references/meeting_cost_canon.md
6.3 KB查看排版效果| 1 | # Meeting Cost Canon — what a meeting really costs, and the should-this-exist gate |
| 2 | |
| 3 | The first discipline of the `meetings` skill is treating a meeting as a purchase. Nobody would sign |
| 4 | a $600 invoice without asking what it buys, yet a six-person hour at a $90 fully-loaded rate is |
| 5 | exactly that — approved with one calendar click. This file documents the canon behind |
| 6 | `meeting_cost_calculator.py`: why the price must be said out loud, why the refocus overhead is real, |
| 7 | and why "no decision, no meeting" is the correct default. |
| 8 | |
| 9 | ## The scale of the problem |
| 10 | |
| 11 | - **Executives average ~23 hours a week in meetings**, up from under 10 hours in the 1960s, and 71% |
| 12 | of senior managers surveyed said meetings are unproductive and inefficient (Leslie Perlow, |
| 13 | Constance Noonan Hadley & Eunice Eun, "Stop the Meeting Madness," *Harvard Business Review*, |
| 14 | July–August 2017). Perlow et al.'s core finding: dysfunctional meeting cultures self-perpetuate |
| 15 | because no individual meeting ever gets billed for its cost. |
| 16 | - **Most meetings fail by design, not execution.** Steven Rogelberg's research program (surveying |
| 17 | thousands of employees and leaders) finds roughly half of meeting time is rated ineffective by |
| 18 | attendees, and the strongest lever is what happens *before* the meeting: whether it should exist, |
| 19 | who is invited, and what the agenda demands (Steven G. Rogelberg, *The Surprising Science of |
| 20 | Meetings*, Oxford University Press, 2019). |
| 21 | - **Deleting meetings works.** Shopify started 2023 by removing all recurring meetings with more |
| 22 | than two people (~12,000 events) and later shipped an internal **meeting cost calculator** that |
| 23 | embeds a dollar figure into every calendar invite — a typical 30-minute, 3-person meeting priced |
| 24 | between $700 and $1,600. The point was behavioral: people decline purchases they would have |
| 25 | accepted as invites. |
| 26 | - **Industry surveys agree on the waste.** Doodle's "State of Meetings" report estimated poorly |
| 27 | organized meetings cost businesses in the tens of billions annually across the US and UK alone; |
| 28 | Atlassian's long-running "you waste a lot of time at work" meeting statistics put the average |
| 29 | employee at ~31 hours per month in unproductive meetings. Exact figures vary by survey; the |
| 30 | direction never does. |
| 31 | |
| 32 | ## The refocus overhead (`--include-refocus`) |
| 33 | |
| 34 | The invite says 30 minutes; the attention bill is larger. Gloria Mark's interruption research at |
| 35 | UC Irvine measured an average of about **23 minutes to return to the interrupted task** after a |
| 36 | context switch (Gloria Mark, Daniela Gudith & Ulrich Klocke, "The Cost of Interrupted Work: More |
| 37 | Speed and Stress," *CHI*, 2008). A mid-morning meeting is an interruption for every attendee doing |
| 38 | focused work, which is why the calculator's `--include-refocus` flag adds 23 minutes per attendee |
| 39 | at the same hourly rate. It is an estimate — deliberately conservative (one refocus, not two) — and |
| 40 | it routinely doubles the sticker price of short meetings, which is the point. |
| 41 | |
| 42 | ## The gate: decision, agenda, owner |
| 43 | |
| 44 | The calculator refuses a MEET verdict unless three things exist: |
| 45 | |
| 46 | 1. **A decision** (`--has-decision`). Andy Grove's taxonomy separates *process* meetings |
| 47 | (one-on-ones, staff reviews — scheduled, informational) from *mission* meetings, which exist to |
| 48 | produce a decision (Andrew S. Grove, *High Output Management*, Random House, 1983). An ad-hoc |
| 49 | meeting with no decision is a status update wearing a meeting's clothes — Grove's rule was that |
| 50 | a mission meeting that produces no decision is a failure of the chairman, not the attendees. |
| 51 | Status flows async: Amazon's practice replaces presentation meetings with a silently-read |
| 52 | six-page narrative memo, and its "two-pizza team" rule caps the audience — both documented across |
| 53 | Jeff Bezos's shareholder letters and his explanation that "PowerPoint-style presentations |
| 54 | somehow give permission to gloss over ideas." |
| 55 | 2. **An agenda** (`--has-agenda`). Rogelberg's finding is blunt: agendas per se don't correlate |
| 56 | with meeting quality — *prepared, question-shaped* agendas do (see |
| 57 | [`agenda_discipline.md`](agenda_discipline.md)). The gate checks existence; the agenda builder |
| 58 | enforces quality. |
| 59 | 3. **An owner** (`--has-owner`). Perlow et al. and Rogelberg converge here: someone must be |
| 60 | accountable for the meeting achieving its outcome and for follow-through afterwards. A meeting |
| 61 | nobody owns produces minutes nobody reads. |
| 62 | |
| 63 | Missing the decision → **ASYNC** (the cheapest meeting is no meeting). Missing agenda or owner with |
| 64 | a real decision on the table → **NOT-READY**: the meeting may deserve to exist, but not yet. |
| 65 | |
| 66 | ## Why the verdict is ternary, not a score |
| 67 | |
| 68 | A 0–100 "meeting quality score" invites negotiation. Three verdicts with hard edges do not: |
| 69 | ASYNC ends the conversation (write the memo), NOT-READY names exactly what to fix, and MEET prints |
| 70 | the invoice — total cost plus cost-per-minute — so the agenda's timeboxes get budgeted like money. |
| 71 | This mirrors Shopify's design insight: the number changes behavior only when it arrives *before* |
| 72 | the commitment is made. |
| 73 | |
| 74 | ## Sources |
| 75 | |
| 76 | 1. Leslie Perlow, Constance Noonan Hadley & Eunice Eun, "Stop the Meeting Madness," *Harvard |
| 77 | Business Review*, July–August 2017 (23 hrs/week; self-perpetuating meeting cultures). |
| 78 | 2. Steven G. Rogelberg, *The Surprising Science of Meetings*, Oxford University Press, 2019 |
| 79 | (meeting-science canon; pre-meeting levers dominate). |
| 80 | 3. Shopify meeting-cost calculator + recurring-meeting purge, 2023 (cost surfaced in the invite; |
| 81 | ~12,000 events deleted; widely covered, incl. Bloomberg/Reuters reporting). |
| 82 | 4. Jeff Bezos, Amazon shareholder letters + public remarks (silent six-page narrative memo; |
| 83 | two-pizza team rule capping meeting size). |
| 84 | 5. Doodle, "State of Meetings" report (survey-based estimate of the cost of poorly organized |
| 85 | meetings across US/UK). |
| 86 | 6. Atlassian, "You Waste a Lot of Time at Work" meeting statistics (~31 unproductive meeting |
| 87 | hours/month per employee). |
| 88 | 7. Andrew S. Grove, *High Output Management*, Random House, 1983 (meetings as the medium of |
| 89 | managerial work; process vs mission meetings; a decision-less mission meeting is a failure). |
| 90 | 8. Gloria Mark, Daniela Gudith & Ulrich Klocke, "The Cost of Interrupted Work," *CHI* 2008 |
| 91 | (~23-minute refocus after interruption — the basis of `--include-refocus`). |
文件内容是平台检查时读出的文本,只供查看。平台不会运行包里的任何脚本或程序。

