nike-allocation-analysis
How inventory moved through Nike’s US stores
I logged 886,825 stock changes across 255 Nike stores. Factory, Clearance, and full-price stores get restocked in very different ways.
I built Flow Monitors, a stock-alert service. To send alerts, it polls Nike’s in-store availability for every size of every shoe it tracks, at every store it knows about. After a few months, that becomes a detailed log of how stock moves through Nike’s stores.
Nike’s fiscal 2026 10-K counts 212 US Factory stores and 75 US in-line stores, but it reports them together with the apps as NIKE Direct. It doesn’t say how inventory moves through each kind of store. In its fiscal fourth quarter (March–May), NIKE Direct revenue fell 7% worldwide while company inventory held at $7.5 billion. So I wanted to know how restocks differ between outlet and full-price stores, and whether those differences hold up when both carry the same shoes.
The short version:
- Factory stores get bigger restocks and refill slightly faster than they sell down.
- Full-price stores get thinner restocks and slowly drain.
- Clearance stores have the most broken size runs, with two sizes per shoe at the median.
- At first the data seemed to show full-price stores selling faster. That came from a biased sample.
The data
Each record is one change for one store, shoe, and size, with a timestamp and the old and new stock label. Nike reports four labels (OOS, LOW, MEDIUM, HIGH) and never says how many pairs each one means.
So I use two terms. A restock is a move from OOS to any in-stock label. A drop is any move to a lower label. A drop might be a sale, but it could also be a transfer, a recount, or theft, so I never treat drops as sales.
| Measure | Coverage |
|---|---|
| inventory level changes | 886,825 |
| Nike-owned US stores | 255 |
| SKUs with recorded changes (1,254 monitored) | 1,089 |
| tracked product-size IDs | 13,024 |
Flow Monitors found these stores by searching Nike’s store locator from 36 ZIP codes, so this is a sample, not every Nike store in the US.
| Store format | Stores | What it is |
|---|---|---|
| Factory | 145 | Outlet stores |
| Clearance | 29 | Deeper-discount, marked-down stock |
| Full-price Nike Store | 74 | Standard in-line stores |
| Community Store | 3 | Grouped with full-price in the charts |
| Company Store | 2 | Employee stores, shown separately |
| House of Innovation | 1 | NYC flagship, shown separately |
| Jordan World of Flight | 1 | Philadelphia flagship, shown separately |
The 886,825 total is every change recorded from March 9 to August 20. The store-format charts cover April 15–August 20. The weekly, random-sample, and population analyses start April 29, after a two-week stretch when the monitor was still discovering stores.
Factory stores get the deep restocks
I started by comparing restocks with drops. Factory stores logged 108 restocks for every 100 drops, and full-price stores logged 84. Factory stock refills a little faster than it leaves, while full-price stock slowly runs down.
The restocks themselves look different too. At Factory stores, 41.4% of restocks jump straight from OOS to MEDIUM or HIGH, compared with 15.5% at full-price stores. Outlets also log 20% of their restocks between 7 and 10am Eastern, against 6% for full-price.
Restocks per drop, by store format
April 15–August 20, 2026. A value of 1.08 means 108 restocks for every 100 drops.
Share of restocks that land above LOW
Restocks that go straight from OOS to MEDIUM or HIGH. 400,483 restocks across the three plotted formats, April 15–August 20.
When restocks appear, by hour of day (Eastern)
Share of each group's restocks in each hour, converted to Eastern time. 400,475 restocks with usable timestamps.
Clearance has the most broken size runs
Using the August 20 snapshot, I counted how many sizes each store had in stock for each shoe. Across everything each store carries, full-price looked best, with a median of seven sizes, against four at Factory and two at Clearance.
That isn’t a fair comparison, because the stores carry different shoes. So I limited it to the 183 exact SKUs carried by both outlets and full-price stores. Full-price fell to four sizes, the same as Factory. Clearance stayed at two, and 58.4% of Clearance listings had only one or two sizes.
Median sizes in stock per shoe, per store
Gray: every shoe each group carries. Red: only the 183 shoes carried by both outlets and full-price stores.
If I were at Nike, I’d test pooling Clearance stock into fewer stores to see whether the size runs fill in.
The result I had to take back
My first look at sell-through used the twelve busiest SKUs. The median time from a restock to the next drop was 145 hours at full-price stores and 413 at outlets. Full-price stores looked almost three times faster.
Then I reran it on two random samples of 30 SKUs, and it flipped both times. Full-price held steady at 363–370 hours while the outlet number moved around a lot. The busiest-twelve list was full of older Air Max, Free RN, and Shox styles that full-price stores barely carried, and the few sizes they did carry changed quickly.
So the data doesn’t show full-price stores selling faster. What it did show me is how much the choice of SKUs can move a result.
Median hours from a restock to the next drop
Same method each time; only the set of shoes changes. Restocks with no drop before August 20 stay in the estimate.
Restocks outpaced drops in June
Restocks averaged 1.87 per drop in June and 0.57 in August. Nike’s fiscal year ends May 31, so June might reflect a new-year product reset, or it might be normal seasonality. With one year of data I can’t tell which.
Restocks per drop, week by week
Values above 1.0 mean more restocks than drops; values below 1.0 mean more drops.
Population didn’t predict how fast stock moved
I also wanted to know whether stores in more populated ZIP codes sell down faster. I matched each store’s ZIP to its Census population and took 57 random SKUs. Then I checked whether each restock saw a drop within seven days, across 38,658 restocks at 242 stores.
After removing each SKU’s and store format’s average, the correlation with population was essentially zero: 0.005 for outlets (p = 0.95) and −0.09 for full-price (p = 0.46).
Share of restocks followed by a drop within seven days, by ZIP population
Stores split into population quartiles within each store format, using 2024 ACS five-year ZCTA population. Weighted by restock windows.
New York shows why. In the same sample, 36% of restocks at Nike Soho saw a drop within a week, compared with 29% at Atlantic Center, 46% at the Harlem Factory store, and 26% at the North Bronx Factory store. The city doesn’t behave like one market.
Census population counts residents, not tourists or commuters, so it’s a weak stand-in for foot traffic, especially at destination outlets.
What the labels hide
Four labels throw away a lot of detail. Flow Monitors also tracked a retailer that reports exact unit counts (for Pokémon cards, not shoes). I used it to see what gets lost when exact counts are squeezed into buckets like Nike’s.
Across 11,430 changes, one visible drop stood for a median of 1 to 22 units, depending on the labels. And with the label cutoffs I assumed, 40.2% of changes never crossed a label at all, so data like Nike’s would never show them. That can’t tell me Nike’s actual quantities, but it shows how coarse this data is.
How many units one visible drop can hide
A trading-card feed with exact counts, bucketed into LOW, MEDIUM, and HIGH like Nike's labels. 11,430 verified changes. This doesn't estimate Nike's units.
Methods and caveats
- Broad labels: a drop records direction, not quantity.
- Selected SKUs: the tracked shoes are an initial import plus three users’ picks, not Nike’s full catalog.
- No opening balance: the event history agrees with full-inventory snapshots (about twice a day from late July, 216 stores) on 75.2% of cells visible in either source.
- Store sample: the 36-ZIP locator search found 255 stores, 68% of them outlets. It’s a sample, not a census, and its format counts don’t map exactly onto the 10-K’s (it has 81 non-outlet stores; the 10-K lists 75 US in-line stores).
- Start date: starting the store-format charts on April 29 instead of April 15 gives 1.11 restocks per drop at Factory and 0.90 at full-price, with 41.7% and 15.9% landing above
LOW. Same direction. - Time zones: restock hours are recorded in Eastern time across stores in several time zones, so they don’t show whether stock arrived before opening.
- Sell-through: Kaplan–Meier medians, which keep restocks with no later drop before August 20. Random samples A and B are 30 SKUs each (seeds 7 and 13, three overlap), drawn from SKUs outside the top 12 with at least 200 events. No confidence intervals.
- Population: 2024 ACS five-year ZCTA population (
B01003_001E), matched for 247 of 249 store ZIPs. Stores needed at least 20 restock windows. Spearman correlation with 4,000 store-level permutations. - Label cutoffs: the unit-count comparison assumes where
LOW,MEDIUM, andHIGHstart. Other plausible cutoffs put the never-crossed share between 26% and 49%. - One year: the June change can’t be separated from seasonality.