Executive Summary

For hardware manufacturers and distributors, inventory is one of the largest line items on the balance sheet. Operations teams are asked to keep high service levels for key components, while finance would like to release working capital locked in slow-moving stock. In this case study we focus on a single high-value spare part – a laptop battery with item code FW1MN – and demonstrate how basic forecasting and inventory simulation can support smarter stocking policies.

The business question is simple: Can we reduce average stock levels for this battery while maintaining an acceptable service level for customer orders? Even single-digit percentage improvements translate into meaningful cash savings when scaled across thousands of SKUs.

Data &Methods

The data comes from a public Kaggle inventory dataset.

Time-series preparation and smoothing
We first convert the demand week field to a proper date index and aggregate weekly MRP_FCST_QTY to obtain a clean time series. To represent a simple, automatically updated forecast, we compute a 4-week moving average.

Inventory policy modelling
Next, we embed the demand series in a simple single-item inventory model. We assume:

• a fixed lead time of roughly two weeks (based on the 13-day value in the dataset),

• an (s,Q) policy with reorder point and fixed order quantity,

• lost sales when demand exceeds available stock.
  Safety stock is calculated via the standard textbook formula: ROP equals mu times L, plus z times sigma times the square root of L — where mu is average demand, L is lead time, and sigma is the demand variability.

We simulate two variants:

• Baseline policy – relatively conservative:

  • service level = 1 – (total lost sales / total demand),

  • initial stock ≈ 4.5 weeks of average demand,

  • order quantity ≈ 3 weeks of demand.

• Optimized policy – leaner but still robust:

  • safety factor z=1.8,

  • initial stock ≈ 2.5 weeks of demand,

  • smaller fixed order quantity.

For each policy we simulate week-by-week inventory, incoming orders, lost sales, and compute:

• average on-hand inventory (proxy for working capital), and

• service level = 1 – (total lost sales / total demand).

Results

From a business perspective, this case study illustrates three points:

• Even with very simple statistical tools – a moving average forecast and a textbook (s,Q) policy – it is possible to quantify the trade-off between inventory and service level for a specific component.

• For the FW1MN battery, a modest re-tuning of safety stock and order-up-to levels can yield about 9% lower average stock with almost unchanged service quality in the simulation. For a high-value spare part, this translates directly into lower working capital and potentially reduced write-offs.diate rule-of-thumb: which days require systematically higher limits or earlier refills.

• The same pipeline can be automated and rolled out across hundreds of SKUs, enabling “what-if” analysis (different service targets, lead times, or demand scenarios) and producing management-ready visualisations.