Micron's page on this site renders R40: 219 · 2026 Q2. That is correct: 219 reconciles exactly to the quarter ended 2026-02-26, and the pill says which quarter it came from.
The Micron article published here the same day is about the quarter ended 2026-05-28. That is also correct, and it says so in its first paragraph.
Both surfaces are labelled. Neither is wrong. And nothing anywhere tells a reader that the "2026 Q2" on the score ended in February while the article beside it is about a quarter that ended in May. That seam — between two correctly labelled things — is what this piece is about, and it is wider than the two companies where we have already reported it.
We have said this twice, on one company each
This is not a discovery. It is a count.
On August 1 we published the Palantir score being five quarters stale: the site scores PLTR 99 from the quarter ended 2024-12-31 while its revenue series runs to 2026-03-31. That piece named the mechanism — the free-cash-flow series stops, so the score falls back to the newest quarter where both halves of the formula exist — and counted eight other tickers with the same shape.
Today we published the Micron follow-up, which put a size on the gap. It cites a stored score of 219 for the quarter ended 2026-02-26 against 388 derived from the filed quarter ended 2026-05-28 — a Rule of 40 the piece then argues is meaningless for a cyclical manufacturer, which is a separate point. That derivation is XBRL-verified there, with the provenance and the CIK, and it is not repeated here; one worked case is enough and re-deriving it would only risk breaking a figure the site already has right.
What is new is neither the observation nor the example. It is the population.
Two different mechanisms, almost no overlap
There are two ways a score on this site can be older than the coverage next to it, and they are unrelated.
Mechanism one — the cash half stops first. The Rule of 40 needs revenue growth and free-cash-flow margin. Where the free-cash-flow series ends before the revenue series, the score is computed from the newest quarter that has both. This is the Palantir case. Nine companies, at HEAD:
| ticker | revenue runs to | cash-flow series stops at |
|---|---|---|
| PLTR | 2026-03-31 | 2024-12-31 |
| AMZN | 2026-06-30 | 2025-09-30 |
| GS | 2026-06-30 | 2025-12-31 |
| JPM | 2026-06-30 | 2025-12-31 |
| XOM | 2025-12-31 | 2025-09-30 |
| BAC | 2026-06-30 | 2026-03-31 |
| QCOM | 2026-06-28 | 2026-03-29 |
| SCHW | 2026-06-30 | 2026-03-31 |
| SOFI | 2026-06-30 | 2026-03-31 |
Mechanism two — the whole series stops before the filings. Nothing internal is inconsistent here; the data simply ends before the company's most recent filed quarter. Fourteen companies, comparing the newest period end in the revenue series against the newest 10-Q, 10-K or 20-F filing date we hold:
| ticker | score computed from | filings run through | gap |
|---|---|---|---|
| AFRM | 2025-09-30 | 2026-03-31 | 182 days |
| NBIS | 2025-09-30 | 2025-12-31 | 92 |
| CVNA | 2026-03-31 | 2026-06-30 | 91 |
| INTC | 2026-03-28 | 2026-06-27 | 91 |
| MSFT | 2026-03-31 | 2026-06-30 | 91 |
| MU | 2026-02-26 | 2026-05-28 | 91 |
| NOW | 2026-03-31 | 2026-06-30 | 91 |
| PYPL | 2026-03-31 | 2026-06-30 | 91 |
| TSLA | 2026-03-31 | 2026-06-30 | 91 |
| V | 2026-03-31 | 2026-06-30 | 91 |
| VRT | 2026-03-31 | 2026-06-30 | 91 |
| COIN | 2025-12-31 | 2026-03-31 | 90 |
| XOM | 2025-12-31 | 2026-03-31 | 90 |
| JNJ | 2026-03-31 | 2026-06-28 | 89 |
The two lists share exactly one name. XOM is on both. Nine plus fourteen minus that one overlap is 22 of the 66 companies in the corpus whose displayed score comes from an older quarter than the newest data about them — for two causes that have nothing to do with each other and would need two different fixes.
Period ends are given rather than quarter labels throughout, because the labels are ambiguous exactly where it matters. MU, V and MSFT are fiscal-shifted: Micron's "2026 Q2" ended 2026-02-26, and Visa's most recent published analysis covers its fiscal Q3.
The screen has a clean threshold, and one row is not a real gap
The second table excludes AVGO, which the mechanical screen returns. Its stored period end is 2026-04-30 against a filed period end of 2026-05-03: a three-day date-convention difference, not a missing quarter.
That exclusion is not a judgement call. AVGO's gap is 3 days; the next-smallest genuine gap is JNJ's at 89, and the rest cluster at 90–92 with one at 182. There is nothing between 3 and 89, so the screen separates cleanly into "the stamp disagrees with the filing" and "a quarter is missing".
One further caveat on that table. NBIS files a 20-F — it is a foreign private issuer, so its filings are annual and its quarterly figures come from releases rather than filings. Its 92-day gap is measured against an annual filing, which makes it the weakest row on the list even though the underlying lag is real.
Eleven of the fourteen sit under a recent article
For eleven of those companies we published something in the last six weeks. Ten of the eleven discuss a quarter later than the one the score is computed from:
Microsoft and Tesla on the quarter ended 2026-06-30, Intel on 2026-06-27, Carvana, ServiceNow, Vertiv, PayPal and Visa on their June quarters, Coinbase on a June quarter our filing index does not yet hold, and Micron on 2026-05-28.
The eleventh is NBIS, and it does not support the claim: that piece is a drawdown comparison rather than an earnings analysis, and it discusses no quarter at all. Saying so is cheaper than letting a reader find it.
The remaining three — AFRM, JNJ and XOM — are behind with no recent article. AFRM is the largest gap on the site and has no coverage at all.
Stale is not wrong
Every score in both tables is correctly computed from the data present, and correctly labelled with the quarter that data came from. None of them is a miscalculation, and none of the eleven articles is inaccurate about its own quarter. A reader who checks any single number will find it reconciles.
The failure is a join. Two labels, both true, sitting on the same company, with nothing on screen saying they refer to different periods — and a composite score is exactly the kind of figure a reader treats as current unless told otherwise.
The measure inherits the stamps
One last thing, because it bears on how much any table like this should be trusted, including this one.
The screen above reads period-end stamps. It cannot see a defect in those stamps — and this corpus has had one. AFRM's series was stamped a quarter early: every point carried the value of the quarter ending three months after its recorded date. Repairing that moved AFRM's newest stamp from 2025-06-30 to 2025-09-30 while its newest value stayed at $933,340,000, unchanged to the dollar.
The same screen therefore returned 274 days for AFRM before the repair and 182 days after it, with no new data of any kind. The lag was always 182; the measure was wrong about how wrong things were.
That is why the tables here carry the date they were computed on: 2026-08-02, at HEAD. The backfill for mechanism two is tracked as a data card and its sequencing is not settled, so no date is promised for it here. When it does land, it will move a dozen scores at once — and it will make the second table above obsolete, which is the ordinary fate of a piece that documents a data defect and the reason to date one rather than to leave a reader guessing whether they are looking at drift or at an error.
The two populations are computed on 2026-08-02 across the 66 companies we track. Mechanism one compares the newest period end in the free-cash-flow series against the newest in the revenue series; mechanism two compares the newest period end in the revenue series against the newest filing date on a 10-Q, 10-K or 20-F in our SEC filing index, counting gaps of 30 days or more. Micron's 219 and 388, the XBRL provenance behind them and the CIK are cited from our Micron piece and are not re-derived here; the nine-ticker cash-flow population and its mechanism are from our Palantir piece, re-run for this article. The AFRM stamp shift and its repair are recorded in this repository's history, where the before-and-after values are unchanged. Article titles are linked rather than restated; where a title carries a share-price move, that figure belongs to the linked piece and is not repeated as fact here. No price, share-move or market-capitalisation figure appears above — the corpus stores no previous close for any company.