How Quorum measures

The base for share-of-the-week figures and the team grade. A workspace with different hours sets its own working day.

It is the week the studies on this page measure against — Slack’s “two hours a day”, Microsoft’s weekly meeting hours, Rogelberg’s weekly survey totals — so a share of your week can be compared with theirs.

Sources: Slack Workforce Index 2023 (opens in a new tab), Microsoft Work Trend Index 2023 (opens in a new tab), Rogelberg & Otter.ai 2022 (opens in a new tab)

Decides when the time after the last meeting stops counting as focus time. Each workspace sets its own day in Settings.

A gap only counts as usable time while people are still at work. The default matches the working week above; a workspace that starts earlier or later changes it.

Up to this point, meeting hours cost a team nothing in its grade or load score. It is 25% of a 40-hour week.

In Slack’s 2023 Workforce Index of 10,333 desk workers, more than two hours a day in meetings was the point past which most said meetings hurt their work. Microsoft’s 2023 Work Trend Index puts its heaviest quarter of Teams users at about 7.5 hours a week, so this is a target, not an average.

Sources: Slack Workforce Index 2023 (opens in a new tab), Microsoft Work Trend Index 2023 (opens in a new tab)

Where the meeting-load part of a team’s grade reaches zero.

Rogelberg’s 2022 survey of 632 US employees found 18 hours of meetings a week on average, about a third of which they would decline if they could. Perlow, Hadley and Eun found senior managers at about 23 hours.

Sources: Rogelberg & Otter.ai 2022 (opens in a new tab), Perlow, Hadley & Eun 2017 (opens in a new tab)

A team is graded only when platform attendance covers at least this share of its measurable meetings in the period: those with a conferencing link on a platform the workspace has connected. Meetings with no link, or on a platform that is not connected, are left out of that share and shown beside it, and their booked hours count in the load the grade scores. Below the line the grade is withheld and the screen says why; shares are rounded down, so a figure below the line is never shown as the line.

Our own choice. The grade’s unattended share comes from measured meetings, so most of what could be measured has to be. Counting rooms and unconnected Zoom links against coverage left every Google-only workspace ungraded; requiring every measurable meeting let one late report withhold a grade.

A gap this short between meetings counts as fragmented, not as usable time.

Indirect evidence: Gloria Mark’s interruption research is the origin of the often-quoted 23 minutes to get back into a task — best read as an order of magnitude — and Leroy (2009) showed that attention left behind on one task lowers performance on the next.

Sources: Gloria Mark, UC Irvine (opens in a new tab), On the “23 minutes” figure (opens in a new tab), Leroy 2009 (opens in a new tab)

The shortest gap Quorum calls a focus block.

Microsoft Viva Insights books focus time in blocks of this length, and the deep-work literature treats it as a real working session. The interruption research above is why a shorter stretch does not count.

Sources: Gloria Mark, UC Irvine (opens in a new tab), Leroy 2009 (opens in a new tab)

A run of meetings this long with no break is named as a chain with no recovery in it.

In Microsoft’s 2021 EEG study, brain activity associated with stress rose steadily across two hours of back-to-back meetings and did not reset. With ten-minute breaks between the same meetings it held level.

Sources: Microsoft Human Factors Lab 2021 (opens in a new tab)

The share of days that should keep at least one focus block.

An MIT Sloan Management Review study of 76 companies found three meeting-free days a week was the optimum, with productivity up 73% and cooperation up 55%. One unbroken block is a weaker test than a whole meeting-free day, so this is a floor, not a stretch goal.

Sources: Laker et al., MIT Sloan Management Review 2022 (opens in a new tab)

More separate clusters of meetings than this in a day lowers the load score.

Our own derivation: in an eight-hour day that needs one focus block, this is the most separate clusters likely to leave room for it. Interruptions carry weight because the CIPD’s 2023 evidence review found that the number of meetings, not total hours, goes with lower effectiveness and wellbeing.

Sources: CIPD evidence review 2023 (opens in a new tab)

The reference point for the size rules below.

Bain & Company’s “rule of 7”: once a decision group has seven members, each additional person cuts decision effectiveness by about 10%. Rogelberg (2019) gives the same figure for decisions and problem-solving.

Sources: Bain & Company, rule of 7 (opens in a new tab), Rogelberg 2019

A larger meeting loses design points, and this is the default invitee cap for routine meetings.

Microsoft Viva Insights classes a meeting with nine or more invitees as large, and Rogelberg’s range that works only with strong facilitation starts at eight.

Sources: Microsoft Viva Insights (opens in a new tab), Rogelberg 2019

A second design penalty applies above this.

The top of Rogelberg’s facilitated range. By this size, Bain’s rule has taken away about half of a group’s decision effectiveness.

Sources: Rogelberg 2019, Bain & Company, rule of 7 (opens in a new tab)

A longer meeting loses a few design points.

Viva Insights calls a meeting over an hour long, and Google Calendar’s Speedy Meetings trims defaults to 25 and 50 minutes. The CIPD review finds no link between length and how effective a meeting feels, so the penalty is small and a shorter meeting counts as time saved, not as better quality.

Sources: Microsoft Viva Insights (opens in a new tab), Google Calendar Speedy Meetings (opens in a new tab), CIPD evidence review 2023 (opens in a new tab)

Punctuality points fall steadily to zero at this point; starting 5 minutes late earns half.

Allen, Lehmann-Willenbrock and Rogelberg (2018) found a meeting that started five minutes late was already rated less satisfying, and one that started ten minutes late also less effective, with fewer and weaker ideas. Punctuality is one of the more consistent predictors of how effective a meeting feels.

Sources: Allen, Lehmann-Willenbrock & Rogelberg 2018 (opens in a new tab), CIPD evidence review 2023 (opens in a new tab)

A meeting that runs this far past its booking is flagged as an overrun.

Google’s Speedy Meetings exist to leave five minutes between meetings, so an overrun of this size hands the next meeting the late start Allen and colleagues measured. Microsoft’s EEG study shows the lost break costs something too.

Sources: Google Calendar Speedy Meetings (opens in a new tab), Allen, Lehmann-Willenbrock & Rogelberg 2018 (opens in a new tab), Microsoft Human Factors Lab 2021 (opens in a new tab)

A series whose median occurrence reaches two people present this long after its booked start has its waiting time priced on its page and in its recommendation. It times the room, never a person.

Half the late-start line above: Allen, Lehmann-Willenbrock and Rogelberg (2018) found five minutes late already rated less satisfying.

Sources: Allen, Lehmann-Willenbrock & Rogelberg 2018 (opens in a new tab)

The default line for a meeting booked at short notice, measured from when the calendar says the event was created. A meeting whose creation time is unknown is never counted. Not costed; each workspace sets its own line.

Our own choice: one working day, the same planning horizon as the response-time convention. Fellow measures short-notice meetings as one of its guidelines; no study names the hour count.

Sources: Fellow guideline analytics (opens in a new tab)

The default cap for the share of a team's people above it — reported for teams of at least 3 people, never as a list of names.

The same line as the meeting-hours target: past two hours a day, most people in Slack's 2023 Workforce Index said meetings hurt their work.

Sources: Slack Workforce Index 2023 (opens in a new tab)

A meeting that starts this soon after the previous one on the same person's day — or overlaps it — counts toward the team's back-to-back share.

Microsoft's Human Factors Lab (2021) compared meetings run with no break against the same meetings with ten-minute breaks; stress built only without one.

Sources: Microsoft Human Factors Lab 2021 (opens in a new tab)

Below this a series is kept and collecting data; a phantom series or a late-start habit needs this many measured occurrences.

Our own choice: one empty room is an accident, three in a row is a pattern.

When every measured occurrence had nobody meeting — never two people present together, or less than this share of the invited time attended — the series is recommended for retirement.

Viva Insights' recurring-meeting audit excludes series no one joined. Quorum never stores an empty attendance report, so “no one” is measured as nobody meeting; the share is our own choice.

Sources: Viva Insights recurring meeting audit (opens in a new tab)

A recurring series is in a reset's scope when it met or is booked at least this often in the last 90 days or the next 14 days.

Our own choice: one occurrence is not a recurring commitment, and a series silent for a quarter has already been retired.

A retired series is confirmed from the calendar once its next occurrence would have been due — the last one plus its measured cadence, or 28 days when the cadence is unknown — plus this grace for a late sync.

Our own choice: the calendar is read a week ahead, so an empty week proves nothing on its own.

A meeting whose attendees rate it below this on average is flagged as poorly rated.

Ratings use the five-point scale Rogelberg and colleagues (2006) established for meeting effectiveness, and this is its midpoint: neither good nor poor. How attendees rate a meeting is the outcome meeting research measures against, so ratings carry the most weight in Meeting Health when there are any.

Sources: Rogelberg et al. 2006 (opens in a new tab), CIPD evidence review 2023 (opens in a new tab)

A stable random sample of meetings with at least 5 people who stayed 5 minutes or more is asked about; nobody is asked about every meeting.

Microsoft Viva Insights samples 15% of qualifying Teams meetings, and only those with five or more participants who stayed at least five minutes. A rare, short question is one people answer.

Sources: Viva Insights meeting effectiveness surveys (opens in a new tab)

Below this, a meeting’s or a series’ pulse shows only how many people answered: no average, no reasons, no poorly-rated flag and no effect on Meeting Health or recommendations.

Viva shows organisers only aggregated results, and its minimum group size for any aggregate cannot go below five. Fewer answers are a handful of people’s opinions with their names nearly on them.

Sources: Viva Insights meeting effectiveness surveys (opens in a new tab), Viva Insights minimum group size (opens in a new tab)

When at least this share of a series’ pooled answers names the same reason — could have been async, no clear purpose, no outcome, too many people — the series recommendation follows it.

Our own choice: a majority of the room, not a vocal minority. “Ran over” is left to measured attendance, which says by how much.

The time counted for each message written in a thread.

Our own estimate: a hundred-odd words at a working typing pace, plus the thinking behind them. It is the least certain number in the model, which is why the floors below exist.

The time counted for every other participant to read each message.

Adults read non-fiction at about 238 words a minute, according to Brysbaert’s 2019 meta-analysis of 190 studies. A typical work message runs a hundred to a hundred and twenty words; re-reading the context it answers takes the rest.

Sources: Brysbaert 2019 (opens in a new tab)

A thread is compared with a call of this length with the same people, priced the same way: minutes × people × the loaded rate. Finding a slot and sending the invite counts too.

The most common default meeting length in calendar tools. A shorter call would make threads look expensive sooner, so this errs toward leaving threads alone.

Quorum proposes a call only for a thread at least this long, among this many people.

Our own choice, so the feature does not nag. The message floor is a backstop: within this range of people, no thread can cost more than the call before reaching it. Above the people cap a thread is an announcement rather than a discussion to convene, and the cap sits under the size Rogelberg treats as the limit for a working session.

Sources: Rogelberg 2019

A thread must have run longer than this and been touched recently. A shorter burst is left to play out; one untouched for longer has settled itself.

Our own choice.

A conversation or review waiting longer than this for a first reply counts as slow.

Google’s published engineering practices name one business day as the longest a code review request should wait for a response. Applying the same line to written threads is our own choice.

Sources: Google Engineering Practices (opens in a new tab)

A team, repository or channel with this share of conversations still waiting is marked slow, however quickly the rest were answered.

Our own choice. A median over answered conversations cannot see the ones nobody picked up.

Below this, no response-time figure is shown for a team, repository or channel.

Our own choice. With fewer, one thread can swing the median and the figure describes one person’s habits rather than a team’s.

The published figure a workspace’s own ratio — how long conversations AI agents opened wait for a first human response, against those people opened — is read against on Conversations and in the report. Nothing is marked slow by it.

LinearB’s 2026 Software Engineering Benchmarks Report, across its customers’ pull requests. Its pickup time is close to, but not the same as, Quorum’s first human review in working hours, so the two are set side by side, not compared as one measure.

Sources: LinearB 2026 Benchmarks (opens in a new tab)

Where a new working agreement starts; each team or place sets its own target and share.

One working day is Google’s published maximum for a first code review, the same line the response-time screen uses. Swarmia and LinearB state working agreements as a time and a share of work meeting it; the 80% is our choice.

Sources: Google Engineering Practices (opens in a new tab), Swarmia working agreements (opens in a new tab)

The targets an agreement may be set to.

Our own choice. Below one working hour a reminder cannot arrive before the breach it warns of; a week is the longest wait worth agreeing to; below half is not an agreement.

A conversation still unanswered after this much of its target is named, once, to the scope’s owner — never who should answer it.

Our own choice: on a four-hour target it leaves an hour to act. Earlier nags about threads that would have been answered anyway.

A budget is ahead of pace when its planned spend is more than this above the share of the month that has passed.

Our own choice. Meetings are not spread evenly through a month — a planning week front-loads it — so a narrower band would flag every first week.

When the average person meets this much, the screens lead with meeting cost and attendance. It is the meeting-hours target above, not a second value for it.

In Slack’s 2023 Workforce Index, more than two hours a day in meetings was the point past which most desk workers said meetings hurt their work.

Sources: Slack Workforce Index 2023 (opens in a new tab)

Below this, the screens lead with written work (thread cost, response times and focus) with meeting cost beside it, never added to it. Between the two lines a workspace is balanced.

Our own choice: half the line above, an hour a day. It sits well under the roughly 7.5 hours a week Microsoft’s 2023 Work Trend Index found for its heaviest quarter of Teams users. No study names a line for async-first.

Sources: Microsoft Work Trend Index 2023 (opens in a new tab)

A workspace is not classified until this much calendar history has been seen. Until then the screens keep their usual order, or lead with written work when no calendar is connected.

Our own choice. A weekly meeting has then been seen twice, and one unusual week is at most half the sample.