Item Recommendations

A Power BI model that projects demand from historical sales patterns and generates replenishment recommendations (reorder point, min/max inventory, and order quantities).

Power BIDAXInventoryDemand PlanningReplenishment

Summary

Problem
Replenishment decisions relied on tribal knowledge and manual judgement calls, making it hard to standardize ordering across thousands of SKUs—especially when demand trends changed.
Solution
Built a projection model using historical demand + trend signals, then translated that into recommended ordering controls (reorder point, min/max inventory, and order quantities with rounding rules).
Result
A repeatable, explainable replenishment view that highlights what to buy, how much to buy, and which items are accelerating, declining, or have limited history.

Business impact

Stockout risk reduced
Uses projected demand + sale-frequency gap metrics to recommend reorder points and order quantities—helping prevent missed sales from under-ordering.
Overstock controlled
Recommends max inventory and rounds orders to practical multiples (when applicable), reducing cash tied up in slow movers and preventing inflated max levels.
Standardized decisions
Creates a consistent framework for replenishment across buyers and categories, replacing one-off “gut feel” decisions with transparent logic that can be reviewed and improved.

Gallery

Screenshots and artifacts from the build (sanitized where needed).

1) Demand projection design

I started by defining a stable way to estimate next-year demand using historical years and trend direction, so partial-year distortions didn’t overwhelm the forecast.

2) Frequency + gap-day metrics

Sales frequency and gap-day measures (average/median/last gap) were used to detect how consistently an item sells and how “late” it is relative to its typical cadence.

3) Recommendation logic
  • Calculated reorder point guidance using cadence + coverage logic.
  • Calculated minimum order quantity with rounding rules (order multiples when applicable).
  • Calculated max inventory recommendations to cap excess stocking while still supporting demand.
  • Added trend labels (increasing / decreasing / mixed / no history) to guide review and confidence.
4) Explainability

I included a calculation legend so users can audit the logic quickly and understand why the model recommended a value— making it easier to build trust and tune assumptions over time.