COMMERCE StoriesSubscribe
← Commerce courseCOMMERCE FROM ZERO · MODULE 9 · 10 min READ · FULL COURSE

Inventory Management

How Much to Hold, and Why It Matters

About access

Buyers receive the course-start link after checkout. This lesson link is unlisted, not account-based access control; anyone with the direct link can open it.

Labels: [Sourced] checked against a source listed at the end. [General] standard industry knowledge. [Our view] opinion, labelled as such. [Illustrative] made-up numbers to teach an idea.

1. The story: the toy buyer who knew by nightfall

In April 1987, K mart, then the second-largest US retailer, was converting its checkouts from hand-keyed registers to scanners that read bar codes printed by manufacturers. Its information chief said that when all stores were connected, about 10 million pieces of information would be added to its files every night. The LA Times described the payoff: when a boy in Topeka pesters his mother into buying a GI Joe, the toy buyers in Michigan know the same day. The paper also noted discount retail had been slow to adopt this. K mart had installed its first scanner about three years earlier, in 1984, and by then had automated 410 stores; supermarkets had been scanning since 1974 (Module 1). [Sourced: LA Times, Apr 20, 1987] Wal-Mart was already running its own inventory links (Module 9A, Module 1). The lesson: inventory is not just stuff on a shelf. It is stuff plus an accurate, timely record of the stuff. [Our view]

2. The one idea

Inventory is money waiting to become sales. Too little and you lose sales. Too much and you tie up cash, pay to store it, and risk marking it down. Managing inventory is balancing those two costs, SKU by SKU. [General]

3. The kitchen-table version

Your pantry again. Bottled water you drink daily: keep a week. Fancy vinegar you use twice a year: keep one. Ice cream: little, because it spoils. The rule is not "more" or "less", it is "match the stock to how it sells, how long the refill takes, and what a miss costs." [General]

4. Why hold inventory at all? Five reasons [General]

  1. Cycle stock: you buy in lots, not one at a time (Module 9A).
  2. Safety stock: protection from surprises in demand and lead time.
  3. Pipeline stock: goods in transit that you own but cannot sell yet.
  4. Anticipation stock: built ahead of a peak (holiday, promotion) or a price rise.
  5. Strategic stock: held for scarcity or supplier risk.

5. Safety stock, step by step

Safety stock is the cushion above expected lead-time demand. The standard teaching version: safety stock = z x standard deviation of demand over the lead time, where z is a number from the normal curve for the service level you want (about 1.645 for 95%, 2.326 for 99%). An APICS-magazine article states the demand-variability case this way and notes that a higher service level, e.g. 98%, needs significantly more safety stock. [Sourced: King, APICS, 2011] IBM adds there is no universal formula; the right one depends on the use case. [Sourced: IBM]

Worked example [Illustrative, computed by script]. Our $12 water bottle: demand averages 20 a day, daily standard deviation 6, lead time 7 days.

  • Lead-time standard deviation = 6 x square root of 7 = 15.9 bottles.
  • 95% service: 1.645 x 15.9 = 26 bottles. 99% service: 2.326 x 15.9 = 37 bottles.
  • Going from 95% to 99% costs 11 more bottles of safety stock, about 42% more, to cut stockout risk during the lead time from about 5% to about 1%.
  • In Module 9A we used 30. That sits between the two. This assumes demand is roughly normal and independent day to day, which is a simplification. Slow, lumpy items break it (Module 4 intermittent demand). [General]

6. Which items deserve attention: ABC and XYZ

  • ABC analysis: rank SKUs by annual sales or cost. A few items drive most of the money, so A items get tight control, C items light control. The idea is usually credited to H. Ford Dickie of General Electric in the early 1950s, applying the Pareto 80/20 idea. [Sourced: secondary summary]
  • XYZ: classify by how predictable demand is (X steady, Z erratic). Combining the two gives a grid for choosing policies. [Sourced: ASCM article]
  • Example split [Illustrative]: 20% of SKUs = 70% of sales (A), 30% = 20% (B), 50% = 10% (C). Count A items weekly, C items yearly.

7. The scorecard: turns, days of supply, GMROI

  • Inventory turnover = cost of goods sold / average inventory (at cost). Days of supply = 365 / turns.
  • GMROI = gross margin dollars / average inventory at cost. It asks: for each dollar sitting in stock, how many dollars of margin did I earn? [Sourced: Investopedia, formula; General]

Worked example [Illustrative]. Bottle: price $12, cost $5, 7,300 sold a year, average stock about 552 bottles (half the 764 order size plus 170 as in 9A, rounded).

  • COGS = 7,300 x $5 = $36,500. Average inventory at cost = 552 x $5 = $2,760.
  • Turns = 13.2; days of supply = 27.6.
  • Gross margin = 7,300 x $7 = $51,100. GMROI = 18.5. This is far above a real store's number because we looked at one fast item; a whole store carries slow items too. Evidence on the trend: academic studies of US retailers found inventory holdings fell from the 1980s to 2000s (abstracts; I have not read full papers). [Sourced: Springer/Exa abstracts]

8. Accuracy: the unglamorous problem

If the system says 12 and the shelf holds 0, no forecast or reorder rule helps. Fixes: cycle counts (count a slice every day rather than shutting down once a year) and, increasingly, RFID. Auburn University's RFID lab published a study asking whether RFID improves inventory accuracy (preliminary analysis); I have the title and summary only, so no figures are quoted. [Sourced: Auburn RFID Lab] Shrink, losses from theft, error and damage, averaged 1.6% of sales for fiscal 2022 in the NRF's last full survey, up from 1.4% in fiscal 2021, equal to $112.1 billion. [Sourced: NRF, National Retail Security Survey 2023]

9. Where AI fits

  • Setting safety stock and reorder points per SKU-location from forecast error, not rules of thumb. [General]
  • Flagging records that look wrong (system says in stock, but no sales for days) and prioritizing counts. [General]
  • Computer vision and RFID for shelf counts. [General]
  • Forward thread: inventory becomes a "network" decision, a system placing each unit where the next order will probably come from. Zara's approach, with models deciding each shipment from two central warehouses to its stores, is an early example. [Sourced: Caro and Gallien, Interfaces; abstract]

10. Exercise

Same bottle, but daily demand sd is 4 and lead time is 9 days. Find 95% safety stock. Answer: 4 x 3 = 12; 1.645 x 12 = about 20.

11. Quiz

  1. Name three reasons to hold inventory. Cycle, safety, pipeline, anticipation, strategic.
  2. Why is a 99% service level much more costly than 95%? Safety stock grows faster than service level.
  3. What does GMROI measure? Margin dollars earned per dollar of inventory at cost.
  4. Why does record accuracy come before forecasting? Wrong stock records defeat any rule.

Sources

CURIOUS? TEST THE CLUES

Curiosity check

Pick an answer and see why. No scores, no pressure. All shop examples are invented practice scenarios.

01 Why is a 99% service level much costlier than 95%?
02 What does GMROI measure?