From reactive reordering to intelligent ordering
A long-established maker of handcrafted lanterns and antique lighting faced a familiar tension: keep best-sellers available without trapping capital in stock. The board wanted an evidence-based answer before changing its operating model. Barrington ran a discovery built to answer one question with rigour.
The challenge
Ordering was reactive and judgement-based. Understock the wrong designs and customers wait 18–24 weeks; overstock and capital is trapped — the team still refers to a pre-pandemic “butter mountain” of excess inventory. The question: does our own history contain signals strong enough to order more intelligently?
What we did
We analysed eight years of ERP transactions, deliberately isolating repeatable, manufacturable demand — a clean dataset of roughly £40m in historical revenue.
- Identified true best-sellers by realised revenue, not raw order counts.
- Measured how revenue concentration shifts year by year — the winners keep changing, so static top-seller lists go stale.
- Reconstructed an inferred stock-availability signal from fulfilment data to test whether availability had quietly constrained sales.
- Stayed deliberate about what the data can and cannot prove — a conservative case, not a speculative one.
The outcome
Around 25–35% of revenue was fulfilled while products were partly or fully unavailable, and higher-availability designs systematically outperformed. Lifting availability on core designs toward roughly 95% could support £500–600k a year in additional revenue — assuming no change in demand, pricing or customer behaviour.
We always refer to that time as the “butter mountain” — the warehouse was groaning with stock. Each one had a reasonable data story, but we were holding far too much.