How to Track Management Guidance Across Quarters

Here is something that took far too long to learn, and the lesson was not the one it looked like. In April 2025, on Centene's first-quarter call, chief executive Sarah London told analysts: "Our full year 2025 adjusted EPS expectations remain unchanged at greater than $7.25." The quarter before, the finance chief had called that figure a guidance floor. Three months later, on the July call, the same chief executive said: "Our current forecast calls for full year adjusted diluted EPS of approximately $1.75." The floor had fallen about seventy-six percent in a single step.

The obvious reading is that management misled investors, and that reading does not survive the transcripts. Centene disclosed the cause in detail, took every analyst question on it directly, and named the mechanism — a Marketplace risk-adjustment shortfall and higher Medicaid medical costs. What failed was not disclosure. What failed was that a commitment made in February and repeated in April had no place to live in anybody's notes, so when it broke in July, most people compared the new number against last quarter's result rather than against the commitment it replaced. The company did the hard part. The reader did not do the easy part.

That is the entire case for tracking guidance across quarters, and everything below is how to do it. It is a lesson in process, not a view on any company named in it.

Why one call cannot show you this

An earnings call is a snapshot with an unusual property: it contains claims about the future that will be graded later, by a document nobody ever puts beside it. Read the July call on its own and you learn that Centene expects $1.75. Read it against April and you learn something categorically different — that this management's confidence in a full-year figure, expressed twice, was worth about a quarter of its stated value. Neither reading is wrong. Only one of them is about management rather than about a quarter.

The mechanics of fixing this are unglamorous. You need somewhere to put a number when it is made, so that you have something to check it against when it comes due. That is the whole innovation, and it is why so few people do it: there is nothing clever about it, only the discipline of reopening the file.

What a guidance ledger actually holds

Six fields per commitment. That is the system in its entirety.

The reason it works has nothing to do with the sophistication of the fields. It is the status column, which forces a revisit of every open commitment every quarter — which is precisely the thing nobody manages to do from memory, and the thing that quietly stops happening around March.

The four things that can happen to a commitment

Once the ledger exists, every open row eventually resolves into one of four states, and they are not equally informative.

The first is that it gets delivered, the number arrives and management says so. Useful, and the least interesting of the four. The second is that it is missed and named — the number arrives short and management says that too. Centene's July cut belongs here, and it is better behaviour than most investors credit, because an early, explained revision hands you a checkpoint. The third is that it is revised with a reason attached, in which case the reason is the information and the direction is not. The fourth is that it is never scored at all: the period ends, and the figure is simply never mentioned again.

Most people assume the fourth dominates. Across the tracked US record it does not, and what happens instead is considerably more useful to know about.

Which guided numbers are actually scoreable?

Across the 37,529 forward commitments tracked from US earnings calls on Inve, covering 885 companies, outright ghosting — a commitment that came due, had a natural occasion to be addressed, and was not — accounts for about one percent of the ledger, though 226 of those 885 companies carry at least one.

The bigger practical problem is quieter and has nothing to do with evasion. A large share of what management guides is never reconciled out loud on any call, not because anybody is hiding it but because the metric is simply not one that companies are in the habit of scoring. And the rate varies enormously by what was guided:

A tax-rate guide is roughly three and a half times more likely to vanish unscored than a revenue guide, which is not a scandal in the slightest. A company guides an effective tax rate every quarter, reports the actual in the 10-Q where it belongs, and never has occasion to stand up and say "we told you 23%, it came in at 24%." But it means something quite specific for anybody building a research process, which is that a large fraction of a naively-built ledger will consist of rows that were never going to resolve. S&P Global guided roughly $5.2 billion of full-year adjusted free cash flow on a 2024 call, and no subsequent call reconciles that figure against the outturn — nothing improper, simply a number that does not get closed out on a call. If you wrote it down expecting a verdict, you were going to be disappointed by convention rather than by management.

The rule that falls out is to weight the ledger toward the metrics companies score themselves on: revenue, EPS, margin, segment growth, and anything management has publicly attached its own credibility to. Track the tax rate if you like. Just don't read its silence as a signal.

How do you score a recovery honestly?

This is where guidance tracking usually goes soft, so take Centene forward from the wreck.

Four consecutive raises. On the face of it, a management that has rebuilt its forecasting and is now beating its own guides — the mirror image of 2025, and a story that writes itself. Now the uncomfortable observation, and the reason to keep a ledger rather than a narrative: the sentence construction has not changed at all. "Greater than $3.00" is the same shape as "greater than $7.25" — a floor, phrased so that any outcome above it counts as delivery. What broke in 2025 was never the arithmetic. It was that a floor was being read as a forecast by everyone involved, including, on the evidence, the people who set it.

So the honest reading is neither "credibility restored" nor "same trick twice." It is that this management sets floors, has raised them four times running, and last shattered one about eighteen months ago. That is a description you can actually act on, and two quarters of data would never have produced it.

A quarterly routine that survives contact with real life

Per company, per quarter, about fifteen minutes once the ledger exists. Open last quarter's rows before you open the new transcript, so that you read your own notes first and the call cannot quietly reframe them for you. Find this quarter's status for each open row — reaffirmed, revised, resolved or absent. Where something is absent, work out whether it has been ghosted or is simply not the kind of number this company ever reconciles aloud, because that distinction decides whether you have found a signal or a convention. Add the new commitments with metric, number, period and speaker. Then write one line: what got harder, what got easier, and which single row you will be watching next quarter.

Doing this for one company is easy. Doing it for fifteen, every quarter, without fail, is the part that defeats almost everybody — which is the job Promise Tracker exists to absorb, and the reason the ledger above is the shape it keeps.

How Inve keeps the ledger

The six columns above are the whole method, and a spreadsheet does the job — for two or three companies, if you reopen it every quarter without fail. Almost nobody does. The ledger dies in March, and it dies quietly, which is the worst way for a research process to fail: you keep the belief you formed in January and lose the record that would have corrected it.

Promise Tracker is that ledger, kept for you. Across 885 US companies it holds 37,529 tracked commitments — each with the metric, the original wording as management phrased it, the quarter it was born, and what became of it: delivered, missed, revised, or never scored at all. Centene's $7.25 floor, its reaffirmation in April, the cut to $1.75 in July and the four raises since are one screen, not an afternoon with seven transcripts.

The friction ledger sits alongside it, recording the topics a management keeps stepping around and for how many consecutive quarters — the pattern that is invisible in any single call and damning across six.

All of it is free for US readers while the US market has no billing rail.

Where this approach can mislead you

The strongest case against it is that a guidance ledger measures forecasting, and forecasting is not the same thing as running a business well. A management that guides conservatively and clears the bar every quarter will score beautifully on this system while doing nothing more impressive than setting a low bar. Meanwhile a team guiding honestly into genuine uncertainty — a health insurer facing a policy shift, a cyclical facing a turn — will accumulate misses while making better decisions than the sandbagger. Rank managements by this alone and you will systematically prefer the cautious to the candid, which is not what anybody set out to do.

There is also a limit on the numbers themselves. "Never reconciled aloud" is a statement about a corpus of calls, not about disclosure: these figures are reported in filings, and the only thing being observed is whether they get closed out on a call. Some companies publish prepared remarks as a separate document that is not held in that corpus, which mechanically inflates their silent count. And the US record runs about two years deep — enough to describe how a management communicates now, nowhere near enough to judge one, and it should not be presented as such.

Frequently asked questions

The owner's question

A five-year owner does not need management to be right about next quarter. What they need is to know what a sentence from this management is worth — whether "greater than $7.25" is a considered estimate or a placeholder, and whether a number that stops appearing was closed out somewhere or simply abandoned. You cannot learn that from a single call, and nobody is going to hand it to you. It accumulates one row at a time, in a document you keep and they don't.

Related reading

• How to Read an Earnings Call Transcript
• Concall Red Flags: A Checklist
• Sector Earnings Trends: How to Read Results Season