There is a specific kind of meeting where everyone in the room is correct and nothing gets decided.
I sat in one about a go-live date. The technology side had run the analysis and held that an earlier week was cleaner: fewer overlapping changes, better coverage, a lower-risk window. The operations side held that the earlier week did not work, because the people who had to be present for it to work would not be there. Both positions were supported. Both teams were green on their own measures. There was one date, and it moved.
Nobody in that room needed better information. Nobody needed to be more collaborative. What they needed was a number that both of them owned, and there was not one.
That pattern repeats. I have watched it inside a health system, a payer services business, a global technology platform, and a clinical AI deployment, and it almost never gets named correctly in the room where it happens. When two functions each hit their own targets and the shared outcome still fails, the defect is not the handoff. It is that there are two scorecards.
The reflex is to schedule more contact. A standing sync. A shared channel. A joint readout. Better alignment.
Put two scorecards in a room more often and you do not get a decision. You get better-informed disagreement, delivered more punctually.
A scorecard is not a report. It is the statement of what an organization actually pays for. When technology is measured on change risk and operations is measured on coverage, both teams optimizing hard against their own measure is not dysfunction. That is the system working exactly as designed. The design is the problem.
So the tell is worth learning. When you hear that a team needs better alignment, or that this is really a communication problem, stop and count the scorecards first. Most of the time the communication is fine. The measurement is broken, and communication is only where the breakage becomes visible.
Naming the problem does not fix it. A shared intention is not a shared measure, and the difference between them is whether anything changes on Monday. Two moves from the current work, both small, both instruments rather than intentions.
The first was a subtraction. We had been reading case count as a signal of whether a deployment was working. It is not one. Case volume moves for a dozen reasons that have nothing to do with the platform: referral patterns, staffing, seasonality, one high-volume reader on leave. A number that moves for reasons outside the thing you are measuring is not a measure, it is a mood. We replaced it with time per case, taken before and after cutover and filterable by platform. Operations and technology now read the same number, and neither side can quietly hold a different one.
The second was a framing decision, and the framing was the entire mechanism. We stood up an hourly per-site health metric: the share of studies drafted within a short window of image receipt. The threshold was the easy part. The hard part is that the same number can be presented as a measure of the system or as a measure of the person, and only one of those survives contact with a clinical audience. Framed as a productivity check on radiologists, the clinical side would have rejected it, and they would have been right to. Framed as a measure of whether the system is delivering work where it is supposed to, both sides can hold it and both sides can defend it to their own people.
A shared measure that one function cannot defend internally is not shared. It is imposed, and imposed measures get complied with rather than used.
Most organizations that believe they have a shared measure have two measures with one name on the slide. Three questions settle it.
Can both functions state the number from memory? If one of them has to look it up, it is not their measure, it is the other team’s measure that they agreed not to argue with.
Can either function be green on it while the other is red? If yes, you have two scorecards wearing one label. A shared measure has one value, and it is the same value in both rooms.
Does anyone’s week change when the number moves? If the number moves and no work gets reprioritized on either side, it is a report. Reports get read. Scorecards get obeyed.
If you fail those tests, the temptation is to split the difference between the two existing measures. Do not. A midpoint between two function metrics is a third function metric, and now you have three scorecards.
Start from the outcome both functions are actually in the building to produce, and work backward to a number that expresses it. In the go-live case that outcome was never the date. It was whether readers could work on day one, which neither the change-risk measure nor the coverage measure described.
Then do three unglamorous things. Give it one owner from each side, named, not the teams. Put it on a review cadence that is faster than the thing it governs, because a measure reviewed monthly cannot steer a weekly decision. And write down in advance what would turn it red, while everyone is calm and nobody is defending a position yet.
That last step is the one that gets skipped, and skipping it is how a shared measure quietly turns back into two. Once a number goes red and the definition is still open, both functions will argue for the reading that clears them, and both will be sincere.
The instruments above are small. The pattern is not.
At Prisma Health, clinical and operational leaders were put under one shared scorecard across 18 hospitals, 300+ practices, and 1.2 million patients. That structure is Dyad Leadership, and it produced $17.5 million in annual savings and moved patient satisfaction from 55% to 82%. Nobody in it was told to care more about the other function’s problems. The measure changed, and the behavior followed the measure.
The same shape shows up wherever two functions had been grading themselves separately. A value management office gave finance and clinical operations one governed pipeline instead of two competing views of what counted as savings, and committed $22 million. In a global technology business, support, workforce management, and account teams were put under one operating model with a single shared measure of account risk, and net promoter score moved from -52 to +20 across 4,000+ customers on a $400 million P&L.
Different industries, different decades, same move. Not better coordination between two measures. One measure, with two names on it.
The strongest objection here is worth stating properly, because it is correct as far as it goes. Collapse two measures into one and somebody stops being accountable for their own work. Technology and operations do genuinely different jobs. If the only thing either is graded on is a joint outcome, the specific craft of each function goes unmeasured, and unmeasured work decays.
That reading is right about the risk and wrong about the design. The shared measure sits on top of the function measures, not in place of them. Technology still owns uptime and change risk. Operations still owns coverage and staffing. What changes is that one outcome carries both names, and neither function can be green on that outcome alone. You are adding a row, not deleting two.
The second objection is sharper: you cannot share a number you do not both control. True, and that is the point. Writing down a shared measure forces an explicit argument about who controls which input, and that argument is exactly what a two-scorecard arrangement lets an organization postpone indefinitely. The discomfort of that conversation is not a side effect of the fix. It is the fix.
One honest limit. The two instruments from the current deployment are new and the verdict on them is not in. What is settled is the older work: those consolidations ran to completion and the numbers held. Read this as a proven mechanism with metrics still under test. Anyone telling you their brand-new metric is already proven is selling something.
In every one of these, both teams were doing what they had been asked to do. That is what makes it a design problem rather than a character problem, and it is why the fix is never a conversation about commitment.
If two teams keep hitting their numbers while the thing you actually care about keeps failing, stop trying to fix the handoff.
Go count the scorecards.
Next in the series · 90 for 90: How I Know a Turnaround Is Actually Complete