CoS Metrics: Meeting Throughput, Decision Velocity, Follow-Through Rate

Picture the QBR. Twenty minutes in, the CEO turns and asks, "So what did the CoS function deliver this quarter?" You open your mouth and out comes "the leadership team works better now."

That sentence is why Chiefs of Staff get cut in downturns. Not because the work wasn't real. Because the work didn't have a number on it.

I once watched a CoS get pushed out three weeks after a board meeting because the CFO ran a spreadsheet of every function's measurable output and the CoS row was blank. The CEO defended the role for about four minutes. Then the CFO asked, "what would change next quarter if we didn't refill it?" Nobody had a good answer. The role was gone by Friday.

This piece is the scoreboard. Five metrics, one slide, and the diagnostic that catches the trap most CoSes fall into around quarter four.

Why CoSes Don't Measure Their Own Work

Most of the job is invisible by design. You're killing meetings the CEO forgot to cancel. You're catching a VP-versus-VP fight in the hallway before it lands in a Slack thread. You're rewriting a board narrative at 11pm so the CEO doesn't get blindsided. None of that shows up in a dashboard.

The instinct is to say "my output is the CEO's output." It feels right. It is wrong. Two reasons.

First, a CEO's output is a function of a hundred variables, and you don't control ninety of them. Sales hired a bad VP, churn ticked up, fundraising slipped — you absorb the blame for things you couldn't fix. You also absorb credit for things you didn't do, which feels nice until the board starts asking how much of the credit is real.

Second, the CFO's argument is structural. Every other function on the leadership chart has its own row in the operating model. Sales has bookings. Marketing has pipeline. Engineering has shipping cadence and incident metrics. The CoS sitting in a strategy seat with no row of their own is, on paper, an expensive headcount with diffuse outcomes. That's the version the CFO writes up when the burn conversation gets serious.

You need a scoreboard. Not because the work isn't real, but because the people who decide whether the role exists next quarter need a defensible row in their model. Your job, on top of running cadence, is to give them that row.

The Five Metrics

Pick these five. Report the same five every quarter. Don't add a sixth until you've reported these three quarters in a row, because every metric you add weakens the slide.

1. Decision Velocity in the WBR

Count the decisions logged in the Weekly Business Review. Track time-to-decision per agenda item from when the topic opens to when a decision lands in the log.

Healthy baseline at a 50-500 person company: six to eight decisions per sixty-minute WBR, average time-to-decision under eight minutes per item. Below four decisions per WBR and the meeting is a status read. Above ten and you're probably skipping debate on things that deserve it.

How to instrument: a simple decision log column in the WBR doc with a timestamp when the topic opens and another when it closes. Your scribe owns the timestamps. You own the rollup. Most CoSes I know use a tracking column directly inside the WBR running doc and pull the numbers at the end of the quarter.

What to watch for: a sudden spike in decisions per WBR usually means people are pre-cooking decisions in 1:1s and rubber-stamping in the meeting. That's not bad, but it changes what the WBR is for. Note the change explicitly.

2. Follow-Through Rate on Commits

Of every decision logged in the WBR, what percentage shipped by the committed date? This is the metric that separates an operating cadence from decision theater.

Healthy: 75% or above. Acceptable: 65-75% in a quarter where you're shipping a lot of structural change. Red flag: below 60%. If your follow-through rate is below 60%, the WBR has stopped being a decision-making forum and is now a place where leaders agree to things they don't intend to do.

I've watched two companies hit 45% follow-through. In both cases the diagnosis was the same: leaders were saying yes in the room because saying no felt confrontational, then quietly de-prioritizing the commit afterward. The fix was not a new tool. The fix was the CoS calling out, in week six of the quarter, every decision that hadn't moved, by name, in the WBR. Awkward for two cycles. Follow-through went to 71% in one quarter.

Pull this number from the same decision log you're already running. The data is there. Most CoSes just don't roll it up.

3. Leadership Team eNPS

Quarterly five-question pulse to your L1 (executives) and L2 (their direct reports) only. Not the whole company. The CoS owns this (not People Ops) because it's a reading of the operating system, and you run the operating system.

Five questions, anonymous, takes four minutes:

1. How likely are you to recommend leading at this company to a peer at another company? (0-10)
2. The decisions made in the last quarter were the right ones. (1-5)
3. I have the information I need to do my job well. (1-5)
4. The leadership team is aligned on priorities. (1-5)
5. One thing the operating cadence should change next quarter (free text).

The score matters less than the trend. A score that drops from 7.4 to 6.1 in a quarter is a louder signal than a flat 6.8. You're looking for shape and direction, not absolute number.

Why the CoS owns it and not People: People Ops surveys are about engagement and culture across the whole org. This survey is about the operating system at the top. Whether decisions are clear, whether priorities are aligned, whether the cadence is working. That's your function. People Ops doesn't have the context to interpret the free-text on question five. You do.

4. Meeting Hours Per Leader Per Week

Pull from the calendar API. Segment by IC versus Director versus VP. Most companies are stunned by the actual number when they see it.

Working ranges I've seen hold up across 50-500 person companies:

• ICs: 8-14 hours/week is healthy for an IC who collaborates a lot. Above 18 and the IC is being used as a meeting prop.
• Directors: 18-25 hours/week. This is the band where Directors are doing real cross-team work without losing maker time entirely.
• VPs: 20-28 hours/week. Above 30 is a delegation problem dressed as a scheduling problem. The VP is in too many rooms because they don't trust their team to be in the room without them.

The diagnostic value is in the segments, not the average. If your VP average is 32 hours/week, fixing it is a structural conversation about delegation, not a calendar audit. If your IC average is 17 hours/week, fixing it is a calendar audit and probably involves killing some standing meetings the CEO forgot to cancel.

How to instrument: most calendar APIs (Google, Outlook) expose meeting time directly. A 30-line script pulls a quarter of data per leader. Don't make this a tool project. Make it a quarterly pull.

5. OKR Scoring Discipline

At the end of each quarter, OKRs get scored. The metric isn't the average score. It's the distribution shape and the percentage scored honestly.

A healthy OKR culture has a distribution centered around 0.6-0.7, per Google's original framing. If your average OKR score is 0.95 across the leadership team, you don't have a high-performing team. You have sandbagging. People are setting OKRs they already know they'll hit, scoring a 1.0, and walking out of QBRs looking like heroes.

Track two things:

• Distribution shape. What percentage of OKRs scored 0.7 or below, what percentage 0.7-1.0, what percentage above 1.0 (rare but possible). Healthy: roughly 30/55/15.
• Honesty rate. Percentage of OKRs where the score reflects the actual outcome and not a generous interpretation. This is judgment work. You read the OKR, the score, and the underlying data. If a "launch new pricing tier" OKR is scored 0.8 because the tier launched but adoption is zero, that's a 0.4 dressed as a 0.8.

The CoS scoring honesty rate is awkward in year one. It's also the single most useful intervention you can make in the cadence. You're not policing. You're recalibrating.

The "More Decisions But Worse Outcomes" Diagnostic

Here's the trap. Most CoSes optimize decision velocity in their first two quarters. The number goes up. Six decisions per WBR becomes nine. Average time-to-decision drops from eleven minutes to six.

The CoS reports the number proudly. The CEO is happy. Two quarters later, follow-through drops from 78% to 54%. Nobody connects the two.

What happened: you optimized throughput at the cost of decision quality. The decisions getting made faster are the wrong ones, or the under-baked ones, or the ones where the room agreed to a path that nobody actually believes in. Velocity went up. Outcomes got worse.

Walk through a worked example. Imagine two quarters of data:

The throughput-only read of this table is "decision velocity up 80%, average time-to-decision down 48%." That's the version that goes on a slide and gets the CoS a favorable QBR.

The honest read is: by Q3 the leadership team is making twice as many decisions, half are not shipping, and the team is less confident in the operating system than they were two quarters ago. You weren't running a cadence. You were running a decision factory with no QC.

The fix is not slowing down. The fix is adding a friction point: any decision that ships in under three minutes gets flagged for a follow-up review. If a decision was that easy, either the topic didn't need a meeting or the room wasn't surfacing the disagreement. Both are worth catching.

Watch the four numbers together. Velocity and follow-through have to move in the same direction. If they diverge for two consecutive quarters, you have decision theater.

The QBR Slide

One slide. Five numbers. Last quarter versus this quarter. One diagnostic callout in red.

Layout:

CHIEF OF STAFF FUNCTION — Q3 SCOREBOARD

                          Q2        Q3       Δ        Target
Decisions per WBR         7.8       8.1      +0.3     6-8
Time-to-decision (min)    7.6       7.4      -0.2     <8
Follow-through rate       71%       77%      +6pp     75%+
Leadership eNPS           7.0       7.3      +0.3     trending up
Avg meeting hrs/VP/wk     27        24       -3       <25
OKR honesty rate          61%       74%      +13pp    >70%

DIAGNOSTIC: VP meeting load came down via the Tuesday review consolidation.
Honesty rate jump reflects three OKRs re-scored from inflated 0.9 to honest 0.6 after pricing-tier adoption review.

NEXT QUARTER FOCUS: Hold follow-through above 75% while pushing decisions per WBR closer to 8. Cap VP meeting time at 25 hrs/wk via Friday no-meeting block.

That slide takes a board member sixty seconds to read. They walk out knowing what the CoS function did, what changed, and what's next. That is the deliverable.

If your slide has nine metrics and three charts, you don't have a scoreboard, you have a dashboard nobody reads. Cut to five. Earn the right to add a sixth.

Vanity Metrics to Avoid

Five traps I've watched CoSes fall into.

"Meetings reduced by 22%." Useless without a follow-through baseline. If you cut meetings and decisions still happen, great. If you cut meetings and decisions just moved into Slack threads where nothing converges, you didn't reduce overhead, you hid it.

"Alignment scores up 14%." Survey scores without behavior change are placebo. If alignment is up but follow-through is flat, the survey is measuring vibes, not operating health.

Slack threads closed. Counting closed threads says nothing about whether decisions in those threads shipped. A closed thread can mean "decided and shipped" or "everyone got bored."

Docs written. Document count is an input metric. The CoS who measures themselves on docs written is the CoS who writes more docs than anyone reads. Read-rate matters. Read-rate is hard to measure honestly. So don't put doc count on the slide.

"CEO time saved." Self-reported CEO time savings is the most over-claimed CoS metric in the industry. The CEO will tell you in a 1:1 that you saved them five hours a week. Their calendar will say something different. Don't put it on the slide unless you have calendar data backing it.

What This Means for Your Role

The CoS who can't put five numbers on a slide is the CoS who gets cut first when burn becomes the conversation. Not because the work isn't real, but because the row in the CFO's spreadsheet stays blank. A blank row gets cut.

Build the scoreboard before you need to defend it. Run it for one full quarter before your first QBR with these numbers, so the second QBR has a comparison. By the third QBR, you have a defensible trend, a diagnostic story, and a forward-looking commitment.

That's the difference between "CoS function" being a label and being a measured function in the operating model. The job security isn't the work. It's the scoreboard for the work.

Learn More

• How to Run a Weekly Business Review
• The CoS Decision Log Template
• Killing Meetings the CEO Forgot to Cancel
• Business Operations Manager JD