Rev'Up pricing deployment in Spain

Context

Internship at Pernod Ricard HQ (Global Digital Acceleration, Rev'Up team), Feb–Jul 2025. Rev'Up is the group's data program for Revenue Growth Management: pricing and promotion optimization across international markets. My main assignment was the Spain rollout of the pricing module, from promotion-free data preparation through Bayesian elasticity estimation to price scenario generation. Work was done with a senior data scientist and in weekly sync with Spanish commercial teams.

The Spanish market had no structured, model-backed tool to prepare annual distributor price negotiations. RGM teams needed interpretable price elasticities (own and cross-product effects) and simulations that respect local business rules.

Constraints

• Sparse and noisy retail data: ~3 years of weekly data, promotions to strip out before any elasticity fit
• Model quality judged by business plausibility and ranking within categories, not only statistical error on volumes
• Deployment cadence driven by market reviews and annual negotiation calendar

Approach

Standard Rev'Up deployment in three blocks: data scope and QC, Bayesian training on non-promotional weeks (PyTorch on Azure, Snowflake + dbt upstream), then P&L simulation and price optimization exported to Power BI. Promotion weeks flagged with internal tooling plus manual corrections where needed. Three iteration cycles with weekly feedback from the Spanish market.

Tech stack

• Python 3.8, PyTorch, pandas
• Snowflake, dbt, Azure ML pipelines
• Power BI (market-facing outputs)
• Azure DevOps (PR reviews, pair programming)

Outcome

Spain moved from ad hoc pricing inputs to a repeatable elasticity workflow validated with local RGM and commercial teams. Elasticities and cross-effects were reviewed in weekly sessions and embedded in dashboards for negotiation prep. Final market kick-off was scheduled just after the internship. The model is planned for periodic retraining on fresh data.