Context Is Not Free

Last week one of my leads brought me a model for his team. Three levels. The habits a person controls at the bottom, what the team ships in the middle, the business result at the top. Every piece of work threaded to an outcome, so nobody ends the quarter having shipped ten things without knowing whether any of them mattered.

He is solving a real problem. Teams fall into activity. They hold twelve triages, close forty tickets, launch the dashboard, and the number the company cares about doesn't move. Tying work to outcomes is the honest fix. I agreed with the diagnosis on the spot.

And I still hesitated, at the top rung, where each engineer holds a business outcome. I pictured myself as that engineer, and then I remembered the platform team.

The outcome I couldn't give them

A year ago we had one team building integrations. They took requirements from the people who knew the customers and built what was asked. The only outcome that shape could hold was "build better integrations", which is not an outcome at all. Nobody could tell you when a customer would be live.

The business needed a new customer live in 90 days. That is a real number. It moves revenue. It is exactly the kind of thing the model says to thread down to the people doing the work.

I could not give it to them.

Handed "live in 90 days", a capable team does the obvious thing. They go to the customer with the loudest problem, work through the requirements as they change, and ship whatever gets that customer live. Then the next one. Every week of that is reasonable, and at the end of it there is no platform, because nobody was ever given the room to build one. The requirements churn would have eaten the abstraction.

So I split the job. One group took the customers, one at a time, with the changing requirements and the 90 days. The platform engineers took the connector every launch would run on, and one outcome: every launch gets faster than the last. The customer mess never reached them. I kept it away on purpose.

In the last two months we launched more customers than in the whole year before. The team that never saw the mess built the thing everyone else now stands on.

The instinct this contradicts

Everything I have read about leading good people says the opposite. Lead with context, not control. Give people the whole picture and they will make the right call. Netflix built a culture on it. Marty Cagan tells a story of Reed Hastings watching a manager make a decision he thought was awful, digging in, and finding it was reasonable given what the manager knew. Cagan's lesson is that the leader failed to share enough context.

I believe that story. I just think it was written by a big company.

Netflix's own preconditions for context over control are talent density, a bias toward innovation over error prevention, and loose coupling between teams. Read those again and you are reading a description of a large, stable company. Strategy that moves once a year. Teams that can be told everything because their work barely touches each other's. Enough slack that a person rethinking their week costs nothing.

None of that is true at a hundred and fifty people with a number to hit next quarter.

Why context isn't free

Three things are different here, and they are getting more different.

Capable people act on what they are given. In a high-agency team, context is an instruction. Tell a strong engineer about a problem and they will go and solve it, whether or not it was the one you needed solved. Too much context does not confuse them. It sends them, confidently, in the wrong direction.

Each person's output is now large. With AI in the loop, a single engineer's week produces what a small team's month used to. That is the whole reason we can launch as fast as we do. It also means a misdirected week costs what a misdirected quarter cost before, so alignment now decides whether the talent lands anywhere.

And the one thing moves. At any moment there is one problem that, once solved, makes the others easier. That is what I ask the team to find. But it changes, sometimes monthly. Context shared today has a half-life. Share the whole plan and every shift in it becomes a re-plan for everyone who heard it. People stop and think, "now I have to rethink everything I'm doing", and the thing they were doing was fine.

Big companies solve all three with layers. A director hears the strategy and decides what a manager needs. The manager hears that and decides what the team needs. The translation happens by default, and nobody calls it withholding. A startup has no layers, and AI has given it big-company output anyway. The translation still has to happen. The leader does it by hand, or it doesn't happen.

What I told him

Keep the model. The bottom two rungs are his team's. The top rung is mine to supply, not theirs to choose.

The difference is altitude. I can give a feature team "raise basket completion by ten percent". They own the lever, the bet is already made, the signal comes back in weeks, and the only way to hit it is to make checkout better. I cannot give them "raise revenue per customer by building a new product". Nobody on that team owns that lever. Hitting it means choosing what to build, which is the strategy, which is my job. Hand it down and a capable team will do the strategy for me, from a smaller window than mine.

So every person gets four things: the ask, the reason one rung up, the limits, and what comes next. They do not get the rest of the portfolio. The day they need it, I will give it to them, and until then it would only cost them. That is the translation a big company does with three layers of management, done by the one person who can see across the whole thing.

How I know when I've got it wrong

I watch for two signals. People surprised by a decision that would have made sense with one more rung of context, which means I held too much. And the other important things quietly rotting while everyone chases the one thing, which means I broadcast too well. Both happen. I watch the second one harder, because it is the one I am prone to.

Back to the platform team

They know now. They can see the launches running on their connector, and they can see the requirements churn they were never asked to absorb. I told them after the platform existed.