All case studies / Case study 12 / 12
Mirvane Pharma
Predictive supply disruption radar
Seeing supplier delays coming before they stop production
In short
The system matches open purchase orders against world news (port strikes, weather, authority notices): as soon as an event threatens a supplier, it finds the affected orders, estimates the delay and proposes a solution to procurement.
- Client
- Mirvane Pharma
- Sector
- Pharmaceutical manufacturing
- Size
- 800 staff, 120 international suppliers
- 9 daysaverage warning ahead of a supplier delay
- −40%line stops caused by supply issues
- 24/7monitoring of 120 suppliers
The context
Mirvane Pharma imports its active ingredients and packaging from Asia and Europe. Until now, procurement found out about a delay when the delivery didn’t arrive, often with production lines standing still.
The challenge
- 01
Warning signs exist (announced strikes, typhoons, port congestion), but no one can follow them all.
- 02
A distant event has to be linked to the actual orders it threatens.
- 03
One false alarm too many, and teams stop trusting the system.
- 04
The answer must be actionable: what to do, not only what to fear.
Our approach
Open orders, suppliers and their routes are synced from the ERP. Scheduled jobs continuously collect news, weather alerts and port notices, analysed by a RAG pipeline that only keeps events touching a tracked place or route. The system estimates the likely delay, lists the affected orders and drafts an alert with alternatives (another port, another qualified supplier).
The solution, step by step
- 01
Your open orders
Each order is linked to its supplier, its port and its route.
- 02
The world, watched
News, weather and port notices are analysed continuously, day and night.
- 03
An event detected
A strike is announced at the port of Genoa from the 12th.
- 04
The impact, quantified
Two orders are affected; the likely delay is estimated at 9 days.
- 05
An alert with solutions
Procurement gets the alert and an alternative: route through Marseille.
Results
- 9 daysaverage warning ahead of delays
- −40%production stops caused by supply
- 85%of alerts rated useful by buyers
- 120suppliers monitored continuously
What we deliver
- Order and supplier sync from the ERP
- Scheduled collection: news, weather, port notices
- RAG pipeline for event detection and matching
- Delay estimates and suggested alternatives
- React dashboard and procurement alerts
Technologies
- Python
- Celery
- Redis
- Pinecone
- News API
- React
Timeline
8 to 10 weeks
Team
1 AI engineer, 1 full-stack developer
Illustrative cases: clients and figures are fictional, representative of the solutions we deliver.