data quality.

03 — My Role

Owning the layer between the model and the decision.

01
Translated the model into a decision tool

Framed response curves and ROI projections as scenarios a commercial team could read, compare and defend — not as statistical artefacts.

02
Designed the scenario & constraint workflow

Built the flow where planners set constraints (brand roles, spend floors and caps) and watch the optimizer respond — keeping humans in control of the machine.

03
Bridged data science and commercial teams

Ran discovery with both sides, turning statistical outputs and business intuition into a shared product language and a single roadmap.

04
Drove the multi-market rollout

Standardized the experience so a tool proven in Brazil, Mexico and the UK could scale across OpCos with varying data maturity.

04 — How It Works

From three years of data to one allocation.

Step 01
Decompose history

Regression models trained on 3+ years of data separate sales growth into ATL, BTL, commercial levers and control factors.

Step 02
Build priors

Bayesian priors are built and updated per touchpoint, then models are trained and aggregated for each distinct channel.

Step 03
Plot response curves

Each curve maps spend to expected volume and incremental ROI, simulating impact at every spend level.

Step 04
Optimize under constraints

The optimizer weighs 250+ ROI projections at once, allocating to the best touchpoints within the Brand Role Framework.

Bayesian regression Response curves Industry-standard optimizer Heineken Azure Platform Automated training & validation
05 — Impact

A structural shift, staged over years.

Replacing intuition-based budgeting with precision modelling compounds — pilots prove the lift, global rollout scales it.

Pilot
Mexico & Brazil — 2–4% net-revenue uplift

First deployments established the baseline: measurable revenue gains from smarter allocation alone.

2027
€305M gross-profit uplift

Optimizing ABTL allocation across top OpCos, at a cost-to-achieve of roughly €16M — a ~19× return.

2029
~€600M yearly revenue uplift

The enterprise-scale target as the capability rolls out across every applied market under the EverGreen strategy.

06 — What We Measured

The KPIs behind every allocation.

Financial
  • Return on Investment (ROI)
  • Gross Profit (GP) impact
Volume
  • Gained volume
  • Sell-out volume per channel
Investment
  • ATL touchpoint spend
  • BTL budget contributions
07 — In Context

The macro engine of a three-part system.

Allocation AI sets the sandbox — how much each channel and brand gets — that Smart Flighting and Promo Advisor then execute against. Allocation secures the funding; Smart Flighting schedules the impact; Promo Advisor executes the offer.

Where & how much · You are here
Allocation AI

Sets the annual budget across OpCos, brands and channels.

When · Coming soon
Smart Flighting

Stretches each budget across the calendar for maximum pressure.

What & why · Coming soon
Promo Advisor

Fills the calendar with the specific winning promo mechanics.

Keep exploring

Two more case studies on the way.

Smart Flighting and Promo Advisor complete the picture. In the meantime, head back to the work or get in touch.

Back to all work Get in touch
Tomasz Czarnecki © 2026 · czarnecki.ai