Detect, Analyze, Reduce OOS
KSS RETAIL PROVIDES UNIQUE CAPABILITIES to identify, analyze, and reduce a retailer’s shelf out-of-stock (OOS) problems, by continuous monitoring of shopper behavioral response. Our KSS Retail Heartbeat® machine learning techniques identify the historic nature and scope of the OOS problem, analyzing OOS trends by enterprise, banner, and individual store.
Additionally, we provide detailed profiling at the individual store/SKU level of OOS patterns, thus helping the retailer to identify and correct systemic problems, such as insufficient shelf space allocation and replenishment from the backroom, insufficient order quantity to meet promotional demand, and distribution voids. This OOS analysis includes:
- Historic trending of OOS rates (number of incidents, actual dollars lost) by enterprise, banner, and individual store.
- OOS profiling by start time, day, and duration.
- Root cause analysis, including: Identification of perennial store/SKU problems. These may be characterized by such factors as: insufficient shelf space allocation and replenishment from the backroom; insufficient order quantity to meet promotional demand; insufficient weekend supply or shelf replenishment; and distribution voids.
Once the scope and nature of OOS problems are understood, procedures for OOS reduction can be developed. These include the deployment of a unique real-time OOS alerting system.
RETAIL HEARTBEAT® SCIENCE DETECTS OOS by analyzing transaction log data and developing a large set of “Poisson” probability models that predict the rates of sales for each item, at each store, at any moment in time. These are blended into a comprehensive model by applying Bayesian classification tree methods that allow for accurate forecasting of item interactions.
The total data model is updated and refined nightly, and accessed minute by minute. Sales in each store are monitored via “trickle feed” every 15 minutes and movement of each item is compared against the rate predicted by the sales velocity model. When an item fails to sell at the rate predicted by the velocity model, the system generates an automated alert that is transmitted immediately to a store manager, clerk or even a vendor or merchandiser.
