Profitable on paper, short on cash, and still reordering products the market has started to leave.
Here is a mistake that never feels like a mistake. A product's sales turn bumpy. Some months move, some do not. It reads as an ordinary slow patch, so the reorder goes out on schedule.
The trouble is that a product quietly dying looks almost exactly like a slow patch. So does a new product still finding its market. From the numbers on the screen, a product on its way up and one on its way out show the same symptoms. Every quarter the guess is wrong, cash goes into something the market is already walking away from.
What is found when someone finally looks
Decline does not show up in the total first. It shows up in the pattern. Orders get irregular, the gaps between them stretch, the same few customers keep buying and no new ones arrive. Set one quarter against the last, and that is indistinguishable from a seasonal dip.
The one view that separates a real fade from noise is the same quarter compared across 3 years, one product at a time. That cancels the seasons and shows the direction. Almost nobody runs it.
It is harder still because 4 different situations produce the same chart:
- A real decline. The market is leaving.
- A stockout. It stopped selling because there was nothing to sell.
- A lost account. One large customer walked.
- A new product. Low and jumpy because it is climbing, not falling.
Read those wrong and something healthy gets written off, or something dead keeps getting fed. That judgment is the whole job, and it is why a standard aging report does not solve this.
What the blind spot costs
A dying product treated as a healthy one is a number on the balance sheet that is quietly wrong, and real decisions sit on it. Prices are set from it. Reorders are set from it. The bank lends against it.
Put a rough size on it. Take the inventory line on the balance sheet and a conservative 10% of it, the share that is non-moving or quietly fading. For a business carrying a few million in inventory, that is 6 figures, frozen, at full value. Run the real number and it is usually higher.
It hides because the cash dies in the part of the catalog that is too dull to look at. Each fading product is small, a few hundred dollars, never worth an afternoon. Together they are the largest position on the balance sheet that nobody has looked at. Each one a rounding error. The sum a write-down.
Why it is not caught in time
Nothing exotic is needed to find it. Pull every product into a spreadsheet, compare the same quarter across 3 years, calculate months of cover on each, and read the patterns. It is a long, careful day of work that has to be redone every quarter, and it loses every time to whatever is on fire that day. So it runs once before a crisis, or never.
The 80/20 rule makes it worse. Trapped cash does not live in the top sellers. A product in decline is one sliding down the ranking, so by the time it is a problem it has already left the 20% anyone watches. The rule was never wrong. It was a workaround for scarce attention.
The fix: look at all of it, every quarter
Half of what AI does here a person could do, and half a person could not.
The mechanical half: comparing each product to its own 3-year history and to its neighbors. AI does it across everything at once, the same way every time, every quarter and not once a year. That half is speed and consistency.
The other half cannot be done by hand at all. A person reviews products one at a time and has forgotten the 12th by the 400th. AI holds the entire catalog at once, so it sees when a cluster is fading together for a shared reason. It can weigh a decline against the market outside the business, so the answer says whether demand is leaving the category or only leaving this company. And it judges the hardest call, new product or dying one, against far more patterns than one analyst carries in memory.
The judgment stays with a person. AI hands back only what needs a decision. It also fails on bad inventory data: the cleanest logic cannot fix a number that was wrong before it arrived.
The result
When I worked at 3M, non-moving inventory went from 8.5% to 5% of inventory value, with fill rates held at 100% through the semiconductor shortage. Nothing about the inventory itself changed first. The classification did: every product was read against current demand and not against its history.
To see what the finding looks like on one page, open the furniture dashboard: $264,590 sitting still for 180 days, inside a month that looked fine on the P&L.
Why now
Found early, this is recovered cash and a cleaner balance sheet. Left alone, it surfaces the day the number has to be right: a lender reviewing the line, an audit, a buyer looking closely. They find it first, and the gap is money already spent. The cash is the same either way. The timing is the whole difference, and it decides who in the room can defend the number.
Is this your number?
If sales are fine and the cash never is, this is the work: where the money is stuck, and which decisions put it there.