Every business carries things that used to earn and no longer do. A service in the price list that nobody has booked since spring. An event type a fund keeps paying entry fees for, with results that stopped justifying it. A client who ordered every month and then simply did not. A stock line gathering dust. None of them shows up as a loss in a monthly report: they just stop contributing, while the cost of keeping them around stays.
Two measurements find them: how revenue is spread across items, and how long ago each item last earned anything.
Most revenue comes from a few items
Sort everything you sell, by revenue, from largest to smallest, and plot the running total. The curve almost always rises steeply and then flattens: a small share of items does most of the work. This is the Pareto pattern, and ABC analysis turns it into classes: A items make the first 80% of revenue, B the next 15%, C the last 5%.
To show what this looks like on real data, we use a public example dataset of an online retailer, UCI Online Retail II: a year of sales of a UK online gift seller, 3,803 distinct products (SKUs, stock-keeping units) sold between 2010-12-01 and 2011-12-09.
Source: UCI Online Retail II (Chen, 2019), sheet Year 2010-2011, CC BY 4.0; our calculation
21.5% of items make 80% of revenue. In ABC terms that is 816 A items, 965 B items and 2,022 C items sharing the last 5%. The same shape appears in a list of services, a client base or a calendar of events. Class C is not useless by definition, but it is where things that stopped earning hide.
Days since last sale
Revenue share says how much an item earns; it does not say whether it still earns. For that, take the date of the last sale, booking or order for each item and count the days to today. Read it next to the run-rate: average units or revenue per week over a recent window, 90 days by default.
Source: UCI Online Retail II (Chen, 2019), sheet Year 2010-2011, CC BY 4.0; our calculation
75.4% of items sold in the last month, while 643 items, 16.9% of the catalogue, had not sold for over 90 days. Of these, 614 items (16.1%) sold in the first half of the year and then went silent; together they had made 1.9% of revenue. Small money, but each of them still occupies space in a price list, a catalogue, a schedule or a warehouse.
Starting thresholds
Starting points, not laws: tune them by how often a normal item sells and by what keeping an idle one costs you.
| Condition | Status | Action |
|---|---|---|
| Run-rate falls by half against the previous 90 days | Slowing | Check the reason: season, price, competitor |
| No sale or booking for 90+ days | Gone quiet | Stop spending on it; test price or offer |
| No sale or booking for 180+ days | Retire decision | Remove, clear or replace |
| Regular client silent for twice the usual gap | At risk | Contact before the next cycle |
If you hold stock: days of supply
For a business that keeps physical stock, one more number joins the list: days of supply = units on hand ÷ average daily units sold. Compared with lead time, it says whether to reorder; above roughly 90 days for an item that sells every week, it signals overstock. For items with no sales at all it divides by zero, which is exactly why days since last sale has to sit next to it: those are the lines where cash is frozen.
The example dataset has no stock levels, so we stop at the demand side here. In your own data, stock on hand and cost turn the idle list into a figure of cash tied up.
How to set it up
Export transaction lines with three fields: date, item (or service, or client) and amount. Compute revenue per item and its cumulative share, the date of the last sale, and the 90-day run-rate. Flag items by the thresholds above and review the list once a month, before planning the next period. A spreadsheet handles a few hundred items; above that, or with several sources, use a dashboard that refreshes by itself.
Limits
- Seasonal items look dead out of season: compare with the same period last year before retiring anything.
- A new item has no history yet; give it a full cycle before judging.
- Low revenue is not the same as no value: some items exist to bring clients in. Check what else those clients buy.
Data: UCI Online Retail II (Chen, 2019), licensed CC BY 4.0. All numbers on this page are calculated by our script on that dataset.