Bad news velocity is the go-to-market metric nobody tracks

Bad-News Velocity

Deloitte put a simple question to 64 health system and health plan CFOs this year: what happened to all those AI initiatives you launched over the past two years? The answers landed in near-perfect thirds. A third scaled. A third are still in pilot — which in healthcare is not a stage, it's a state of matter. A third were quietly taken out behind the barn. [1]

That's the weather. This newsletter is about who's flying the plane, and the plane is your company.

Read those thirds from the vendor side of the table. For every three logos on your pipeline slide, one is dying right now inside a customer's finance review you were not invited to, and nobody is going to call you about it. Which means the question is not whether your customers are disciplined. It's how long it takes that fact to travel from the person in your company who noticed it first to the chair you're sitting in.

Here's what took me too long to learn as an operator. Whether your product works is mostly settled by month four. The variable that decides your next eighteen months is how many weeks pass between the moment your champion stops believing and the moment your CEO finds out. Call it bad-news velocity. I think it's the most underrated operating metric in a digital health company, and it isn't set by your CRM, your QBR cadence, or the dashboard your Head of RevOps built that nobody opens. It's set by two things sitting in the CEO chair: whether the leader asks more than they assert, and what happens to the person who walks in holding the number nobody wanted.

Your customer will never tell you no. Neither will your team.

The customer part is well understood. Health system executives don't kill things; they extend them. The capital committee meets in March. The CMIO who used to reply in a day replies in a week, then two, and when he does reply it's the word "circling" attached to no object.

The part founders underrate is what happens next, entirely inside their own building.

Your account exec reads exactly the same signals you would. She writes "positive momentum, timeline pushed to Q1." She isn't lying. Her comp plan, her manager, and every instinct she has sharpened in this job tell her that's the correct sentence — and she's right, because she has watched what happens around here to people who write the other one.

Meanwhile your implementation engineer has known for five weeks. He raised it once, in standup, phrased as a question, and your VP of Product said "good flag, let's take it offline," and it went offline, permanently. Your customer success lead has a hunch she hasn't written down anywhere because the last time someone brought a soft signal to a leadership meeting it turned into a two-week fire drill and a reorg of her team.

None of these people did anything wrong. Each one made a locally rational decision. Collectively they built a company where the CEO is the last to know, and every one of those decisions is downstream of behavior the CEO modeled without noticing.

Amy Edmondson found the cleanest version of this thirty years ago, and the setting was a hospital, but the mechanism is any team anywhere. She expected the better teams to make fewer errors. The better teams reported more of them. [2] They weren't sloppier. They were willing to talk. Run that distribution across your own org chart: the team with the prettiest pipeline is very often the team that has gotten best at reading your face.

Humility, in the operating sense, has nothing to do with modesty. Plenty of humble-sounding founders run companies where the truth cannot travel. It's a much narrower thing — whether being wrong in front of you is survivable. Your people already know the answer. They worked it out in your first ninety days from about four data points, none of which you remember.

Curiosity is just asking while the answer still stings

The founders I watch climb out of purgatory are almost boringly consistent about one behavior: they keep asking questions during the stretch when the answers are unflattering, which is exactly when everyone else stops asking.

The part where I argue against myself

Boards fund conviction, not curiosity, and they're not wrong to. Nearly every company worth studying went through a stretch where the founder was publicly incorrect, ignored the feedback, and turned out to be early rather than delusional. A leadership team that treats every customer objection as ground truth ships a roadmap that's the arithmetic mean of the last ten complaints, which is a worse product than one bad decision made confidently.

And psychological safety, run badly, degrades into consensus-seeking, then into a workshop, then into a Miro board with 200 stickies that everyone agrees was a valuable conversation. Slow is also fatal.

I don't have a clean test for telling conviction from denial in real time. The best proxy I've found is disconfirming evidence: a founder with real conviction can state precisely what would change their mind, and has instrumented the company to notice if it happens. A founder in denial changes the subject, usually to the roadmap. That's imperfect. I know operators who can't articulate a word of this and are right anyway, and I've watched the whole discipline curdle into theater.

What I'd do tomorrow

Pull your last four board decks and find the slide where a number went the wrong way. Then trace it backward through your own company: who knew first, how many weeks did they sit on it, and what happened to that person afterward. If you and your board found out on the same day, your company's learning rate is capped at one person's field of view. Yours.

Investors can run the same test from the outside in about ten minutes, and it's a better use of a diligence call than the product demo. Skip the CEO — the CEO will be outstanding at this question. Ask the VP of Engineering, or whoever owns customer success, what the company changed its mind about in the last six months and who caused the change. If nobody can name a specific instance and a specific person, the diligence is finished and you can go get lunch.