Case studies › ANET · WDC · PWR
Arista and the momentum cohort have one input left: order backlog
A screen that selects on four-quarter average earnings surprise and a positive 50-day moving average is selecting on two things that cannot be forecast from the same place: the past, and the trend in the price. Neither is order backlog.
The published screen cut 7,743 companies to 16. Three of the survivors are Arista Networks, Western Digital and Quanta Services, on average surprises of 10.6%, 10.5% and 21.5% and expected current-year growth of 40.9%, 138.4% and 62.7% respectively.
| Company | Ticker | Avg 4Q surprise | Expected growth | Momentum score |
|---|---|---|---|---|
| Arista Networks | ANET | +10.6% | +40.9% | B |
| Western Digital | WDC | +10.5% | +138.4% | B |
| Quanta Services | PWR | +21.5% | +62.7% | A |
What an average surprise actually measures
A four-quarter average beat of 10.6% is a statement about the relationship between a company and the consensus, not about the company. It can persist for one of two reasons: the business keeps handing over more than the analysts modelled, or the analysts keep modelling low.
Both produce the same screen output. Only the first survives a deceleration in end demand. In a period where the AI infrastructure order book is the single variable, the distinction is worth naming, because a beat manufactured by sandbagged guidance does not carry the multiple when the order book rolls.
The 138.4% expected growth at Western Digital is the extreme case. A number that large is a cycle turn in the estimate base and not a trend line. Pricing it needs the mid-cycle unit volume and the mid-cycle price, and neither appears on a momentum screen.
Why the moving average is the wrong stress test
The screen requires a positive percentage change in the 50-day moving average and positive relative strength over four weeks. That is a trend filter on price, and price is the last thing to break.
A company with a falling order book can hold a rising 50-day average for roughly a quarter, because the orders signed ninety days earlier are still shipping. The overlap between the moving average and the bookings trend is precisely one quarter wide, which is the same window in which the trend filter looks most convincing.
That sets the useful test. Not "is the price above the average" but "was the backlog signed at a price and volume the next quarter can repeat". For Arista that is the switching order book and the campus line. For Quanta it is the utility transmission award calendar, which is the physical precondition for the data centre load growth the whole cohort is levered to.
Where the case breaks
In any of the three, if the next print delivers the beat from a backlog that was signed two quarters earlier while new bookings flatten, the screen keeps the stock and the arithmetic does not. The 50-day average will not warn first; the bookings line will.
Quanta is the cleanest and also the most exposed. Its work is transmission and substation build, so a slower permitting calendar at state level is a direct order-book event rather than a sentiment one.
What the arithmetic says
The cohort is not wrong. It is unverified. Three historically reliable beaters, a trend filter that lags the input that matters, and no backlog figure in the selection. Run the bookings first and the moving average second, and the screen becomes a forecast instead of a receipt.
A four-quarter average beats record sits on top of an order book that can only be as good as the orders still open. The moving average knows the last quarter. The bookings line knows the next one.
Sources
Surprise averages, expected growth rates, momentum scores and the screen criteria are as published by the screening authority. The quarter-shift between price trend and bookings conversion is a derived observation on the stated criteria. No company financial is restated here.
Published for information only. Nothing here is investment advice or a recommendation. This note is not prepared in accordance with the legal requirements designed to promote the independence of investment research. No position is held in the securities covered.