Industry Food service

Inventory managed by quantity alone loses product in the places the ledger cannot see.

-28%Dry goods inventory loss rate
-23%Fresh goods inventory loss rate
93% → 98%Ingredient quality inspection pass rate

Figures reflect relative change before and after implementation, presented in de-identified form under client confidentiality terms.

A central kitchen production line, stainless benches, crate dollies and cooking equipment behind
AI-generated illustration

Central kitchens and multi-site restaurant groups have long managed inventory by quantity rather than condition. Shelf life comes off the packaging, picking follows first-in-first-out — but ingredients within a single delivery arrive in genuinely different states, so loss accumulates where the ledger cannot see it, and is usually only discovered when the only remaining option is disposal.

Add batch-to-batch yield variation in fresh produce, long lead times on imported goods and exposure to currency movement, and prep accuracy becomes very hard to hold together on floor experience alone.

Our approach is to make quality a computable variable: physical measurements taken at goods-in, combined with storage environment data, give each batch a real decay curve, so picking order becomes first-to-spoil-first-out rather than first-in-first-out. Food safety rules go into the algorithm for the same reason — rules that live in an SOP are the rules that get bypassed when the kitchen is busiest.

Case 01Food service · Supply chain

Central kitchen: from managing quantity to managing quality and food safety

Prep accuracy, ingredient quality control and procurement cost are where multi-site food service hits its margin ceiling — and conventional practice tracks quantity while ignoring quality.

-28%Dry goods inventory loss rate
-23%Fresh goods inventory loss rate
93% → 98%Ingredient quality inspection pass rate

The problem

Prep accuracy, ingredient quality control and procurement cost together determine whether a multi-site food service brand can break through its profit ceiling.

Static shelf life and high spoilage
Managing to the printed date and first-in-first-out ignores the real condition of goods at intake, so fresh items degrade unpredictably and losses run high.
Yield varies batch to batch
Produce output rates differ enough that prepping to a standard bill of materials regularly leaves sites short or oversupplied.
Import cost exposure
Long-lead imported spices and high-value seasonings move with exchange rates and season, with no data behind the buying decision.
A shelf-life date is printed on. Real condition has to be computed.

What we built

  1. 01
    Demand forecasting and dynamic prepPOS data from each site is combined with weather, temperature, holidays and local activity, and run through machine learning and time-series models to project item-level sales one to two weeks out, then worked backwards into daily base prep volumes for the central kitchen.
  2. 02
    Quality-driven storage and food safety (ServSafe Ready)Moisture content and surface temperature at intake, combined with zone temperature and humidity data, produce a real degradation curve per batch and a first-to-spoil, first-out picking order. ServSafe time and temperature control and cross-contamination rules are built into the algorithm itself, so batches approaching a quality threshold are routed to the safest available location and the whole storage and handling chain meets an international food safety standard.
  3. 03
    Yield learning and dynamic procurementHistorical loss data feeds back into the next purchase quantity, ingredient quality grades are recommended per dish based on end use, and price monitoring on high-value imports triggers strategic buying recommendations at favourable points in the cycle.

The business case

Ingredients near end of life used to be discovered too late to use, leaving disposal as the only option. Knowing the real degradation curve early means they now go into normal production first — which removes not just the loss, but the labour and handling that disposal consumed.

What we planned to do, and did not

The original plan called for RFID to verify storage locations automatically. Field testing showed the volume of liquids and sauces in a central kitchen interfered badly with radio reads, and accuracy never reached a usable standard.

The team changed the architecture quickly, moving to handheld barcode scanning at process checkpoints backed by environmental IoT sensors — keeping the system's instructions fully enforceable on the floor without adding to the staff's workload.

Need Professional Assistance?

Our team are senior engineers and former executives with 20+ years in commercial product development — from apps serving hundreds of millions of users to B2B SaaS platforms serving tens of thousands of merchants. In recent years the focus has been enterprise AI adoption, from mapping the opportunity through to tuning what is live. Whether that means:

  • Mapping AI opportunities and sequencing them
  • Preparing data flows and integrating existing systems
  • Building forecasting and optimisation models
  • Monitoring and tuning after go-live
  • Test strategy and process review
  • Building and adopting test automation

We can provide professional advice and concrete solutions for teams of different scales and types. If you're facing similar challenges, feel free tocontact our consulting team.