Case Study
How High is High: Breaking the Negative Feedback Loop in Automated Replenishment
Identified critical flaw in Walmart's automated replenishment, developed custom OSA algorithm, and drove $3M incremental revenue across two P&G categories in 4 months.
While working as Replenishment Analyst at P&G's dedicated Walmart Canada team, I identified a critical flaw in RetailLink's automated replenishment that was causing golden stores to systematically underperform. By developing a custom On-Shelf Availability algorithm using R and Knime, I proved automation was creating negative feedback loops — and convinced stakeholders to pilot a counterintuitive solution: strategic overstocking. The pilot delivered immediate results, scaling from 20 stores to all-Canada implementation and expanding to P&G's highest-margin category.
Key Results
The Transformation
The Challenge
Walmart's RetailLink automated replenishment system was creating a hidden negative feedback loop, systematically destroying sales at P&G's best-performing stores. The mechanism was subtle: bulky Femme Care products arrived at stores and went into backend storage. If staff didn't bring them to shelves on time, on-shelf availability dropped. RetailLink couldn't distinguish 'product in backroom' from 'product not needed.' It saw low sales, interpreted them as low demand, and ordered less next cycle. Less inventory meant even lower shelf availability, even lower sales, and even smaller orders. Golden stores with initially excellent sales were declining to single-digit weekly forecasts through no fault of actual demand.
- The pattern repeated across the entire Femme Care category — dozens of stores showing the same systematic decline, suggesting a system-level issue rather than isolated demand problems
- Hundreds of SKU-store combinations made manual intervention impossible — we needed a data-driven approach to distinguish availability-driven low sales from genuine low demand
- The longer the loop ran, the harder it was to reverse — stores with months of suppressed orders had built up such low forecasts that even returning to normal shelf stocking wouldn't immediately fix the automated ordering
Our Approach
Algorithm Development:
- Developed a custom On-Shelf Availability (OSA) algorithm using RetailLink POS data, inventory levels, and R/Knime workflows. The algorithm compared actual POS velocity against expected demand based on store demographics and historical performance. This let us identify which low-selling SKU-store combinations were availability issues vs. genuine low demand
- The key insight was the 'golden store' pattern — stores that had performed well historically but were now in decline. We built a scoring model that flagged stores where the decline trajectory matched the negative feedback loop signature rather than organic demand shifts
Pilot Design & Stakeholder Conviction:
- Proposed the "How High is High" pilot: strategic overstocking of 10-20 Femme Care products in approximately 20 golden stores across Canada. The thesis was counterintuitive — we were asking Walmart to order more of a product that their system said was in low demand
- Presented findings to Walmart replenishment and buyers alongside my manager and sales lead. We made a bold commitment: "If this doesn't work after one month, my manager and I will rent a U-Haul and personally buy back excess inventory from all Ontario stores." The data was convincing, but that personal stake sealed the approval
Rapid Scale & Expansion:
- Launched the pilot within one week of approval. Overstocked stores showed POS recovery within two weeks as product returned to shelves consistently. Walmart requested expansion before the initial one-month evaluation period ended
- Scaled to all-Canada implementation after one month of strong performance. We then applied the same methodology to Shaving — P&G's highest-margin product line — which showed identical feedback loop patterns with bulky razor multipacks
The Outcome
Revenue & Market Impact:
- $3M incremental revenue generated in 4 months: +$2M Femme Care POS (4% category increase) and +$1M Shaving POS (4% category increase)
- 10% reduction in stockouts across managed categories, with a 5% increase in on-time deliveries as corrected ordering patterns stabilized
- The entire revenue lift came from fixing an existing system flaw — no new products, promotions, or marketing spend. Pure recovery of demand that the automation had been suppressing
Recognition & Strategic Value:
- Pearl Award from P&G Walmart Canada team — one of the highest recognition honors for the Walmart account
- Methodology adopted as standard practice for high-velocity, bulky product categories across P&G's Walmart business. The lesson: automation is powerful, but unmonitored automation can silently destroy value
What the Client Says
Arturo quickly became the go-to data person—organized, diligent, and a strong partner. We co-created ecommerce replenishment processes; his out-of-the-box thinking consistently improved how the team worked.
Bebul Soomro
Director of Supply Chain, Procter & Gamble
Want similar results for your organization?
Get in touchReady to turn data into decisions?
Let's discuss how Clarivant can help you achieve measurable ROI in months.