The Predictability Need
Inside every company there are people who need to know when.
Someone runs billing and has to staff month-end close. Someone runs operations and has customers who were promised something. Someone in finance has a forecast to defend in front of people who don’t care what discovery needs. When they ask a product team for a date, they aren’t being unreasonable. They’re trying to keep their own promises to other people.
The Transform Sub is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Marty Cagan’s recent piece at SVPG, Predictability vs Outcomes, makes an argument I mostly agree with. The trade-off we’ve argued about for a decade, output versus outcomes, was never the real one. In his words: “The real trade-off is the relative priority of outcomes versus predictability.” Output is the symptom. The desire for predictability is the cause. And in a time when AI makes output feel nearly free, a company that rewards predictable shipping will simply ship more of the wrong thing, faster.
He concedes that predictability isn’t inherently bad, and that some situations genuinely require it. But that’s where the piece goes quiet. Those cases are exceptions to be managed inside an outcome-first culture. The person who needs to know when appears mostly as an obstacle.
I’m beginning to wonder. I don’t think their need is the problem. I think we’ve been handing them the wrong kind of predictability, and they’ve accepted it because it’s the only kind on offer.
What people mean when they ask for a dateAsk the billing manager what she actually wants and it isn’t a date. It’s not being surprised. No scramble in the last two weeks of a quarter. No go-live that turns her team into a call center. No discovering in production that the thing doesn’t do what everyone assumed.
A date is a proxy for that. It’s the visible thing you can hold someone to on day one. The project model is very good at producing dates. What it can’t produce is the thing the date was standing in for.
Here’s a case where it went the other way.
A project that slipped and didn’t breakA team in my practice recently took a large vendor integration to general availability: document automation. A little over a year from the first vendor lunch to version one launch.
We never committed to a long-term date. We made high-integrity commitments we could defend within the next two to three sprints, and held everything beyond that as quarterly markers. We updated stakeholders every two to three weeks, and they weren’t watching from a distance. They were in the work with us.
We slipped by a quarter. We kept every high-integrity commitment we made.
Then it launched. Fourteen support tickets in the first week of hypercare, for a multi-system integration wired into approvals, disputes and document visibility across several groups. One was critical. Ten were closed by the end of the week, the critical one among them.
By the standard of a date-driven plan, the project missed. By the standard the billing manager actually cares about, it was pretty predictable.
Where the predictability wentIt didn’t disappear. It moved. To three places.
Earlier into discovery. During validation, nearly half our test documents produced no artifact at all. No error, no log, nothing. The tempting reads were vendor defect or a design gap. The PM traced it to an upstream formatting rule that silently rejected certain inputs. That surprise was going to happen to someone. Discovery made it happen in testing, to us, instead of in production, to the billing team. A quiet launch is mostly that: surprises you already paid for.
Into the system. After the fix, she pushed for observability so the next silent failure would announce itself. That’s predictability built into the product rather than promised about it.
Into the team. Mid-project, a key person left for another company. In a plan built around the people who wrote it, that’s a recovery event. Here it wasn’t, because decisions lived in briefs, stories and backlogs, and the daily standup kept shared understanding stale by no more than a day. The replacement onboarded against artifacts, not memory. The team stayed predictable even when its membership didn’t.
None of that shows up on a Gantt chart. All of it showed up in the first week of hypercare.
Two kinds of predictabilitySo here’s the distinction I’d add to Cagan’s framing.
The project model makes the plan predictable. The product model makes the result predictable.
Leaders choose the first because it’s visible immediately, while the second only arrives at the end. That isn’t irrational. It’s a reasonable response to uncertainty from someone whose own credibility rides on what happens next. But it means they’re buying a forecast and paying for it with the very surprises they wanted to avoid.
High-integrity commitments are how you serve the real need without faking the forecast. Commit only to what you understand well enough to defend. Hold the rest as honest markers. Then replace the comfort of a date with something better: a steady rhythm of updates, and stakeholders close enough to the work that a slip is never news. Our quarter-late launch surprised no one. They’d watched it happen, two weeks at a time, and knew why.
That’s the part I think the product world undersells. Trust doesn’t come from hitting dates. It comes from never being blindsided. A team that says “here’s what we know, here’s what we don’t, here’s what we’ll know in two weeks,” and then shows up in two weeks, earns more trust than a team that hits its date and ships a problem.
Cagan frames the choice between predictability and outcomes as a choice about culture and incentives. I agree. But I’d ask which predictability you’d prefer to reward. The kind you can see on day one, or the kind you feel on launch day?
Most of us reward the first and hope for the second.
The Transform Sub is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.


