Intelligent Inventory Forecasting for Hun**** Industries, AustraliaCase Study

The purchasing manager was making reorder decisions across a catalogue running to thousands of SKUs using a monthly report and a well-developed instinct. For the fastest-moving lines that worked; across the long tail it did not, and the business was carrying significant capital in stock that had not moved in over a year while periodically air-freighting fast movers at cost. Lead times measured in months from overseas suppliers meant errors compounded before they were visible. The harder problem was data quality. The ERP held years of transaction history, but stock adjustments, returns and inter-site transfers had been recorded inconsistently, so raw consumption figures were unreliable for training anything. Seasonality was genuine but irregular — demand tracked construction activity rather than the calendar. There was also justified organisational scepticism: the purchasing manager had decades of experience and no interest in a black box telling him he was wrong, which meant explainability was a functional requirement, not a nice-to-have.

We started with a data reconciliation exercise before any modelling, classifying historical transaction types and separating genuine consumption from adjustments and transfers — this took longer than the modelling itself and was the difference between a useful forecast and a confident wrong one. Forecasting runs as a Python service producing per-SKU demand projections with explicit confidence intervals, using different approaches by SKU behaviour: steady movers, intermittent demand and genuinely unpredictable lines are modelled separately rather than forced through one model. Recommended reorder points account for supplier lead time and its variability, not just average demand. Results surface through Power BI for management reporting and a React dashboard for the purchasing team's day-to-day work, where every recommendation shows the history and reasoning behind it — that explainability is what earned the purchasing manager's engagement. The service reads from and writes recommendations back to the ERP through its API, so purchasing continues to happen in the system the team already uses. Forecasts regenerate on a nightly schedule with each run versioned, so a recommendation made months earlier can be reviewed against what actually happened rather than being overwritten. We also added a manual override path that records who changed a reorder point and why, which keeps the purchasing manager in control while still building an audit trail the business did not previously have. Related: Power BI development, ERP integration, and our manufacturing practice.