Meta published which surfaces grew. Here's what it won't tell you.
Meta's Q3 disclosure shows where attention went. None of it says whether your ads sold anything. The three questions to run before you move budget.
Meta's Q3 disclosure tells you where attention went. It does not tell you whether attention there sells. Every figure Meta published on October 22 is denominated in time, share or count — none in price, conversion or revenue — and not one of them can be lifted into your ad account. So before a budget moves, run three questions over the number you are about to cite: which denominator, which objective, which attribution window. Survive all three and you have a place to test. Fail one and you have a headline. The money still moves only on a figure your own system recorded for the same period.
What Meta actually published on October 22
Alongside its Q3 2026 earnings release, Meta put out usage and engagement notes for Facebook, Instagram and Threads. Social Media Today read them app by app on October 22, and the same day it reported what Zuckerberg said about Muse. The underlying numbers come from Meta's Q3 2026 earnings call, which is worth reading yourself rather than through a summary.
Both write-ups are worth reading in full — as disclosures, not as forecasts. Here is the same material arranged for someone who buys media:
| Surface | What Meta published | What it cannot answer | The question to ask |
|---|---|---|---|
| Instagram Reels | Global time spent per user at an all-time high; relevance launches raised Reels time spent by over 4%; nearly 45% of recommended Reels are less than a day old, up more than 11 points from Q2. | What a thousand of those viewers costs, and what any of them does after watching. | Out of which population, and compared with which quarter? |
| Facebook Feed and Reels | Reels time spent at all-time highs globally and in the US, growing double digits year over year; a new generative retrieval architecture covering roughly a quarter of unconnected Feed views and half of Reels views, worth about +1% watch time. | Whether the extra minute is inventory you can buy, and at what price. | Which objective is this metric optimizing? |
| Threads | Time spent per user up more than 20% year over year; Zuckerberg said it has overtaken X in monthly active users. | Whether a placement on that surface reports into your attribution setting at all. | Can I buy it, and does my own measurement see it? |
| Creator Marketing Hub | Global rollout to all advertisers finished in Q3, with more than 10 million creators discoverable; partnership ads grew over 97% year over year and passed a $15 billion annual revenue run-rate. | Your return. The run-rate is Meta's revenue, not your ROAS. | Which of these numbers belongs to the seller? |
| Muse | More than 10 million downloads; about 25% of people who set it up use it daily; shopping is the most common use case, with more than 25% of users asking it to find deals or make purchases, and an average purchase over $200. | Whether you can advertise there, and whether your measurement would see the sale at all. | Is this a placement, or a PR line? |
Read down the third column and the pattern holds: every row answers "did people use the surface", and not one answers "did it sell". That is not a gap Meta is hiding. Company-level engagement statistics and account-level outcomes are drawn from different populations, so one cannot be derived from the other — which is exactly why a platform disclosure can be completely accurate and still be useless as a media plan.
The three questions to run before any budget moves
Which denominator?
"+4%" of what, measured over which baseline, for which users? An all-time high in time spent per user can be driven by a ranking change, a content cycle, a change in the population of people who open the app, or all three. None of those are things your campaign caused. Ask what the number is a percentage of and who is inside it. If the answer is "global, all users, versus last quarter", you have learned something about the platform and nothing about your addressable audience.
Which objective?
Engagement metrics optimize for engagement. If you are buying leads, qualified conversations or purchases, an attention metric sits upstream of your outcome — closer to it than a brand-lift number, further from it than a conversion. Ask which objective the published metric is attached to and whether that objective is the one you are paying for. Reels watch time is a strong argument for testing a Reels placement. It is not evidence that the placement returns what your pipeline needs.
Which attribution window?
Meta itself redefined click-through attribution in March 2026, requiring an actual link click — likes, shares and saves stopped counting (Meta's own announcement). When the ruler changes, a comparison across that change measures the ruler. So ask two things: which attribution setting the figure is reported under, and whether your destination-side system recorded the same events in the same window. That second question is the one that settles arguments — we covered why platform-reported conversions and CRM-verified leads drift apart and how to read the spread without picking a side.
The one number the disclosure structurally cannot give you
It is not a secret number. It is a kind of number: any figure denominated in money for your account. No CPM, no cost per outcome, no revenue per surface, no incrementality. Meta's disclosure is precise about Meta's money — a $15 billion partnership-ads run-rate is a real, useful figure — and silent about yours, because yours is not knowable from the outside.
This is where the reading method earns its keep. In the accounts we run, the rule is procedural rather than clever: two columns stay separate at all times, one for what the platform reports and one for what the business's own system recorded for the same window, and no sentence in a report is allowed to blend them. Nothing about that requires privileged access. It requires refusing to let a platform-published engagement figure stand in as a business outcome, even when the figure is flattering and even when it is true.
The decision rule for next quarter's budget
- Move on outcomes, test on engagement. Platform engagement decides where the next test runs. Your own recorded results decide how much money moves. Reversing that order is how a quarter gets spent on a chart.
- Fail one question, cap instead of shift. If the number cannot name its denominator, its objective or its attribution window, fund the surface at a test-level share rather than reallocating into it. You are buying information, not yet buying demand.
- Require the destination-side record for the same period. A surface earns a larger share only after your CRM or order system sees qualified actions from it, not after the platform reports engagement from it.
- Keep organic and paid as separate decisions. Growing an organic surface and buying against one are different jobs with different proof. That split is the reason we run social media and digital ads and PPC as separate workstreams rather than one budget line.
The question to ask every month
One question, to whoever runs the account: which of our surfaces produced qualified actions last month, on which attribution setting, and what did our own system record for the same window? If the answer opens with an engagement figure, redirect it. The person who can answer this one cleanly is telling you something about their reporting; the person who cannot is telling you something too.
The direct answer
Meta published an honest account of where attention moved and no account at all of what attention is worth to you, because those are different questions asked of different populations. Read the disclosure for where to test next quarter, not for what to reallocate. Run the three questions — denominator, objective, attribution window — over any figure you intend to repeat in a meeting, and let the money move only on a number your own system recorded for the same period. Engagement tells you where to look. Outcomes tell you when to move.
If you want the reading method applied to your own reporting — two columns, one monthly question, a budget rule that survives a platform's disclosure cycle — it is part of how we run digital ads and PPC management, and the fastest next step is a working session on your account.
Frequently Asked Questions
Does a per-platform engagement number predict ad performance?
No. A per-platform figure such as time spent per user or Reels watch time measures attention inside the platform, on a population the platform defines, for a period the platform chose. It says nothing about cost, conversion or incrementality in your account. Treat it as evidence about where to test next, not as evidence about what a test returned.
How do I compare a social surface with a paid placement on the same objective?
Put both on the same outcome definition and the same attribution setting, then read them against what your own system recorded for the same window. If one side only has a platform-published engagement figure, it is not comparable yet — it is a hypothesis with a headline attached. Fund it as a capped test until it produces a number your CRM or order system also sees.
