·

8 min read

What a platform’s earnings call can tell you, and can’t

What a platform's earnings call can tell you, and can't

A headline tells you a platform executive said video engagement is growing, and within a day the same sentence has become advice about what you should be posting. The sentence is usually real. The advice attached to it usually is not, because an earnings call is written for a different audience under a different set of obligations. Open the transcript yourself, see which parts carry weight and which are a headline writer’s inference, and you stop having to trust the intermediary.

What actually happens on an earnings call

An investor relations officer opens, names who is on the call, and reads a disclaimer. The CEO delivers prepared remarks, then the CFO covers financials and outlook. An operator then opens a queue and sell-side analysts ask questions, which the executives answer live. Two halves: scripted, then unscripted.

On Meta’s second quarter 2025 call, held July 30th, 2025, Kenneth Dorell of investor relations opened this way:

Our remarks today will include forward-looking statements, which are based on assumptions as of today. Actual results may differ materially as a result of various factors including those set forth in today’s earnings press release, and in our quarterly report on Form 10-Q filed with the SEC. We undertake no obligation to update any forward-looking statement.

That paragraph looks like throat clearing, but it tells you what the next hour is. Everything after it is a statement to investors about business performance, made under conditions where being wrong has legal consequences. It is not guidance to the people running accounts, and that difference in audience is why the language sounds the way it does.

Why the wording is more careful here than in a blog post

Two things constrain what gets said. The first is Regulation FD, codified at 17 CFR Part 243, adopted as SEC Release 33-7881 on August 15, 2000 and effective October 23, 2000. Its general rule, 17 CFR 243.100(a), requires that when an issuer discloses material nonpublic information about itself to certain outside parties, including brokers, investment advisers and holders of its securities, it must make that information public: simultaneously if intentional, promptly if not. That is why the transcript is published rather than shared with the analysts alone.

The second is antifraud liability under Section 10(b) and Rule 10b-5, which attaches to materially misleading statements made to investors. Regulation FD is explicit that it does not displace this; 17 CFR 243.102 reads: “No failure to make a public disclosure required solely by 243.100 shall be deemed to be a violation of Rule 10b-5 (17 CFR 240.10b-5) under the Securities Exchange Act.”

Now compare that to a company blog post announcing a feature, or a conference keynote. No equivalent disclosure obligation applies to either, and nobody gets sued for a marketing headline that oversells a product’s reach. This is why the same company can sound definitive in one venue and hedged in another, on the same subject, in the same week.

What the 10-Q says about a metric, in writing, from the company itself

The filing is where this gets useful, and it is the part almost nobody reads. Meta’s Form 10-Q for the quarterly period ended June 30, 2025 was filed on July 31, 2025, the day after the call. It defines its headline engagement metric:

We define a daily active person as a registered and logged-in user of Facebook, Instagram, Messenger, and/or WhatsApp (collectively, our “Family” of products) who visited at least one of these Family products through a mobile device application or using a web or mobile browser on a given day.

It reports that Family daily active people, or DAP, was 3.48 billion on average for June 2025, an increase of 6% year-over-year, and that “We view DAP as a measure of engagement across our products.”

Then it does something a press release never does: it tells you what the number cannot bear. Because the techniques and models used “require significant judgment, are developed based on internal reviews of limited samples of user accounts, and are calibrated against user survey data, there is necessarily some margin of error in our estimates.” It sizes that error: “we estimate that such margin generally will be approximately 3% of our worldwide DAP.” And it goes further: “it is also possible that our Family metrics may indicate changes or trends in user numbers that do not match actual changes or trends.”

Read that last clause again. The company is telling its own investors, in a filed document, that a movement in its headline engagement metric may not correspond to a real movement in engagement. That caveat sits in public while the same quarter’s number circulates without it. The same instinct applies to reading a platform transparency report.

Prepared remarks versus the Q&A section

Prepared remarks are written in advance, reviewed before delivery, and read aloud. Q&A answers are spoken live to questions the executives have anticipated but not scripted. Both carry the same disclosure obligations, so the difference is not legal weight. It is precision. Prepared remarks have had every qualifier placed deliberately; live answers have not.

Here is what a scripted engagement claim looks like. In her prepared remarks on the same call, CFO Susan Li said:

We continue to see momentum with video engagement in particular. In Q2, Instagram video time was up more than 20% year-over-year globally. We’re seeing strong traction on Facebook as well, particularly in the US where video time spent similarly expanded more than 20% year-over-year. These gains have been enabled by ongoing optimizations to our ranking systems to better identify the most relevant content to show.

Notice the shape. “We continue to see momentum” is an observation about an internally measured aggregate. “More than 20%” is a floor, not a figure. The causal sentence attributes the gains to Meta’s own ranking changes, a claim about what Meta did, not what any publisher did. Nothing in it is a commitment.

Later, in Q&A, an analyst asked what would drive further engagement lift on the core platform. Li’s answer named directions rather than mechanics: recommendations “even more adaptive to what a person is engaging with during their session”, and models the company is “planning to scale up” with techniques “that should improve the overall quality of recommendations”. “Should improve” is not “will”. A summary that reports this as a confirmed change has removed the load-bearing word.

The distance between a printed line and an account-level claim

Set the transcript sentence next to the filing language and the gap is visible without any outside commentary. What the sentence commits Meta to is narrow: that in the second quarter of 2025, an internally measured aggregate of Instagram video time rose more than 20% year-over-year, globally, and that the company attributes that rise to its own ranking work. That is a statement about billions of accounts, measured by a system the same filing says may show trends that do not match actual trends.

What it does not commit Meta to is anything about a specific account. It does not say video is favoured over other formats in ranking, or that a publisher who posts more video will see more distribution, or that the rise was spread evenly across categories, regions or account sizes. An aggregate rise is compatible with large parts of the population seeing a decline. Companies do not discuss individual accounts on these calls, so every step from the quoted line to a posting recommendation is inference added afterwards.

A short checklist for the next call

  1. Find the original document, not the writeup. Transcripts are usually a PDF on the investor site; filings are on SEC EDGAR.
  2. Note which half the line came from. Prepared remarks, or an answer to a named analyst in Q&A. The transcript labels both.
  3. Check whether the metric named has a published definition in the most recent 10-Q or 10-K. If it does, read the definition and the limitations section before the number.
  4. Read for hedges. “We believe”, “we continue to see”, “should”, “we expect”. Each is there on purpose, and a summary that dropped them changed the claim.
  5. Ask what population the claim covers. It will be platform-wide, so any account-level conclusion came from somewhere other than the call.

What this reading skill cannot give you

None of this tells you how ranking works. A platform has no disclosure obligation to explain its recommendation systems, and an earnings call would be the wrong venue anyway, because the audience is analysts modelling revenue. The most a call offers here is directional: the company is investing in recommendation quality, or scaling models. That is a spending story, not a mechanism. Where a platform has published something concrete about ranking, it published it elsewhere, and the line between what Instagram has published versus what gets assumed is drawn in that material.

So treat a transcript sentence as evidence about where a company says it is putting money and attention, stated under conditions that make casual exaggeration expensive. That is useful when you plan a quarter, because it tells you which surfaces the company is funding. It is not an operating instruction, and the filing published the next day says as much in writing.

Read the source, not the summary

The habit transfers. Applied to reading a platform’s own engineering posts and to reading a developer changelog, it catches the same error before it reaches a plan or a client report.

FAQ

Are earnings call transcripts actually public, or do I need investor access?

Public, both. Transcripts are posted on the company’s investor relations site, usually as a PDF, and filings are on SEC EDGAR, which also publishes filing dates through its submissions data. Some investor relations pages block automated requests, so bookmark the direct document URL, not the index page.

An executive said engagement with a format is growing. Is that a signal I should act on?

It is a platform-wide aggregate stated to investors, so treat it as information about the company’s priorities rather than about your account. Run the checklist: whether the line sits in prepared remarks or a live answer, whether the metric has a published definition and limitations section in the 10-Q, and what the hedging words were doing before a summary removed them.

Sources

Dinesh Agarwal Avatar