A reorder point is wrong when the usage data behind it includes a one-time spike, a stockout, or a promotion that will not repeat. The math looks calm and precise, but it is reporting an unusual month as if it were the normal one. Check the month before you trust the point.
Here is how it shows up on the floor. A distributor's top SKU gets reordered at double the normal quantity because one customer placed a bulk order in March. By May the bin is still half full, cash is tied up in stock nobody is asking for, and the warehouse manager is moving pallets to make room for the next shipment that should not have come in that size. Nobody lied to the system. The system just believed the month it was given.
Run that same mistake the other direction and it costs more. An item goes out of stock for ten days in a slow month. The reorder point, built off trailing usage, reads the suppressed demand as low demand and drops the trigger quantity. The next cycle the item runs out again, this time during a month that would have sold fine, and the order gets delayed while the customer calls twice. Wasted labor chasing a part that should have been on the shelf, unnecessary inventory sitting on a different SKU that absorbed the leftover cash, and a delayed order that did not need to happen.
Most software shows you the problem after the shelf is already empty or already overflowing and leaves you to work out why. A reorder point on a screen does not tell you whether the month behind it was normal.
The reorder point formula trusts whatever month you feed it
The standard formula is simple: average daily usage, multiplied by lead time in days, plus safety stock. Every variable in that formula is an average, and averages do not know the difference between a demand pattern and a one-time event. Feed it a month with a trade show order, a stockout, or a promotion, and the formula returns a number that is confident and wrong.
This is not an argument against the formula. Reorder points still work. The argument is that the formula needs a human, or a system watching on the human's behalf, asking one question before the number goes live: was last month normal for this SKU?
Three months that break the formula
A regional hardware distributor saw a single line item spike because one contractor customer placed a bulk order ahead of a job that would not repeat for a year. The reorder point jumped with it, and the next three purchase orders carried extra stock the warehouse had no reason to hold.
A food co-packer ran a promotion through a retail partner for six weeks. Usage during the promotion window was three times the normal rate. When the promotion window rolled into the trailing average, the reorder point for every SKU in that promotion stayed inflated for months after the promotion ended, long after the real demand had settled back down.
A parts supplier had a two-week stockout on a fast mover because a vendor shipment arrived late. Demand did not drop, the part was simply not there to sell. But the usage number for that month looked low, the reorder point dropped with it, and the item ran short again the very next cycle, right when the supplier's lead time had also stretched by a week.
A furniture maker saw the opposite problem on a slow-moving upholstery fabric. One large custom order pulled 400 yards off the shelf in a single week, a volume that SKU had not moved in two years combined. The trailing average treated that week as the new normal, and the reorder point stayed inflated for the next five purchase cycles. By the time anyone caught it, three months of carrying cost had piled up on a fabric that was back to selling five yards a week, and the cash that should have gone toward a fast-moving line was sitting in a bolt rack instead.
A small electronics assembler ran into the mirror image. A component went on backorder from its supplier for three weeks during what should have been a strong sales month. The usage number for that month came in low, not because demand dropped but because the part was not there to use. The reorder point followed the usage down, and the next purchase order undershot right as the supplier's lead time stretched from four weeks to six. The line waited on a part twice in one quarter for a problem that started with one month of bad data, not two.
The checklist before you trust last month's number
Before a reorder point is allowed to drive the next purchase order, run the month through this check. A planner working through a morning purchasing run does not need to audit every SKU this way, only the ones flagged as different from their own history, which is usually a short list once the system is watching for it.
- Did any single customer order account for more than a quarter of the month's usage on this SKU?
- Was the item out of stock, backordered, or on hold for any part of the month?
- Did a promotion, discount, or marketing push run on this item or a bundle that includes it?
- Does this SKU carry a seasonal pattern that this calendar month does not represent?
- Did the supplier's lead time change since the reorder point was last set?
A yes on any of those is a reason to look at the number by hand before the purchase order goes out automatically. A clean no across all five is a month worth trusting.
What Fishbowl does with the abnormal month
Fishbowl is the platform. Fishbowl Inventory, Fishbowl Advanced Manufacturing, and Fishbowl Advanced Warehouse are purpose-built parts of Fishbowl, matched to how the business runs. On Fishbowl Inventory and Fishbowl Advanced Manufacturing, data-focused AI reads usage by SKU and surfaces what is happening, what is coming, and what needs your attention, so an abnormal month stands out before it quietly rewrites the reorder point behind it.
On Fishbowl Inventory, that visibility goes a step further. The AI Operations Specialist uses live Fishbowl data to prepare purchase orders, sales orders, transfers, receipts, and adjustments for your approval. Nothing is saved or changed without the operator's approval. Ask, Approve, Done: the prepared purchase order shows the usage it was built from, so the operator can see a spike or a stockout sitting in the numbers before the order goes out, not after the bin is already wrong.
Brian Garrett at Telluric Companies, a beta tester, says the AI Operations Specialist saves him at least 45 minutes per purchase order, because he is reviewing a prepared order instead of building one from scratch and guessing at whether last month was a fair representation of what the shelf actually needs. See more results on the case studies page.
Fishbowl runs inventory, production, and the warehouse. QuickBooks or Xero keeps the books. Learn more about how the parts fit together on fishbowlinventory.com.
See how the AI Operations Specialist prepares a reorder with the abnormal month flagged, not buried, at fishbowlinventory.com/demo.
Fishbowl does the work. You make the call.
Frequently asked questions
How many months of usage should a reorder point be based on?
Most operators do better with a rolling three to six month average rather than a single month, because one month is too easy to distort with a single large order, a stockout, or a promotion. Longer windows smooth out the noise but can hide a real shift in demand, so the window should match how fast the item actually moves.
Should seasonal items use the same reorder point formula as steady sellers?
No. A seasonal item needs a reorder point built from the same month or season in prior years, not a trailing average of the last few months, because the last few months may include the slow season right before the spike or the spike right before the slow season.
How long does it take to get live data accurate enough to trust reorder points?
Paid setup and training typically run six to eight weeks, but the system is useful from the first live count. You do not have to wait for the full rollout to start seeing honest on-hand numbers and catching the months that do not belong in your averages.
Can a person still review a reorder point before the purchase order goes out?
Yes. The AI Operations Specialist prepares the purchase order from live data, but nothing is saved or changed without the operator's approval. If support is needed, customers can reach a person through phone, email, or Smart Chat.