A maker: cost per unit, read daily

The reading

A specialty maker in the north. Fifteen formulators, a quality manager, a dispatch bench, and a partners’ room where the figures were assembled quarterly from four unrelated sources: a formulation workbook (Excel), a quality register (a shared document), the accounts system (off-the-shelf), and the dispatch log (a paper book signed by whoever was there).

The partners knew their revenue to the penny and their cost of goods to the nearest two points of margin. The two points — roughly £180,000 a year, on their turnover — lived somewhere in the gap between the four sources. Nobody could tell them where.

Four books, four truths

The formulators worked in Excel because Excel was fast and because the quality system did not know what a formulation was. The quality manager worked in a shared document because the accounts system did not know what quality was. The accounts system knew what a product was but could not cost the formulation that produced it. The dispatch log knew what had left the building but was legible only to whoever was on the dispatch bench that day.

Each book was defensible. The composite was not.

What we built

One record per batch, written at the point the batch was created. The formulation is a child table of the batch record. The quality tests are child tables of the batch record. The cost of ingredients and labour at the moment of production is attached to the batch record. The dispatch signature is a field on the batch record.

The batch is signed on dispatch. Not initialled. Not noted. Signed, by the dispatcher, electronically, with timestamp and username, which is as much ceremony as a dispatch has ever had.

The partners now read, on a dashboard the firm did not previously have the option of reading, the cost per unit of the previous day’s production. The margin they could not see because it was split across four books is now one figure, read daily.

The numbers at ninety days

What	Before	Ninety days in
Books of record	4	1
Lag on cost-per-unit	~ 90 days	< 24 hours
Batches dispatched unsigned	~ 1 in 20	0
Quality tests not linked to a batch	~ 1 in 7	0
Formulation versions tracked	1 (current)	all
Partner time on quarterly assembly	3 days	nil (continuous)

The two points of margin reappeared within six weeks. Three of the seven in the formulation book, two in the dispatch log, two in waste that no one had attributed to specific batches because no one had been looking at specific batches. The firm did not raise its prices. It stopped giving the two points away.

What we did not build

• No new quality system. The firm already knew what to test and when. We added the field to the batch; we did not reinvent the regime.
• No parallel accounts system. The off-the-shelf accounts remained in place. We linked to it rather than replace it.
• No cost model with ten parameters that nobody will maintain. The cost per unit is ingredient cost + labour cost + a fixed overhead percentage per batch. The partners can read it and understand it. If anyone wants a more complex model later, the data to build it is now there.

The standard

A new dispatcher should be signing batches correctly by the end of the first morning. That is the test. The record exists at the point of work. The dispatcher cannot sign without filling the field; the field is on the batch record; the batch record is where the next person goes to find out what left the building.

Cost per unit, read daily. The margin that was lost is now kept.