Earnings Surprise: Why a Beat Tells You Little

Becton Dickinson began its 2025 financial year telling investors to expect organic revenue growth of 4% to 4.5%. By May, the guide had been lowered to 3% to 3.5%. The year finished at 2.9%.

Depending on which bar you happened to keep, that is either a modest shortfall against a recently reduced target or a company that delivered barely two-thirds of what it originally said it would. Both descriptions are accurate, both are defensible, and only one of them is about the business — which is, unfortunately, not the one most earnings-surprise screens are measuring.

It is worth admitting how long the obvious version of this survives before it breaks. Consensus surprise looks like a signal, gets reported like a signal, and moves prices like a signal. It took a great deal of staring at guidance histories to accept the simpler explanation: the bar that a beat is measured against is set by people whose job depends on being close to the number, and adjusted right up until the day before the print. A structurally high beat rate is the expected output of that arrangement rather than evidence about the businesses inside it. What follows is a lesson in what to compare instead, not a view on any company named.

Who sets the bar you are measuring against?

An earnings surprise is the difference between reported EPS and consensus, the average of analyst forecasts. Tucked inside that definition is an assumption almost nobody examines: that consensus is an independent estimate.

It is not quite that. Analysts revise toward company guidance as a quarter progresses, and companies have every incentive to ensure the number they will eventually report sits comfortably above wherever consensus settles. Charlie Munger's line applies with unusual precision — don't ask the barber whether you need a haircut. None of which makes analysts dishonest; it makes the resulting bar a poor instrument for the question most people think they are asking.

There is a second bar available, and it goes almost entirely unscored. The company's own published guidance is specific, dated, attributed to a named executive, and sits in the transcript where anybody can find it. It is also, unlike consensus, the bar management is actually managing to.

How often do companies clear their own guidance?

Across the 37,529 forward commitments tracked from US earnings calls on Inve, 18,478 have reached a definitive outcome, and 89.4% of those were delivered as stated. Split by what was being guided, the pattern gets sharper.

Read the top row and the bottom row against each other, because the ordering is the opposite of what intuition suggests. A number cleared nineteen times out of twenty is not a forecast at all — it is a target being managed to, with all the levers for managing it (discretionary spend, timing, buybacks) sitting inside the company. The surprise was never that EPS guidance gets hit. The surprise would be a company that regularly failed to hit it.

And then capital expenditure, which ought to be the easy one. Capex is the number management controls most directly — nobody outside the firm decides what a company spends — and it is the guide they land least often. That is not incompetence. Capex is the shock absorber. When something else in the plan comes under pressure, the spending plan is the first thing to move, because it is the only line management can adjust without renegotiating with a customer, a union or a regulator. A capex guide that slips is very often the earliest visible sign that a company is protecting a number further up the page.

The bar moves, so record where it started

Guidance is not a fixed target, and it does not move symmetrically. In the tracked US record, favourable revisions outnumber unfavourable ones by more than three to one — 1,858 against 567 — which is precisely what you would expect from people who set their opening number somewhere it can comfortably be raised from.

Which brings us back to Becton Dickinson, and three numbers from one financial year. The opening guide, on the November 2024 call, was organic revenue growth of 4% to 4.5%. By the May 2025 call it had become 3% to 3.5%. The actual was 2.9%. Against the revised bar that is a small miss and a forgettable one. Against the bar the year opened with, the company delivered about seven-tenths of what it set out to do. If your notes were overwritten in May, the second sentence is invisible to you — and the second sentence is the one that describes the year.

The fix is a single habit, and nothing else in this article is as useful: when a guide is revised, keep the old row. Add the new one underneath it.

What happens when the bar disappears entirely?

In April 2025, Cummins withdrew its full-year outlook. The words were: "Absent more clarity about the likely duration of elevated tariffs, we are not able to provide a reliable forecast for the remainder of this year."

A company with no guidance cannot disappoint against it. The segment targets set the previous November — Engine revenue down 2% to up 3%, Components down 5% to flat — stopped being live commitments the moment the outlook was pulled, and were never reinstated with a number. Both segments finished outside those original ranges, and no call ever framed that as a miss, because by then there was officially nothing to miss.

Note carefully what this is and is not. Withdrawing guidance in the face of a genuine tariff shock is defensible, and arguably more honest than publishing a number nobody in the building believes. But it also resets the scoreboard, and a research process that compares actuals only against live guidance will quietly record no result at all for that year. Keep the withdrawn rows too, and score them yourself.

The contrast worth holding beside it is W.W. Grainger, which told investors to expect volume outgrowth in its High-Touch Solutions business at "the low end of the 400 to 500 basis point range" and delivered about 250 basis points — and said so on the call, describing the result as short of the long-term target in its own words. That is a proper miss: a specific, numeric, publicly stated target, compared with the outturn, acknowledged out loud. It carries more information than a dozen consensus beats, because the bar was set by the people who knew the business best and they still came in under it.

A five-minute results-season routine

Before the print, write down the original guide for the year, the current guide, and the date each was set. When the numbers land, compare the actual against both — the gap between the two guides is management's own revision to its view, and the gap to the original is the year's real story. Ignore the beat-versus-consensus headline unless the magnitude is genuinely large, because it mostly tells you where analysts happened to be standing. Check the capex line on its own, and if capital spending guidance slipped while the earnings guide held, ask what was being protected. Then read the Q&A for the reason, because a miss with a named, checkable cause is a different object from a miss with an adjective attached to it.

How Inve keeps the bar from moving

The habit this article turns on — when a guide is revised, keep the old row — is the one a human reader cannot enforce on themselves, because the revision arrives as news and the original is a note you took nine months ago.

Inve keeps both. When management moves a number, the original commitment stays in the ledger with its birth quarter and the wording it was first given in, and the revision is recorded against it rather than replacing it. Becton Dickinson's year reads as 4%–4.5%, then 3%–3.5%, then 2.9% — three numbers in one place — instead of a small miss against whatever the bar had quietly become by May. When a company withdraws guidance outright, as Cummins did, the withdrawn targets stay on the record too, so the year does not simply score blank.

Across 885 US companies that is 37,529 tracked commitments, 18,478 of them resolved — which is where the figures in this article come from: the 94.8% EPS hit rate and the 78.5% capex rate are read off those resolved commitments, not estimated.

The call summaries carry the forward half. Each one closes with a watchlist naming the metrics that will decide the next quarter — current value, target, the quote the target came from, why it matters, and the guidance row that will settle it. For the capex-as-shock-absorber pattern above, that is the difference between noticing a slip a year later and knowing in advance which line to watch.

Free for US readers, Pro surfaces included, while the US market has no billing rail.

Where this approach can mislead you

The strongest counter-case is that guidance-versus-actual rewards sandbagging, and consensus surprise at least has an adversarial party in it. A management that guides deliberately low will score full marks on the method above while telling investors almost nothing at all. Analysts, whatever their incentives, are the only participants with a reason to publish a number the company did not choose, and discarding them entirely would be a mistake. The argument here is narrower than that: a beat against a bar the company effectively set is close to uninformative, and the interesting events are the misses, the revisions and the withdrawals.

The figures come with their own caveat. A 94.8% EPS hit rate is measured on commitments that could be resolved from the calls themselves, which biases toward guidance companies reconcile aloud — and companies are more inclined to revisit a number they hit. Read it as "of the guides that got scored," not "of all guidance ever given." The US record is also about two years deep, which is enough to describe current behaviour and not enough to judge a management.

And "delivered as stated" is not the same as delivered well. A company can hit an EPS guide by cutting the spending that would have earned next year's revenue, and no ledger can see that. The transcript sometimes can.

Frequently asked questions

The owner's question

A quarter's beat or miss changes very little for somebody who intends to hold a business for five years. What changes a great deal is the answer to this: when this management publishes a number and the world moves against it, do they cut early and explain why, quietly re-baseline until the shortfall disappears, or withdraw the target altogether? All three happened in the examples above, in the same two years, to companies of comparable size and quality. Only one of them is visible on a screen of earnings surprises.

Related reading

• How to Track Management Guidance Across Quarters
• How to Read an Earnings Call Transcript
• Sector Earnings Trends: How to Read Results Season