Skip to content
allanninal.dev/build
Collaborate

Part 3 of 7 · Stock transfer planner series ~5 min read

How the source gets protected

The easiest way to make a transfer planner harmful is to let it optimise the destination without any constraint on the source, at which point it will steadily strip the quiet branches.

Key takeaways

  • The source keeps its own cover, computed from its own demand.
  • A slow-selling item still needs one on the shelf if it is displayed.
  • Size the transfer from the surplus above cover, not from the destination’s need.
  • Watch for a site that is always the donor; it is usually a symptom.
  • A site with zero of something it displays is a different problem from a stockout.

Cover at both ends

How the quantity available to transfer from a source site is determinedA vertical chain of five steps entered by a box labelled A candidate transfer, where A has surplus and B needs it. Step one computes A's own cover in weeks of its own demand. Step two adds the display minimum if the item is on a shelf, with a side box noting it is often one or two regardless of demand. Step three computes the surplus above that, giving the movable quantity. Step four asks whether it is enough to be worth it, since fixed costs do not scale; if not it exits to No transfer, too small to pay. Step five moves the smaller of the surplus or B's need. A note says the display minimum is the rule people forget, and customers notice it immediately.AWS ACCOUNTA candidate transferA has surplus, B needs itA's own coverweeks of its own demandPlus display minimumif it is on a shelfOften one or tworegardless of demandSurplus above thatthe movable quantityEnough to be worth it?fixed costs do not scaleNo transfertoo small to paynoMove the smaller ofsurplus, or B's needThe display minimum is the rule people forget, and customers notice it immediately.
Fig 1. How the movable quantity is determined. The source’s requirements are computed first and the destination gets what is left over.
  • Compute
  • App integration
  • Security & identity
  • Analytics
  • Front-end & mobile

Cover is per site

The source’s cover requirement comes from its own demand and its own replenishment lead time, exactly as it would if no transfer were being considered. A branch selling one a month with a four-week supplier lead time needs a couple in stock; a branch selling nine a month needs considerably more.

Using a single company-wide minimum produces the wrong answer at both ends: it strips the busy branches too far and leaves too much at the quiet ones, which is the imbalance the system is supposed to be reducing.

Display stock is not surplus

An item that sits on a shelf where customers can see it needs at least one there, even if it sells twice a year. A transfer that takes the last one leaves a gap on a shelf and removes the only way anybody discovers the product exists.

This is a rule that any experienced retailer states immediately and that essentially no automated system implements, because it is a merchandising fact rather than an inventory one. One field per product, set once.

The permanent donor

Transfers in and out of four branches over a yearA stacked bar chart with four bars. Two series: transfers out this year in orange, and transfers in in green. Branch one: four out and thirty-one in. Branch two: twelve out and fourteen in. Branch three: thirty-eight out and three in. Branch four: nine out and sixteen in. A note says branch three gives and does not receive, so its buying is wrong rather than its stock.020406080~35Branch 1~26Branch 2~41Branch 3~25Branch 4Transfers out, this yearTransfers in, this yearBranch 3 gives and does not receive. Its buying is wrong, not its stock.
Fig 2. A year of transfers by branch. A site that is consistently the donor is usually being over-ordered rather than being a useful reservoir.

This pattern is a genuinely useful finding and it is invisible when transfers are looked at one at a time. A branch that is permanently the donor is receiving stock it does not sell, and the fix is in whatever decides how much it gets, not in moving things afterwards.

Transfers in that case are the symptom management: real money spent every month to correct a purchasing allocation that could be corrected once. Putting the chart in front of whoever sets allocations is worth more than any individual transfer.

Zero is not always a stockout

A site with none of a product it never sells and does not display is fine, and flagging it as a stockout produces a stream of transfer proposals for things nobody wants there. The distinction is whether the site ranges the product, which is a decision somebody made, not something to infer from a stock level.

Ranging is one flag per product per site and it removes a large amount of noise. Without it, the imbalance detector treats every product as if every site should carry it.

Next: waiting for a van.

All posts