Cheapest advertised 10×10 — median across stores
Median web rate by unit size
Advertised inventory signal — last 30 days
Rate-change activity — last 14 days
Median cheapest 10×10 — last 30 days
Promotion coverage
Web rate by size, brand by brand
The same table
Discount depth — web rate vs. street rate
State by state
Independent operator pilot
Median cheapest 10×10 per store
Median store price (all sizes)
Advertised units available
Stores reporting
Public Storage history reaches back to April 2026 from the legacy series; the other brands start the day their first two snapshots were recorded.
Biggest web-rate moves — last 30 days
When “40% off” did not mean 40% cheaper
Daily, exact-listing observations show Public Storage changing both sides of an offer: the promotion and the reference rate used to calculate the advertised savings. July supplies the control, August contains the exceptional event, and September shows the structure being unwound.
One pattern, tested across three months
A discount can grow because the customer pays less—or because the number crossed out above it grows. Daily snapshots let us separate those two mechanisms on the same SKU at the same facility.
Promotions routinely rotated without rate changes. On July 28, 428 listings entered the same four-month offer and none received a price increase.
On August 1, every one of 4,904 listings entering “40% off For 4 Month” also received a higher reference rate. The median rise was $45.
On September 14, 35,235 promotions changed while 30,762 matching rates fell. Promotional value weakened enough to offset the cheaper headline rates.
The control that makes August meaningful
Public Storage’s publishing system can write prices and promotions independently. Promotion-only changes occur throughout the history, including a clean comparison just four days before the August event.
July 28
August 1
This rules out an unavoidable coupled-write artifact. It establishes an unusual pricing decision, not why that decision was made.
One advertisement, reconstructed
On August 24, a Costa Mesa location displayed “Month 1–4 40% OFF,” the full monthly rate struck through, “Months 5–12 In-Store Rent,” and a “Total Estimated 12-Month Savings.” The operator’s savings arithmetic was exactly four months at 40% off the displayed reference rate.
August 1, measured on the 4,904-offer event cohort: headline value rose 35.3% and promotional value rose 216.4%, producing a 2.1% reduction in aggregate modeled four-month cost. That aggregate masks the distribution: the median offer cost $2 more, 55.8% of listings cost more over four months, and 96% cost more by month eight.
| Unit | Reference | Months 1–4 | Claimed savings |
|---|---|---|---|
| 5×5 | $110 | $66 | $176 |
| 5×10 | $165 | $99 | $264 |
| 5×15 | $237 | $142 | $380 |
The $237 listing was observed at $161 on August 20, $237 on August 22–24, and $161 again on August 25. Against the adjacent $161 rate, the $142 promotional payment represents about $75 across four months—not $380. Roughly $304 of the displayed savings came from the temporarily higher reference rate.
September exposes the trade
On September 14, Public Storage cut headline rates but simultaneously weakened or removed promotions. Across 59,045 matched, modelable offers:
| Four-month accounting | Before | After | Change |
|---|---|---|---|
| Headline value before promotions | $32.78M | $27.96M | −14.7% |
| Promotional discount supplied | $10.20M | $5.13M | −49.7% |
| Modeled customer cost | $22.58M | $22.83M | +1.1% |
Lower monthly rates reduced modeled cost by 15.4%. Weaker promotions added 16.5%. The two effects nearly cancel, leaving the four-month customer cost slightly higher.
This is an equal-weight analytical portfolio: one hypothetical rental per matched offer. It measures advertised terms, not transactions, and excludes fees, insurance, tax, and later tenant-rate changes.
Living ledger: the 96-hour walkback
A static paper published on September 14 would have concluded that Public Storage locked in a 25% lower price regime entering Q4. The daily pipeline stayed running and caught the walkback:
| Date | State | Matched Index | Observed Mechanics |
|---|---|---|---|
| 13 Sept | Baseline | 1.0000 | Pre-reset steady state |
| 14 Sept | Flash reset | 0.7667 | Full matched panel (0.7500 within cut cohort of 27,213 SKUs); promo values halved |
| 15 Sept | Predictions sealed | 0.7667 | Git commit 2026-09-15-q3-prediction-and-ledger-gap.md |
| 18 Sept | Walkback | 0.9800 | +12% average price restoration (panel mean; 22,772 SKUs repriced: 18,828 up, 3,944 down); 17,935 promos rewritten |
By September 18, the matched panel index was 0.9800—just 2% below starting rates. A 25% price reset was half-unwound within 96 hours.
The pre-registered predictions (sealed 15 Sept 2026)
All three hypotheses were committed into git history prior to Q3 earnings release:
Data hygiene & provenance: 78 of 92 days observed (84.8%)
Collection began 9 July 2026. 78 of 92 days in Q3 were observed (84.8% of the quarter; 92.9% of the 84-day collection window). 8 days precede the start of collection, 5 were a deliberate pause (26–30 August) to review changed site terms, and 1 is an unexplained collection gap. Missing days are preserved as missing; nothing is interpolated.
What the evidence supports
The study can establish the advertisement, its reference rate, its promotional terms, and the history of that exact listing. It does not observe completed transactions, establish intent, or make a legal determination about deceptive advertising.
Method: immutable daily snapshots; exact facility-and-SKU matching; independent price and promotion event logs; four-month offer terms modeled from the published advertising; missing observations preserved as missing rather than interpolated.