C4, Lesson 3: Inventory: How Much to Hold, and When a Smart System Is Not Enough
Tags: Sourced / General / Our view / Illustrative. About 13 minutes.
1. The story: Nike, February 2001
On February 26, 2001, Nike filed an 8-K with the US Securities and Exchange Commission saying it expected third-quarter earnings of $0.34 to $0.38 a share, against earlier guidance of $0.50 to $0.55. Chairman and CEO Philip Knight said Nike had expected a flat quarter because of weak US footwear revenue, and added: "During the quarter, we also experienced complications arising from the impact of implementing our new demand and supply planning systems and processes which resulted in product shortages and excesses as well as late deliveries." [Sourced: Nike Inc. Form 8-K, filed Feb 27, 2001 (period of report Feb 26, 2001)] CIO magazine (June 15, 2004) quoted Nike's operations head, Roland Wolfram, saying that Nike's problems with its i2 demand-planning software, which it said began in June 2000, cost it "more than $100 million in lost sales," in the magazine's words, and that Nike's own account blamed "tactical" issues: slow performance, poor integration, bugs, and planners who had not been trained enough before go-live. Nike told the magazine that it had fixed these by fall 2000, and that in spring 2001 it moved short- and medium-range sneaker planning into its SAP system. Nike's technology executive said the software was never intended to be the hero of the project and that planning is "a mixture of art and technology." [Sourced: CIO magazine, "Nike Rebounds," June 15, 2004; secondary; the $100 million is Nike's operations head as quoted by CIO, not from the 8-K; Nike moved short- and medium-range sneaker planning to SAP in spring 2001 and kept i2 for apparel, per the verifier] The point: the filing says shortages and excesses happened together. A planning system can get the total roughly right and still put the wrong things in the wrong places. [Our view]
2. The one idea
Inventory exists to cover two things: demand you did not predict and supply that arrives late. The stock you hold for those is safety stock. How much you need depends on how uncertain demand is, how long resupply takes, and how often you are willing to run out. Better forecasts shrink safety stock; they do not remove it. [General]
3. The kitchen-table version
You sell about 100 bottles a week and a refill takes 2 weeks to arrive. You need 200 bottles on hand just to cover the wait. The extra bottles you keep in case a busy week arrives are your safety stock. The more certain you want to be not to run out, the more extra you hold, and each step up costs more than the last. [General]
4. The three numbers
- Lead time: how long from order to arrival.
- Demand variability: how much weekly sales jump around (standard deviation).
- Service level: the share of cycles you want to avoid a stockout (for example 95%). Safety stock is roughly z x demand variability x square root of lead time, where z rises with service level (1.28 at 90%, 1.64 at 95%, 2.33 at 99%). This is the standard textbook rule. [General] (See also the replenishment module in the Commerce course, Commerce Module 9A.)
5. Worked example: the price of certainty
All numbers [Illustrative], computed by script. An item sells 100 a week on average with a standard deviation of 20. Lead time is 2 weeks. It costs $7 and carrying cost is 20% a year.
| Service level | z | Safety stock | Yearly carrying cost |
|---|---|---|---|
| 90% | 1.28 | 36 units | $51 |
| 95% | 1.64 | 47 units | $65 |
| 99% | 2.33 | 66 units | $92 |
| Going from 95% to 99% costs 19 more units, 40% more safety stock, for four more points of availability. | |||
| What better forecasting buys. If an AI forecast cuts demand variability from 20 to 14, safety stock at 95% falls from 47 to 33 units, 14 fewer. That saves about $20 a year per item. On 1,000 items: $20,000 a year. Compare with the tool's cost. [Our view] | |||
| Notice the limit: a forecast cannot cut lead time or shortages caused by a supplier. And if the system's output is wrong, it can raise stock in the wrong places, as the Nike filing's "shortages and excesses" suggests. [Our view] |
6. Where AI helps and where a simple rule wins
- Helps: many items, changing demand, many locations, learning lead-time patterns.
- Simple rule wins: stable, fast sellers; a few suppliers; no one to act on the result.
- Both: set safety stock by rule, let the forecast adjust the demand input, and have a person review exceptions. [General] Nike's executive described the approach as running models against history and then having "people look at it to make sure it makes sense," trusting the model more for reliable sellers. [Sourced: CIO magazine as quoted; secondary]
7. What it costs and what it breaks
- Cash tied up in stock, plus storage and markdowns.
- Stockouts cost sales and goodwill.
- Change risk: a new planning system can disrupt orders during rollout. Nike's 8-K tied its shortfall partly to implementation. [Sourced: Nike 8-K]
- Training: planners need to know how to read and override the system (Nike's account said training had been inadequate). [Sourced: CIO magazine; Nike's statement]
- Data: bad lead-time or demand data produces confident wrong numbers.
8. Monday checklist
- For your top 20 items, list lead time and weekly sales variability.
- Compute safety stock at 95% by the rule in Section 4.
- Compare to what you actually hold. Where are you far above or below?
- Write down who may override the system and when.
- Before any system change, run old and new side by side for several weeks.
9. Try it yourself (no code, 30 minutes)
Pick one item. Estimate lead time (weeks), average weekly sales and their standard deviation (STDEV in a spreadsheet). Compute safety stock at 90%, 95% and 99% with z of 1.28, 1.64 and 2.33. Compare with what is on your shelf.
10. Self-check
- What did Nike say in its February 2001 filing? Implementation of its new demand and supply planning systems caused shortages and excesses and late deliveries.
- Which three inputs set safety stock? Lead time, demand variability, service level.
- In the example, what is the cost of going from 95% to 99%? About 40% more safety stock.
- What can a better forecast not change? Lead time or supplier shortages.
- Why run old and new systems side by side? To catch errors before they cause real shortages.
11. Where this is going
Planning software will keep getting better at estimating demand and lead time. People who understand the basic rule will be able to tell when its output is off. [Our view]
Sources
- Nike Inc. Form 8-K, filed Feb 27, 2001: https://www.sec.gov/Archives/edgar/data/320187/000032018701000003/0000320187-01-000003.txt
- Nike Form 10-Q, quarter ended Feb 28, 2001: https://www.sec.gov/Archives/edgar/data/320187/000032018701000007/0000320187-01-000007.txt
- CIO magazine, Nike rebounds: https://www.cio.com/article/264637/enterprise-resource-planning-nike-rebounds-how-nike-recovered-from-its-supply-chain-disaster.html
Review flags
- The 8-K is primary; I read its opening section. The "$100 million in lost sales" figure is attributed by CIO (June 15, 2004) to Nike operations head Roland Wolfram; it is secondary. ACCEPTED LIMITATION (Oct 7, 2026): the Nike 2001 annual-report letter and the fiscal 2001 10-K text retrievable in this pass did not surface the figure, so it is not confirmed in Nike's filings. The lesson attributes it to CIO magazine and Wolfram "in the magazine's words", not to Nike.
- CIO article date confirmed by the verifier: June 15, 2004. Per the verifier, the whole ERP project cost about $400 million with i2 about 10% of it; this lesson does not state the i2 cost.
- Blame for the problem is disputed between Nike and its vendor in press accounts; the lesson reports Nike's filing and Nike's statements without assigning blame to the software vendor.
- The safety-stock formula is the standard textbook version; real systems use more detail.
- Example numbers are illustrative.
Claim ledger (C4 Lesson 3)
| # | Claim | Source | Verified | Status | Re-check |
|---|---|---|---|---|---|
| 1 | Feb 26-27, 2001 8-K: EPS $0.34-0.38 vs $0.50-0.55 | Nike 8-K | 2026-10-06 | Read | none |
| 2 | Knight quote on implementing demand and supply planning systems | Nike 8-K | 2026-10-06 | Read | none |
| 3 | Over $100M lost sales (Nike ops head Wolfram via CIO, Jun 15, 2004); Nike's "tactical" account; moved to SAP spring 2001 | CIO magazine | 2026-10-06 | Read; secondary | primary source |
| 4 | "Art and technology"; people review model output | CIO magazine | 2026-10-06 | Read; secondary | none |
| 5 | Safety-stock arithmetic | Our computation | 2026-10-06 | Illustrative, script-checked | none |