Supply Chain & Operations Optimization

A store sells 200 units of a product per week. The automated replenishment system sees that and orders 200 for next week. Simple. Correct.

Except 50 of those units are sitting in the backend, never making it to the shelf. Sales drop to 150. The system sees 150 and orders 150. Next week, sales drop to 120. The system adjusts down again.

Six months later, a golden store is ordering 40 units and nobody knows why.

That is a negative feedback loop — and we found it running across hundreds of SKU-store combinations at P&G's Walmart Canada business. Not a bug. Not a data error. A structural flaw in how automated replenishment interprets demand signals when products require manual shelf restocking.

What supply chain analytics actually means

It does not mean dashboards with traffic lights. It means answering questions that the existing systems cannot: Is low demand real, or is it an availability artifact? Which routes are underperforming by revenue but overperforming by margin? If we increase safety stock by 10% in these 20 stores, what is the expected revenue lift — and does it exceed the carrying cost?

These are not BI questions. They are analytical questions that require custom logic layered on top of POS, inventory, logistics, and financial data.

Three engagements that illustrate the range

Replenishment intelligence at P&G/Walmart. We built a custom On-Shelf Availability algorithm using RetailLink POS data, inventory levels, and R/Knime workflows. The algorithm distinguished true low demand from availability gaps across golden stores. The counterintuitive recommendation — strategic overstocking — was piloted in 20 stores, generated visible results in two weeks, scaled to all-Canada within a month, and expanded to P&G's highest-margin category. Impact: $3M incremental revenue in 4 months, 10% stockout reduction, 5% improvement in on-time deliveries.

Operational visibility for the replenishment floor. The same Walmart engagement surfaced another problem: the supply chain team could not act on the algorithm's recommendations without real-time operational dashboards. We built a phantom inventory detection layer — logic that flags product existing in the system but unavailable for sale — and wired it directly into the replenishment workflow. The supply chain contribution was defining what signals matter (backend-to-shelf delay, phantom flags, restock labor capacity); the reporting automation that freed 120+ analyst hours monthly was a parallel workstream that made the insights consumable at scale.

Route-to-market digitization. A regional CPG brand had sales routes managed on paper and distributor performance tracked in spreadsheets. We centralized POS and distributor data, built dashboards for sales leaders by region and route, and applied basket analysis to identify which product combinations drove the highest margin per delivery stop. The result was not a technology transformation — it was visibility into a distribution network that had been running on relationships and intuition.

What we deliver

Custom algorithms and analytical models built on your operational data — not generic software. Operations dashboards that surface the 5-10 metrics your supply chain team needs to act on daily. Automated pipelines that replace manual data pulls. And documentation that explains what the models assume, where they break, and how to adjust them as your business changes.

When off-the-shelf tools are enough

If your supply chain runs standard SKUs through standard distribution with predictable demand, modern ERP systems handle replenishment well enough. Custom analytics add value when you have complexity: bulky products requiring manual restocking, multi-tier distribution, seasonal demand swings, or fresh/perishable inventory with short shelf life. If your supply chain manager says "the system works fine," trust them.

Diagnostic questions

Do you know which of your stores are declining because of real demand shifts versus availability problems your system cannot see? When your replenishment team overrides the automated system, do they track the reason — and does anyone analyze the pattern? Could you calculate the true cost of a stockout at your top 20 locations, including lost future demand from the customer who walked away?