What the best evidence actually says about meeting load.

Almost every article about the cost of meetings recycles the same few numbers: 71% of managers say meetings are unproductive; meeting time jumped 252% after 2020; US companies lose $37 billion a year. The first two survive scrutiny. The third doesn’t, and I’ll explain why below.

I wrote this piece because I got tired of tracing citations and finding they led to someone paraphrasing someone else paraphrasing a press release. What follows is three numbers I’m comfortable citing, and three I’ve stopped using. Primary sources linked throughout.

The three numbers that hold up

71% of senior managers say their meetings are unproductive

From a survey of 182 senior managers reported in Perlow, Hadley & Eun, “Stop the Meeting Madness” (HBR, 2017). Survey, not time-diary, but a large-enough sample of senior operators that the direction is robust. The same paper reports that 65% say meetings keep them from doing their own work.

15% of company-wide time goes to meetings

From Mankins, Brahm & Caimi, “Your Scarcest Resource” (HBR, 2014), based on Bain analysis. Their exact wording: “15% of an organisation’s collective time is spent in meetings — a percentage that has increased every year since 2008.” The article also notes that figure roughly doubles for senior leaders.

+252% weekly meeting time per user since 2020

From the Microsoft Work Trend Index “Great Expectations” report (2022). Based on anonymised Teams telemetry — the number of weekly meetings per user rose 153% and the weekly minutes spent in meetings rose 252% between Feb 2020 and Feb 2022. I wrote a longer piece tracking the follow-on 2023–2025 editions: Why meeting load doubled after 2020.

The three I’ve stopped citing

“$37 billion a year lost to unproductive meetings”

The canonical debunking is Elise Keith’s trace at Lucid Meetings. The number appears to originate in a late-1980s Industry Week item and has never been reconstructed from a primary source. It gets re-cited because it’s round and it feels right. Don’t use it.

“Mintzberg found managers spend 10 hours a week in meetings”

Mintzberg’s The Nature of Managerial Work (1973) observed five chief executives and reported they spent 59%of their time in scheduled meetings, plus ~10% more in unscheduled ones. That’s closer to 25–30 hours a week than 10. The 10-hour paraphrase has taken on a life of its own.

“Atlassian 2023: average meeting length grew by 16 minutes”

Atlassian didn’t publish a State of Teams report in 2023. They did 2022, 2024, and 2025, none of which contain this stat. Skip it — the 2022 and 2024 reports have genuinely useful numbers on distributed-team meeting load if you need an Atlassian data point.

The one stat that deserves a footnote

Gloria Mark’s “23 minutes and 15 seconds” to recover from an interruption is real — and I use it often. But it comes from Mark, Gonzalez & Harris, “No Task Left Behind?” (CHI 2005), not the 2008 Mark/Gudith/Klocke paper it’s usually attributed to. The 2008 paper measures speed and stress effects of interruptions, which is a different question. Cite the right one.

If this publication is going to earn the trust of people making decisions about their calendars, every number in it needs to trace back to a primary source. I’ll keep this piece updated as new research lands, and I’ll fix anything I get wrong.