The solution worked because it connected a dramatic inference-cost improvement to complete transaction coverage and measurable loss reduction.

WorkedDid not

Coverage

100% of events¹

About 20% of events¹

Network scale

2,800 locations¹

Partial network view¹

Inference economics

1,000× cost improvement¹

Coverage constrained by cost¹

Business outcome

$28M reduction¹

Missed network patterns¹

Logswiz’s approach made inference cost reduction the practical bridge between advanced reasoning and full-network operation. The retailer moved from analyzing about 20% of point-of-sale events to applying reasoning to 100% of transaction events across 2,800 locations.¹

The outcome was not defined only by model activity. The deployment identified 67 organized retail crime networks in the first 90 days and supported an annual shrinkage reduction of $28M.¹ The reported first-year return on investment was $28M to $420K.¹ That combination links system design, operating reach, and financial value in one chain of evidence.

LLM inference cost reduction enabled full‑network shrinkage detection