The Reorder Point Nobody Reviews

Strip the arithmetic away from any reorder rule and what is left is a set of claims about the future that somebody made without writing them down. A rule that says reorder when the position falls to thirty is making six claims at once. It claims you will sell at the rate you have been selling. It claims the supplier will take the time they used to take. It claims you will not have a week where three customers all want the same thing. It claims the case you order arrives in the quantity and the condition you expect. It claims nobody will move stock out of that bay for a reason that is not a sale. And it claims you will notice when any of those stops being true, which is the claim that fails most often, because nothing in the system is watching.

The number is the easy part and it is the part people defend. The rule that gets argued over in a meeting is the thirty. The six claims above are almost never in the conversation, and that is why the meeting goes the way it goes: somebody proposes fifty, somebody else says that is too much stock, and both of them are arguing about a number while the assumption that actually matters, the lead time, sits in a field that was last edited when the supplier was first onboarded.

So the first change is not to the number. It is to write the claims down. A reorder rule that states its own assumptions is a different object from one that does not, because the first one can be wrong in a way somebody can notice, and the second one is just a number that happens to produce orders. The useful form of the rule is a sentence you could read out loud to a new person: we order when the position falls to thirty, which covers about four days of selling at the rate we sold last month, and about three days of delay if the supplier slips, and the delay figure came from the twelve receipts we recorded last quarter.

A manual order and an automatic order are both an order. The difference is that a manual order is read by a person on the day it is placed, and an automatic order is not read by anybody at all. That is the whole argument, and it is stronger than it sounds, because the person placing a manual order is doing something else at the same time. They know the customer who has been asking about the item. They know the delivery that went wrong last week. They know the two cases still sitting in the returns bin that are not on the shelf but are, technically, the same units. A manual reorder is a moment when the state of the world is available to the decision, and an automatic reorder is a moment when it is not.

The consequence is that a wrong assumption inside an automatic rule does not produce a conversation, it produces a quantity. A wrong assumption inside a manual rule produces a person saying, actually, give me a moment, and then either changing the number or placing the order anyway with a note that says why. One of those produces information. The other produces a purchase order with a number on it, and the number is the only artefact, and the number looks identical whether the assumption behind it is sound or not.

There is a second reason, and it is about scale. A manual rule fails one item on one day and a person notices. An automatic rule that is wrong does not fail one item, it fails every item that shares the assumption, at the same moment, in the same direction, and the failure arrives as a single large order that looks like a decision rather than like a defect. By the time a rule that has been quietly wrong for a quarter is discovered, it has usually placed several orders, and the investigation has to separate the orders that were right from the orders that were wrong, which is much harder than preventing the first one.

None of this is an argument against automating anything. It is an argument about which errors you want a machine to have. A rule that reflects a decision somebody made deliberately, on a date, with the assumptions written next to it, is a good candidate for automation. A rule that nobody has looked at since the day it was set is not a system, it is a fossil, and automating it only makes the fossil fire more reliably.

One item, a 5 kg pack, supplied in cases of twenty, sold through a single store. Every figure below is constructed for this article so you can check the arithmetic; none of it is a measurement, and no benchmark is implied.

The rule, set nine months ago: reorder when the stock position falls to 30 units, order 60 units, which is three cases of twenty. The 30 was built as roughly one week of selling plus a small buffer. At the time it was set, sales ran at 20 units a week, so 20 units is one week, and the extra 10 was the buffer. Check the buffer in days: 20 units a week spread over 7 days is 2.857 units a day, so 10 units is 10 divided by 2.857, which is 3.5 days of cover. That is the entire reasoning. The supplier quoted a lead time of 5 days at onboarding, so the number covered a week of demand and the 3.5 days of buffer absorbed a supplier who was slightly late.

Then the supplier changed. Nothing about the arrangement was formally renegotiated; the goods simply started arriving later, for reasons that belong to the supplier's own operation. The recorded lead time in the file is still 5 days, because that field was set at onboarding and nobody has written to it since.

The rule does exactly what it was built to do. The position falls to 30 units on a Tuesday, the threshold is crossed, 60 units are ordered, and the order is placed the same afternoon. The rule has fired correctly. Now do the arithmetic with the real lead time rather than the assumed one. With 12 days of lead time instead of 5, the demand that arrives before the delivery is 20 units a week multiplied by 12 divided by 7, which is 20 times 1.714, which is 34.3 units, so 35 units of demand against 30 in stock. The shelf runs out roughly five units before the goods land, and it does so even though the supplier delivers on the day they promised the last nine times.

That is the shape of the failure, and it is why the memo about discipline does not fix it. The rule did not fire late. The rule did not fail to fire. The rule fired on the day it was designed to fire, the order went out on time, the supplier performed on time, and the customer still found an empty shelf, because the number in the file had been describing a supplier that no longer exists.

Now the part that is genuinely uncomfortable, which is that this rule can be quietly wrong and still be firing usefully for months. Consider the same rule applied to a different item in the same shop, one that sells at 8 units a week. At 8 a week, 30 units is 8 times 7, so 30 divided by 8 is 3.75 weeks of cover, and a 12-day lead time is comfortably inside that. The rule does the right thing for that item and the wrong thing for the fast one, and the two failures are indistinguishable from the outside because both produce a purchase order on the day the threshold is crossed. The rule is not wrong in general. It is wrong for one item, on one assumption, and the reason it is wrong is that the assumption was never recorded anywhere a person could check it against reality.

The second uncomfortable thing is what the wrong rule does to the money while it is wrong. Each order is 60 units at a constructed landed cost of 640, which is 38,400 per order. If the shop continues to reorder at 30 while the shelf actually empties and demand continues at 20 a week, the pattern is a second order roughly every three weeks rather than every four, and the difference between those two numbers over a quarter is real money sitting in a godown rather than on a shelf. We are deliberately not turning that into a turnover or a stock-cover benchmark, because we have not measured one and a printed benchmark would be worse than no number. The arithmetic is above if you want to run it against your own figures.

The 18% GST figure is not used in this article, because a reorder rule has nothing to do with tax and adding one would only invite a question we should not answer here. How purchase invoices and input tax are treated in the value you carry for stock is a question for your own chartered accountant.

The fix is not a better number. It is a rule that is legible enough to be argued with, which means the assumptions are written in words next to the arithmetic, each one has an owner, and each one has a last-checked date. Five of them, in this order. The demand assumption, stated as a rate and a period, so that next quarter somebody can compare it against what actually happened rather than against what somebody remembers. The lead-time assumption, stated per supplier and, crucially, stated as what actually happened on past receipts rather than as what was quoted at onboarding, because the quote is the number that is guaranteed to be stale. The buffer, named as a comfort figure and a number of days, not disguised as a calculation. The order quantity, with the reason for it written down, whether that is freight, case rounding, or cash tied up. And the review itself: a name and a date, so that somebody is answerable for the rule being current rather than the rule merely existing.

A rule with those five lines is a document somebody can hand to a new person. A rule without them is a number in a field, and the field will still be there in three years, unchanged, describing a business that no longer exists. The cost of writing the five lines is perhaps twenty minutes per item. The cost of not writing them is the memo about discipline, which arrives after the customer has already noticed.

The review does not have to be a process with a form. It has to be a date somebody owns, and a small number of questions with numeric answers: did we sell at the rate in the rule, did the goods arrive when the rule said they would, and did the buffer get used. Three questions, once a quarter, per item that matters. If the answer to the second one is no, the number is wrong and no amount of tightening will fix it, because the problem is upstream of the reorder point and in a different organisation.

Stock movement, purchasing, warehouses and low-stock signals live in the Commerce area. Receipts and issues are documents that move stock, a warehouse is a place stock is held in, and a low-stock signal is your own threshold crossing a position you can read. The point worth holding onto is what that signal is: a number crossing a line that you set. It is a prompt for a person. Expenses are not part of this: expense claims live in Nox-Billings only, and an invoice and a payment are two different records, where paid is a projection of allocations rather than a column you set, so a purchase invoice marked settled can always be walked back to the receipts behind it.

The two denials come first because this is exactly the page where a reader expects them and where marketing copy usually goes quiet. There is no demand-forecasting model in NoxOrigin. Nothing here projects what you will sell next month, next quarter, or on a seasonality curve, and we are not going to describe a threshold as if it were one. And there is no automatic purchase-order engine. Nothing here will raise a purchase order by itself because a position crossed a number. If you want a reorder to happen without a person, today it is a person who raises it, and what you can see is the stock position, the reorder level, and the purchase the team recorded against it. There is no Automations area yet: no trigger, condition, action, or run history, and no plan adds one. So the honest answer is that a reorder is a decision somebody made, visible on the record rather than fired by the product. That is a narrower claim than an intelligent purchasing engine, and it is the one we can actually honour.

Also absent, so that this page is not read as a list of things we forgot: no manufacturing capability set, meaning no bill of materials, no shop-floor scheduling and no production costing, which is a different category of software if your stock is consumed by making things. No change-request, milestone or deliverable record; no contract editor and no e-signature, so a scope change is a new quote against the same project. No timesheet record and no payroll. We do not file GST returns or any other statutory return, and we do not pursue statutory recovery. Duplicate detection flags candidates and a person reviews them, with match rules and merge behaviour as a setup decision and nothing merging automatically. Shift close and day-end reconciliation are assisted-setup maturity rather than self-serve switches. And no price appears anywhere on this page, because the commercial terms are a conversation rather than a number on a blog post.