Index
/02 Work
AC/PD Portfolio
P.01 — Case study

Kroger Deli
Reinvention

Temperature, vision, and the problem of "compliant" systems that still fail.

60% → 10%
Out-of-range holding
6 mo.
Build to pilot
~$800M → $1.6B
Deli growth target
IoT Computer Vision Operations Hardware Sensing
Context

Kroger was working toward a major expansion of its deli business — roughly doubling the program from ~$800M toward ~$1.6B. The opportunity was real and the ambition was clear. But the system underneath it was less stable than it looked.

On paper, hot food was being held correctly. In stores, product quality still drifted. Chicken could be technically in process, still saleable, and already moving toward a worse customer experience. The problem was not a lack of procedures. The problem was that the real operating system inside the store was only partially visible.

The problem

Prepared food quality is fragile. It changes with time, temperature, handling, and store behaviour. A process diagram can show the ideal path, but stores do not run as diagrams. They run under labour pressure, rushes, interruptions, uneven execution, and constant small deviations.

What we were seeing was a gap between food safety logic and food quality reality. A product could remain nominally safe and still be held too long, drift too warm, or lose quality in ways that were hard to detect early. The store had no useful system for seeing that drift in time to act on it.

The system looked compliant. It wasn't behaving that way.
Insight

The breakthrough was understanding that temperature alone was not the real issue.

The important question was not just, "Was the product in range?" It was: How long has this product spent outside the right operating band, and what does that mean for quality?

That sounds simple, but it changes the whole system. Once you frame the problem that way, periodic checks stop being enough. You need persistent visibility — not occasional snapshots.

Solution

I helped shape a system that combined wireless temperature monitoring, quality logic, and store-facing operational feedback.

  • A tighter operating band for hot-hold performance (140–160°F)
  • Wireless sensing across the holding environment
  • Computer vision logic to estimate product status and detect over-hold conditions — flagging product held more than 2 hours outside acceptable range
  • Dashboards and alerts that translated raw signals into store action
  • Workflow logic for when product needed to move to refrigerated holding or be handled differently

This was not just a hardware project. It was a sensing and decision system built to survive real store conditions.

Execution

The work required more than technical development. It took cross-functional alignment across store operations, culinary, in-store quality, suppliers, and the broader deli initiative. The technical side had to be credible, but the operating model also had to be practical.

  • Hardware development and store integration over 6 months
  • Defining thresholds people would trust — store operations and culinary buy-in
  • Supplier coordination for sensing infrastructure
  • Pilot-first deployment in real store environments, not conference rooms
  • Iterating thresholds based on real data, not process diagrams

The point was not to create a beautiful sensor network in theory. The point was to create a system that stores could actually use.

Result

The pilot materially improved out-of-range holding performance — reducing noncompliant holding from roughly 60% to 10%. More importantly, it gave the business a clearer mechanism for protecting product quality while scaling deli operations.

It helped turn an invisible quality problem into something measurable, manageable, and improvable.

What this shows

I do not start with the gadget. I start with the operating failure. I look for the place where a system appears fine from far away but breaks down under real conditions — then work backward from that failure mode and build the sensing, logic, and workflow needed to make it visible. That is usually where the real leverage is.

← All work Next: Freshness Sensor Platform ↘