Back to Case Studies
Supply Chain LogisticsGlobal Retail Partners (name changed)
Inventory Resilience & Risk Profiling
Case study. Client names are changed for confidentiality. Results are measured against each client's pre-project baseline.
40%
Stockout Reduction at Peak
Top-200 SKU availability improved from 82% to 96% during the first post-implementation peak season.
15%
Holding Cost Reduction
Tighter, risk-responsive buffers freed £4.2 M of working capital previously tied up in excess stock.
18% → 4%
Top-SKU Stockout Rate
At peak season, critical item availability reached 96% — up from 82% the prior year.
6 hrs → 20 min
Reorder Planning Time
Automated nightly SAP writeback replaced a labour-intensive manual buyer process.
Case studyClient names are changed for confidentiality. Results are measured against each client's pre-project baseline.
The Challenge
After three consecutive years of supply-chain disruptions, Global Retail Partners was simultaneously over-stocked on slow movers and stocked-out on 18% of their top-200 SKUs during peak season. Their safety-stock model was a spreadsheet built on static average lead times with no mechanism to respond to deteriorating supplier reliability or geopolitical risk events.
Our Approach
- 1
Conducted a full data audit across SAP S/4HANA, 3PL warehouse feeds, and 140 supplier portals to establish a single version of truth.
- 2
Built a supplier risk scoring engine ingesting shipping-delay indices, geopolitical risk feeds, and financial-health signals.
- 3
Replaced the static safety-stock spreadsheet with a Bayesian dynamic model that adjusts reorder points daily based on demand volatility and supplier risk score.
- 4
Implemented an Apache Airflow pipeline that runs the model nightly, writes recommended reorder quantities to SAP via BAPI, and flags exceptions to buyers.
- 5
Created a scenario-planning module where supply-chain managers can simulate supplier failure events and see projected stockout risk.
The Solution
We replaced static inventory rules with a risk-intelligence system. A supplier risk engine continuously monitors external signals and translates them into daily safety-stock recommendations per SKU. These recommendations are written back to the SAP ERP system each night, meaning buyers wake up to updated reorder alerts that already reflect yesterday's disruption news.
Tech Stack
ML / AI
Python (PyMC / Bayesian models)
Dynamic safety-stock model with posterior updates on lead-time distributions daily.
Integration
SAP S/4HANA (BAPI / RFC)
Bidirectional ERP integration: reading actuals and writing back recommended reorder quantities nightly.
Geopolitical Risk APIs
Real-time country-level disruption signals (port strikes, sanctions, weather events) feeding the risk model.
Data Engineering
Snowflake
Cloud data warehouse consolidating SAP, 3PL, and supplier-portal feeds into a unified SKU dataset.
dbt
Transformation layer that models supplier risk scores and demand features consumed by the Bayesian model.
SQL Server
Operational staging database for 3PL inbound/outbound feeds before Snowflake ingestion.
Infrastructure
Apache Airflow
Orchestrates the nightly pipeline: risk scoring → model run → SAP writeback → exception alerting.
Visualisation
Power BI
Buyer exception dashboard: SKUs breaching risk thresholds, recommended actions, and scenario outputs.
Face a similar challenge?
Send us your challenge. We will reply to arrange a conversation and scope the next step.
Start with the decision
Turn one operational challenge into a measurable data plan.
Tell us about your challenge. We will reply to arrange an assessment and identify a practical route to delivery.
Request a data assessment