担当替えを伝える同じ1on1で、取締役会資料に載りかけた「MTTR 62%改善」の定義ずれをフィードバックする

"Thanks for making time, Wes. I've got two things today, about fifteen minutes each, and I want to say up front that they're unrelated — I'll say it again when we switch. First, a staffing change that starts today. Second, the MTTR numbers in the board deck."

"Marcus and Elena approved the CloudEdge verification layer on Friday. Starting today, you and Aisha move onto it for six weeks, through November 6th, and I'd like you to be the tech lead. Observability Modernization moves from mid-December to mid-January."

"It's the opposite of a vote of no confidence. The verifier only works if it has a second signal that doesn't come from CloudEdge, and that second signal is your synthetic checkout probes. The piece we need most right now is something you built for Observability."

"I won't pretend the slip is free — you've been selling this project to the team for months, and they'll ask you why it moved. The honest answer is that we have a Black Friday risk with a date on it and Observability doesn't. Observability isn't cancelled; it's where you go back to on November 9th, or on October 26th if the verifier fails its test on the 23rd. I've written both dates into the decision record, so this doesn't quietly become permanent."

"Okay — different topic, and not connected to the first one. On the 24th you sent Ravi the MTTR figures for the October 15th board deck, and draft two now says 'down 62%, 47 to 18 minutes.' The 47 includes the CloudEdge incident from January; the 18 excludes the one from September 15th. Nothing in the deck says the definition changed in August. The impact is that the board would be comparing two numbers measured two different ways, and one of the directors ran engineering at a public SaaS company until last month."

"You're right about the principle. There's a real case for tracking vendor-caused incidents separately — it's arguably the better metric, because it measures what we can actually fix. The problem isn't the exclusion. It's that Q1 was counted one way and Q3 another, and the slide doesn't say so. So pick either rule — I'm genuinely fine with both — as long as every quarter on the slide uses it. Include vendor incidents everywhere, it's 47 to 26, down 45%. Exclude them everywhere, it's 42 to 18, down 57%."

"I'd rather be the one who shrinks it now than the one who explains it in January. 57% on a consistent definition is still a strong story, and it's one we can defend line by line. If a director spots the mixed baseline before the Series C conversations start, the question stops being 'how good is reliability' and becomes 'what else in this deck was counted generously.' That's a much worse conversation for Hannah than a smaller number. One line of definition under the chart is what protects the story."

"By Wednesday noon, send me three things: Q1 through Q3 recomputed on whichever rule you pick, the other rule's numbers alongside it, and a one-line footnote stating the definition — something like 'Excludes incidents caused by third-party provider failures; applied to all quarters.' I'll review it Wednesday afternoon, and you send it to Ravi Thursday morning with me copied. I'll give Ravi a heads-up today that a corrected figure is coming, so he doesn't circulate draft two in the meantime. Going forward, any definition change on the dashboard gets a dated note on the dashboard itself."

"So, to recap: you lead the verifier from today, you're back on Observability on November 9th at the latest, and the corrected MTTR goes to Ravi on Thursday. That's a lot of change to land on one Monday. How are you actually feeling about all this?"