Case Study Retail

Recovering $28M in annual shrinkage through full transaction event reasoning

A global retail chain reduced inventory shrinkage by $28M annually by applying LLM reasoning to 100% of transaction events — identifying organized retail crime patterns that were invisible at sampled point-of-sale data.

$28M

Annual shrinkage reduction

20% → 100%

Transaction event coverage

2,800

Locations covered simultaneously

67

Organized crime networks identified in first 90 days

$28M : $420K

Return on investment Year 1

1000×

Inference cost reduction made full-network coverage viable

The Organisation

Global Retail Chain · Retail

The Challenge

The retailer was analyzing approximately 20% of POS transaction events in their loss prevention analytics system. Organized retail crime operated in patterns specifically designed to avoid rule-based detection — distributing activity across locations, time windows, and product categories to stay below individual store thresholds. The network-level signatures of organized activity required full transaction coverage across all 2,800 locations simultaneously.

The Approach

Full transaction event coverage across all 2,800 locations with LLM reasoning configured to identify network-level behavioral patterns. The model analyzed transaction sequences, refund patterns, and product movement across the full store network simultaneously.

"Organized retail crime is a network problem. Our previous system looked at individual stores. You can't see a network-level pattern from a store-level view — or from 20% of the data."

VP of Asset Protection

Key Finding

The 67 organized crime networks identified in the first 90 days ranged from 3-person teams operating across a regional cluster of stores to a 180-person operation spanning 14 states. The largest network had been operating for 31 months, accumulating $8.4M in shrinkage from no individual store bearing enough loss to trigger an investigation.

Results at a Glance
Annual shrinkage reduction $28M
Transaction event coverage 20% → 100%
Locations covered simultaneously 2,800
Organized crime networks identified in first 90 days 67
Return on investment Year 1 $28M : $420K
Inference cost reduction made full-network coverage viable 1000×
Get in Touch

Talk to us about your data.

Tell us about your event stream and we'll show you what full LLM reasoning coverage looks like for your environment.

Or book a call directly →