Reorder points, and comparing quotes that are not comparable

Two numbers that stop most stock problems

Stock management in a small business usually means noticing you have run out. Two figures, calculated once per important item, fix most of it.

Reorder point: the level at which you order more:

reorder point = (average daily usage x lead time in days)
                + safety stock

Safety stock covers the days the supplier is late and the weeks you sell more than usual. A rough version that works: your worst recent week’s usage, or a third of the lead-time demand, whichever is larger.

A worked case. You sell 8 units a day on average. The supplier takes 12 days. Lead-time demand is 96. Your worst week was 90 units, about 13 a day, so safety stock is around 60. Reorder point: 156. When stock reaches 156, you order, regardless of how it feels.

That arithmetic is trivial and almost nobody does it. It is worth doing for the ten items that matter and ignoring for the rest.

Where AI fits

Not in the calculation. That is a spreadsheet, as always.

In getting the numbers out of your records. “Here are 18 months of sales. Give me average daily usage, the worst week, and the trend, for each product code.” Structure and extraction, done in the sheet.

In noticing patterns. Seasonality, the item that always runs out in the same month, the product whose sales have been declining for a year without anyone noticing. This is analysis nobody in a small business has time for, and the data is already there.

In writing to suppliers. The chase, the complaint about short delivery, the request for better terms. Routine, awkward, and better with a draft to react to.

Comparing quotes that are not comparable

This is where an hour is well spent, and where suppliers rely on you not doing it.

Three quotes arrive. One is per unit, one per box of 12, one per kilogram. One includes delivery, one adds it, one adds it above a threshold. Payment terms differ. One quotes a price valid for 30 days and one for 12 months.

The task is normalisation: putting everything into the same units so the numbers can be compared. That is structured transformation, which is exactly what a model does well:

Three supplier quotes below. Put each into these fields:
price per single unit in local currency, delivery cost per
unit at an order of 500, payment terms in days, minimum
order, price validity, and anything conditional.

Use ONLY figures stated. Where a quote does not state
something, write NOT STATED. Do not estimate.

Then check every extracted number against the original document. The extraction is doing the tedious part; the arithmetic goes in a sheet, and the decision is yours.

The NOT STATED entries are frequently the most useful output. The cheapest quote per unit is often the one with the undisclosed delivery charge and the 60-day validity.

What it cannot tell you

Whether the supplier is any good. Reliability, willingness to help in a crisis, whether they answer the phone in August. This is why you keep two suppliers for anything critical, even when one is cheaper.

Whether the sample is representative. A first delivery is not evidence about the tenth.

What the relationship is worth. The supplier who has extended you credit twice is not competing on price alone, and a spreadsheet that ignores this will eventually recommend something expensive.

The record worth keeping

One line per delivery: date ordered, date promised, date arrived, quantity right or wrong.

Three months of that turns “they’re a bit unreliable” into “average nine days late, and short on one delivery in four”, which is a completely different conversation to have with them. It is also the input for the lead time in your reorder formula, which is otherwise a guess dressed as a number.

Do this for ten items, not everything

Most stock lists follow the usual pattern: a small number of lines account for most of the money and most of the trouble. Sort by annual value, usage multiplied by cost, and work down.

The top ten get a reorder point, a recorded lead time and a second supplier. Everything else gets a simple rule: reorder when the shelf looks half empty. Trying to manage two hundred lines properly means managing none of them, and the effort is better spent elsewhere in the business.

Asking better questions of your own data

Once eighteen months of sales are in a spreadsheet, a handful of questions are worth asking every quarter:

• Which products have declined for three consecutive months?
• Which are sold almost always together?
• What did we run out of, and what did that cost in lost sales?
• Which lines have not moved at all this year?

That last one usually finds money sitting on a shelf. The answers come from formulas, the questions come from you, and the model’s job is to build the formulas and explain what the pattern might mean, never to report the figures itself.