What Is an Earnings Call? Read the Q&A, Not the Script

An earnings call has two halves and only one of them carries information. The prepared remarks tell you what management wants to talk about. The Q&A shows what it would rather not. Real information sits in that second half, and you only see it properly once you put four consecutive quarters side by side.

What an earnings call actually is#

It is a conference call, usually held the day after the quarterly report lands. Management walks through the numbers, then hands the floor to analysts who ask about whatever those numbers left unexplained. Four times a year, once per quarter.

The surprising part is that nobody makes a company do this. The SEC requires the documents, from the annual 10-K to the current 8-K, but it does not require anyone to sit down at a microphone and explain them. Companies hold calls because investors expect them, and Regulation FD polices only one thing: if you are going to say it, say it to everyone at once rather than to three favoured analysts.

Since the call is voluntary, a company that quietly stops holding one has just told you something.

Every call has the same shape. The investor-relations officer opens by reading the safe-harbor notice on forward-looking statements, which in plain terms warns that everything optimistic you are about to hear might not happen. Then the CEO and CFO read a script covering revenue, margins and guidance for the next quarter. Analysts get the floor last, and from that point nobody is reading.

If you have ten minutes for a call, give nine of them to the Q&A.

What the prepared remarks are good for#

The script is written to survive a lawyer reading it and a reporter quoting it, so it rarely contains a sentence management did not want to say. That does not make it useless. The remarks tell you which number management chose to lead with and what word it used for a weak quarter, and both of those will matter to you three quarters from now.

Where it gets genuinely interesting is the seam between the two halves. The same topic gets two versions: the one written at home and the one produced under questioning. If the script says the margin fell "temporarily, on product mix", and the Q&A gives that same decline three sentences without a single figure, those are not one piece of information. They are two, about the same thing.

So why read the script at all? Because you need something to hold the Q&A against. Copy out two sentences: the one carrying guidance and the one management chose for the weakest part of the quarter. Three quarters from now one of them will read differently, and that note is the only reason you will notice.

That is why I read calls in pairs: one passage from the script, and the answer to the question about exactly that passage. The rest is a description of a quarter you already have in the filings. How management explains the distance between a promise and a result is its own subject, and I take it apart in the piece on judging management by promise, action and evidence.

One question in nine gets no answer#

Dodges on earnings calls can be counted, and somebody counted them. Non-Answers During Conference Calls (Gow, Larcker and Zakolyukina, Journal of Accounting Research, 2021) finds that about 11% of analyst questions get an answer that is not an answer, and that the rate is stable over time and similar across industries.

Which means that on a call with fifteen questions, one or two go unanswered, whether you are listening to a chipmaker or a supermarket chain.

Where those dodges land matters more than how many there are. The authors found non-answers more likely on questions with a negative tone, on more complex questions, on questions asking for more detail, and on questions that touch information a company would rather not hand its competitors. Non-answers are not spread at random. They land exactly where the question hurts, which makes the list of unanswered questions a list of topics management would rather avoid.

How do you tell a dodge from ordinary caution? By the shape of the answer. A dodge usually runs longer than the question, warmer in tone, and is built from nouns nobody can measure: customer engagement, a maturing pipeline, cost discipline. A cautious answer sounds different because it names its own limit: we do not disclose that metric, and we will give it after the year closes. The first runs away from a number. The second tells you when the number arrives.

So here is the reading rule, and it is almost embarrassingly simple. Do not count how often the word "demand" appears. Count how many questions got a figure back.

When the third analyst circles back#

Analysts do not agree on questions beforehand and do not share a list. So when the third one in a row returns to inventory, to churn, or to margin quality, it means one thing: the first answer did not convince them and they did not buy the second.

A repeated question is the one moment on a call when the market speaks in its own voice rather than the company's.

A topic three professional modellers keep returning to is a live argument, not a footnote. That does not make them right. It means there is uncertainty there worth checking yourself, in the filings.

It also happens to be the cheapest way I know to tell a topic from noise. New stories collect around any large company every day, and most of them do not survive to the next quarter. The list of questions from a call shows where the people who model the business for a living are looking. That is rarely where the headlines are looking, and if the distinction interests you, I go further into it in the piece on noise versus signal in investing.

The excuse the competitor did not need#

Good management names its own mistake and says what it will do about it. Weak management blames the weather, currency, supply chains and where the holidays fell in the calendar. The difference is not tone. It is whether you can check who is responsible.

The check is free: open the same-quarter call from a direct competitor and see whether they needed the same excuse. If the competitor raised prices and grew in that identical "challenging environment", you are listening to a story about management, not about the environment.

In practice it goes like this. You copy out the one sentence carrying the excuse, open the competitor's material from the same quarter, and look for the same cause. Either you find it almost word for word, because the whole industry took the same hit, or you do not find it at all. The first result closes the question and saves you an hour. The second opens a new one, because if a competitor in the same market had no trouble with the weather, the trouble was not the weather.

This test has one limit worth knowing about: two companies in one industry can sit at different points of the cycle, with a different geographic mix and different customer contracts. Macro really does explain a quarter sometimes. It just explains it for everybody at once, not for one company on its own.

The number that left the slide deck#

Say a company spent six straight quarters disclosing what share of revenue came from its ten largest customers, and the figure came in around 40% each time. In the seventh quarter it is gone. In its place is a new metric, "adjusted contribution margin", which appeared in no material a year ago.

A missing number proves nothing.

A change in disclosure is a signal, though, because a company whose metric looks good does not stop showing it. A new yardstick that arrives in exactly the quarter the old one turned ugly rarely serves clarity.

You check it in one move: pull the deck from four quarters ago and compare the table of contents. Not the content, the contents. A metric that fell off the list usually comes back as an analyst question, and what management does with that question is more interesting than the metric.

A year of calls instead of one#

A single call almost always sounds good, because that is how it was written. The signal is not one confident quarter. It is drift: the way a company talks about the same problem quarter after quarter.

Netflix left behind a textbook example, in its own documents. In the shareholder letter of 19 April 2022 the company disclosed that on top of 222m paying households, more than 100m additional households were using its accounts. It framed that this way: "Sharing likely helped fuel our growth by getting more people using and enjoying Netflix". Which means roughly one in three households watching Netflix was not paying, and the company called it growth fuel.

Nine months later, in the letter of 19 January 2023, the same subject reads: "Later in Q1, we expect to start rolling out paid sharing more broadly". Same phenomenon, same company, two different names three quarters apart.

Drift has a few standard shapes. A specific problem has its own name in the spring, is called "market headwinds" by the summer, and "long-term opportunities" by the autumn. Verbs change too: "we expect" becomes "we're comfortable with", "confident" becomes "encouraged", and the printed guidance range looks identical to last quarter's.

The easiest way to catch drift is through the analyst who will not let go. The same person asks the same question four quarters running, and you read the four answers one under the other and watch management either move toward a number or away from it.

I judge management on that drift rather than on any single appearance. A year of transcripts read back to back shows character: one team says "we misjudged inventory and here is what we are doing", another gives the same problem a new name four times in a row.

I will go further: in most cases the drift arrives a quarter or two before the guidance comes down. The words go soft first, the table follows. I have not counted a sample, so let me name what would sink the claim: a run of companies whose guidance drops with no quarter of vague language in front of it. I lay the comparison method out step by step in the piece on why you read three earnings calls, not one.

Do Warsaw-listed companies hold calls#

They do, on different terms and a different rhythm. Principle 1.6 of the 2021 Best Practice code for companies listed on the Warsaw exchange says a WIG20, mWIG40 or sWIG80 company holds an investor meeting once a quarter, and everyone else at least once a year. The same principle adds that at these meetings management publicly answers the questions put to it, so a Q&A is written into the format.

Which means the largest Polish companies give you the same raw material as the US market, more often as video than as text, while smaller ones give you a quarter of it and you have to replace the rest with something else.

So what do you do with a company that meets investors once a year? Replace the three missing quarters with what it has to publish anyway: current reports, the half-year statement and management's commentary on results. It is weaker material, because nobody asks it awkward questions, but it is still written under the board's responsibility.

The code works on a comply-or-explain basis, so a company may decline principle 1.6 as long as it explains why. That explanation is public, and it is occasionally more revealing than the meeting would have been. There is also principle 1.7: a company answers an investor's request for information within 14 days, which means the question no analyst asked, you can ask yourself.

Where to get a transcript for free#

US companies furnish the results release as an exhibit to a Form 8-K, and that exhibit sits in the EDGAR database at no cost. Nearly every company also posts the recording and the transcript in the investor section of its own site.

Go to that page even when you already have the release from the regulator. The company's own site carries two things the filing does not: the deck management walks the quarter through slide by slide, and the time of the webcast, so you can hear the Q&A live instead of reading it a week later. The deck earns its keep twice over, because its contents page is the one you compare with the deck from four quarters ago.

Polish issuers publish in the same place, more often as video than as text, so the Q&A costs you real time rather than a quick scroll.

Either way it is a first-hand document, which means nobody summarised or graded it for you on the way. Why documents like that beat second-hand analysis is the subject of the piece on primary sources in stock research.

The price is not a verdict#

In theory we could stop at "read the Q&A". In practice there is one more trap, and it is the easiest one to fall into: confusing the price reaction with the value of the call.

The move after results measures the gap between what the company showed and what the market expected. Say the price falls 10% after a good call: all that tells you is that expectations were higher. A rise after a call where management ducked the margin question three times does not cancel the three ducks. The price reaction and the research value of the conversation are two different things that happen to share a date, and what to do about that is the subject of the piece on why you don't trade the earnings print.

The cheaper version of the same trap is reading only the headline. "Company raises guidance" and "shares fall after earnings" describe the same evening, both accurately and both uselessly, because neither says what the analysts asked or what they got back.

The second limit is more serious. Tone is not evidence. A confident answer can be wrong, and a tense one can simply be the answer to a hard question from somebody who knows the model by heart. I do not know whether a management team that avoids a number is hiding it or does not have it; I do know that next quarter the number either comes back or the topic disappears, and only that settles anything.

So tone is for aiming your attention and the filings are for settling the question. The direction is one-way: take the question from the call, take the answer from the documents.

What to read at the next results#

At the next call from a company you hold, do four things:

1. Read the Q&A alone, from the last question to the first. The hardest questions usually come late, once the queue of analysts thins out.
2. Write down every question that did not get a figure in reply.
3. Check whether one topic comes back from three different analysts.
4. Open the deck from four quarters ago and compare its contents page with today's.

Those four steps take about an hour per company. The Earnings brief does the same work on one call: the numbers, what management said, what changed against the previous quarter. Earnings brief pro lines the current call up against the three before it, which is exactly the window where drift becomes visible. Under both sits the Transcript, so you can check a quote at the source instead of trusting a paraphrase, and quotes stay in the original language, because nobody can be held to a translated sentence. See how that reads on finished reports.

You can read one call in an hour. You cannot read drift at all without the three before it, and that is the whole difference between reading a transcript and holding management to account.