Case study
Making Pernod Ricard's AI sales engine trustworthy
Semantic built the data quality foundations of D-STAR, the in-house AI recommendation engine that Harvard Business School later featured as a model of successful AI adoption.
At a glance
Client
Pernod Ricard, the world's second-largest spirits company, active in more than 70 countries.
Engagement
15 months alongside Pernod Ricard's internal AI team, across 12 markets worldwide.
Semantic's focus
Data quality, built on dbt, Snowflake and the Elementary dbt package.
The challenge
D-STAR uses machine learning to optimise sales representatives' store visits and product recommendations. Sales teams had built their results on local expertise and intuition, and many were skeptical of a tool they had not asked for.
A recommendation engine earns trust only if its outputs are right. One bad suggestion from bad data is enough for a representative to stop listening, so the quality of the data behind D-STAR was a precondition for adoption, not a technical detail.
Global impact
Measured across 12 markets.
30% → 2.5%
Data quality errors, after Semantic designed and scaled the resolution process.
28 → 4 days
Average time to resolve a data quality issue.
6,000+ dbt tests
Implemented to detect errors automatically and make the data more reliable.
What Semantic did
We worked in dbt and Snowflake, using the Elementary dbt package to monitor data quality checks, and set up the ownership and process around them.
- Clear ownership. One IT owner and one business owner for each data field in every market.
- Standardized checks. A documented set of checks, with clear instructions to create, maintain and resolve them.
- Trained owners. Training sessions so owners can use the Elementary dashboards and the data quality playbook to resolve issues.
- Monitoring over time. Tracking of resolution trends to spot bottlenecks and keep improving data quality metrics.
Recognised by Harvard Business School
Harvard Business School professors Iavor Bojinov and Edward McFowland III wrote a case study on Pernod Ricard's digital transformation, covered in HBS Working Knowledge on November 21, 2025. They found the company had done “a fairly good job, better than some tech companies” at getting employees to adopt AI. According to the article:
- D-STAR reached 85% adoption across deployed markets by 2023.
- The tools delivered sales increases of 1.5% to 4.5%, depending on the market.
- In France, stores that followed D-STAR's recommendations outperformed control groups on net sales growth and net market share growth.