Across five quarters, this site's P/E series for Johnson & Johnson reads 31.04 → 10.79 → … → 8.72 → 22.58. A mature pharmaceutical company appears to get two-thirds cheaper, sit there for a year, and then re-rate by 159%.
Almost none of that was the market. It was one quarter of earnings arriving in the trailing window, and then the same quarter leaving it.
That second move is the one worth knowing about. Everyone discounts a strange earnings print on the day it happens. Nobody is watching four quarters later, when a perfectly ordinary quarter appears to double a company's multiple — because in the quarter the chart moves, nothing happened at all.
We are the exhibit
Our own P/E explainer lists, under When Is It Lying (Misleading)?:
| One-time events | Non-recurring gains/losses distort EPS (e.g., asset sales or write-offs) |
and closes: "And always ask: what's driving the E in P/E?"
We never asked. Our February piece on Palo Alto embeds that company's P/E series and reasons from it directly:
"$PANW trades at a P/E north of 130x — steep for any software company … but that premium multiple still needs sustained execution to justify."
That sentence is arithmetically true and this piece does not retract it. What it missed is how the multiple got there.
The Palo Alto arithmetic, in period ends
Palo Alto's stored P/E, by period end — its fiscal year ends in July, so its quarter labels run a different calendar and are not used anywhere below:
| period end | 2023-10-31 | 2024-01-31 | 2024-04-30 | 2024-07-31 | 2024-10-31 | 2025-01-31 | 2025-04-30 | 2025-07-31 | 2025-10-31 |
|---|---|---|---|---|---|---|---|---|---|
| P/E | 123.04 | 46.88 | 38.35 | 40.79 | 43.12 | 96.44 | 99.97 | 102.23 | 133.44 |
The single largest step toward "north of 130x" is 43.12 → 96.44, and it is not a re-rating.
Diluted EPS for the quarter ended 2024-01-31 was $2.44, against 0.28, 0.39, 0.50 and 0.49 in the quarters around it. Twelve months later, on 2025-01-31, that quarter left the trailing window:
- trailing EPS 3.82 → 1.76 — a 2.17x fall
- P/E 43.12 → 96.44 — a 2.24x rise
The two ratios agree to within 3%. Whatever moved that multiple, it was the denominator.
And the quarter the chart moves on is unremarkable: Palo Alto's EPS for the quarter ended 2025-01-31 was $0.38, against $0.49 the quarter before. The doubling is entirely a number twelve months old walking out of the window.
Five companies, one signature
Every figure here comes from the database the site actually serves. Den × is the trailing-EPS fall; residual is what is left of the P/E move once that is accounted for — the part that could be price.
| spike quarter (period end) | that quarter's EPS | rolls out | P/E | P/E × | trailing EPS | den × | residual | |
|---|---|---|---|---|---|---|---|---|
| Johnson & Johnson | 2023-10-01 | $10.21 | 2024-09-29 | 8.72 → 22.58 | 2.59 | 15.06 → 5.96 | 2.53 | +2.5% |
| Palo Alto Networks | 2024-01-31 | $2.44 | 2025-01-31 | 43.12 → 96.44 | 2.24 | 3.82 → 1.76 | 2.17 | +3.0% |
| Salesforce | 2020-07-31 | $2.85 | 2021-07-31 | 47.61 → 95.99 | 2.02 | 4.77 → 2.48 | 1.92 | +4.8% |
| Block | 2024-12-31 | $3.05 | 2025-12-31 | 14.54 → 31.00 | 2.13 | 4.96 → 2.10 | 2.36 | −9.7% |
| Coinbase | 2024-12-31 | $4.68 | 2025-12-31 | 24.50 → 48.20 | 1.97 | 11.56 → 4.39 | 2.63 | −25.3% |
Block inverts the sign, which is the sharpest line available. Its multiple rose 113% across a quarter in which the residual price term is −9.7%. The stock got cheaper and the ratio more than doubled.
The effect runs in reverse when the quarter arrives, and it is just as large: Johnson & Johnson 31.04 → 10.79 (−65%), Uber 37.39 → 13.23 (−65%), Coinbase 45.60 → 22.10 (−52%), Block 37.71 → 18.64 (−51%). A company does not become half as expensive because one quarter's earnings were unusual.
We are not naming a cause for any of these spikes. This repository holds no earnings capture for Palo Alto, Johnson & Johnson, Salesforce, Block or Coinbase, and no filing index for Block or Uber. The claim here needs no cause: one quarter was four to twenty times its neighbours, and its exit from the window moved the chart.
Two cases where the price really did move
If every move were the denominator, this would be a single-cause story, and it is not.
Tesla is the clearest counter-example. Its quarter ended 2023-12-31 carried EPS of $2.26 against 0.78, 0.53, 0.41 and 0.40 around it. Rolling out on 2024-12-31, the P/E went 71.68 → 197.96 — a 2.76x rise, the largest in the corpus — against a trailing-EPS fall of only 1.81x. The residual is +52.7%. More than a third of that move was genuinely the market.
Coinbase fails the other way, and the table above shows it: a 1.97x P/E rise against a 2.63x denominator fall leaves a residual of −25.3%. The denominator more than explains the move, which means the price fell materially while the multiple nearly doubled.
So the honest statement is a decomposition, not an accusation. Sometimes the denominator is the whole story (Johnson & Johnson, Palo Alto, Salesforce, all within 5%). Sometimes it is most of it and the price moved too (Tesla, Coinbase). Knowing which requires looking.
What this is not
Three things, because each is a way this could be over-read.
- Not a data defect. The site's P/E series is behaving exactly as a trailing-twelve-month ratio must. Any company with one outsized quarter does this, on any chart anywhere. What breaks is the inference a reader draws from the shape, not the shape.
- Not a claim that any multiple is wrong. Palo Alto's P/E genuinely was north of 130x. The point is narrower and stranger: its change was not a re-rating.
- Not a valuation call. Nothing here says any of these companies is cheap or expensive. Decomposing a ratio's movement is not an opinion about its level, and no price figure appears in this piece — the P/E series is stored, not computed here, and this repository holds no price history to check a back-solved price against.
The habit
Before believing a P/E move, look at the quarter that just left the window, not the one that just arrived. If trailing earnings fell by roughly the factor the multiple rose, you are looking at arithmetic, not at a market opinion.
It takes one subtraction, and it would have caught all five of the companies above — including ours.
Every P/E and EPS figure above is read from the database this site serves, matched on period end; the values also appear in our stored data for Johnson & Johnson, Palo Alto, Salesforce, Block, Coinbase, Tesla and Uber. Quarters are identified throughout by period end rather than by quarter label, because several of these filers are on non-calendar fiscal years — Palo Alto's year ends in July, and Johnson & Johnson reports 52/53-week quarters whose ends fall on 2023-10-01 and 2024-09-29 rather than on month ends. Trailing EPS is the sum of the four stored diluted-EPS points ending at and before the stated period end. "Den ×" is the ratio of trailing EPS before to trailing EPS after; "residual" is the P/E ratio divided by that denominator ratio, expressed as a percentage — it is the part of the move not explained by the earnings change, and it is reported as a ratio rather than converted into a price, because the P/E series is imported rather than computed by this site and no price history exists here to check an inverted figure against. No cause is named for any earnings spike: this repository holds no earnings capture for any of these five companies, and no SEC filing index for Block or Uber, so naming one would be unsourced. Salesforce's stored 2020-04-30 P/E of 0 is excluded — trailing earnings were negative that quarter and the stored zero is a known open defect, not a real multiple. Uber appears only on the roll-in side; a second spike in its 2025-09-30 quarter means its window never fully clears, so its roll-out is not comparable to the five above. The two quoted passages are verbatim from our price-to-earnings explainer and this site's February Palo Alto analysis.