How rebuilding a forecasting spreadsheet saved a fast-growing snack brand roughly £120,000 a year - and what it taught me about operating in the dark.
When I joined Proper Snacks, the ordering process was straightforward. Sales orders went into an Excel spreadsheet. The spreadsheet became a CSV. The CSV was emailed to the factory. The factory made the product. The product shipped to our logistics company. Customers ordered. Repeat.
It worked. It also meant we held four weeks of stock against demand we could not clearly see.
Most of what we could not see was already written down somewhere - just not somewhere that fed into the planning. Our logistics company had a customer portal with live stock holdings. Tesco, Morrisons and WH Smith each shared their own sales data and stock projections, on different cadences and in different formats. Sitting next to each other, those numbers would have told us exactly how much stock we needed and where it should be. Sitting in five separate inboxes and portals, they told us nothing useful - and we compensated for the blindness by overstocking.
The cost of overstocking was the kind that does not show up as a single bad decision. Warehouse fees on units that sat too long. A steady trickle of stock going out of date on the slower SKUs. A quietly bloated line item every month.
The project, when it got priority, was to build a single planning tool that pulled those data sources into one place. I built it in Excel, because Excel was the tool the team already trusted, and because the goal was not to introduce new software - it was to make the existing process see further.
The tool did three things.
First, it pulled SKU-level sales and stock data from each major retailer in whatever format that retailer used, and normalised it into a consistent view. Tesco's orders did not look like Morrisons' which did not look like WH Smith's; the planner did the unglamorous work of making them comparable.
Second, it generated separate demand forecasts per customer based on each retailer's own ordering patterns, then combined them into a single SKU-level master forecast. An 80/20 weighting recognised that our biggest accounts shaped the picture far more than a long tail of smaller ones.
Third, and most usefully day to day, it could be refreshed against the live stock holdings in the logistics company's customer portal. That meant when it came time to send the next order to the factory, the team was not ordering against a guess. They were ordering against what we actually had, where we actually had it, and what the next four weeks of demand actually looked like.
The effect was that we could place tighter orders to the factory and hold less inventory at the logistics company. Stock holdings came down from four weeks of cover to two. The logistics company was pleased; they got back warehouse space they could resell to another customer. Waste dropped on the slower SKUs. Cash that had been sitting in over-ordered product went somewhere more useful.
By the end of the year the rebuilt planner had saved roughly £120,000 in annual operating cost - mostly storage, partly waste - with no change to how the team experienced the work, beyond the spreadsheet being smarter.
Most of what I now do for early-stage consumer brands looks like a version of this. There is usually a real operating answer hiding inside data the brand already has, sitting in three or four different systems that do not talk to each other. The job is rarely to introduce new software. It is to make the existing process see further. The savings on the back of that are usually larger than the brand expects, because the cost of operating in the dark compounds quietly.
If your brand is sitting on the kind of operating mess this story describes, a thirty-minute call is the cheapest way to find out whether I can help. Book a call or say hello.