By Christian Freese ·
Most strategies don’t fail because the idea was wrong. They fail in the distance between the decision and the work.
A few years ago, while I was running Grocery and Retail at Uber, Osk called me. Osk had founded Cornershop, and he was unhappy with the quality of merchants customers were seeing on the platform. We talked it through, the problem was real, and that night we deactivated roughly 20,000 low-quality merchants. A problem became visible, reached someone who could act on it, and the system changed the same day.
A few years earlier, at the same company, I watched a different decision take more than a year to land. In much of Central Europe, the regulatory environment made it clear that partnering with taxis was critical, both to win public trust and to build credibility with regulators. Coming out of COVID, the case was even stronger: taxi supply was still sitting in the cities, and we needed it. The strategy, which we called Hailables, was clear at the top. Getting the organization to move was another matter. Centrally in Amsterdam and in the local teams, there was a deep belief that taxis weren’t Uber, that the product wouldn’t work, and that riders didn’t want it. It took well over a year for the strategy to turn into aligned teams, product and engineering resources, and shipped capabilities. The idea was right the whole time. It just spent a year in transit.
Same company, a few years apart. One decision took a night; the other took more than a year. The strategy wasn’t the difference. The distance between the decision and the work was.
Strategy lives in decks. Execution lives in calendars.
Between the two sits a translation layer: the goals, dashboards, reviews, roadmaps and routines where strategy either becomes work or quietly fades. That layer is what I mean by an Operating System, and most companies never design it deliberately.
I think of the distance it has to cover as organizational latency: the time between reality changing and someone with the authority to act actually changing the system. Every company runs on a picture of reality that’s somewhat out of date. The question is how old it is, and most companies never measure it. Much of what gets called an execution problem is really latency.
When I set goals, diagnose a problem or decide where to focus, I run through five questions almost without thinking: translate, read, assign, check and execute. Together they trace a signal from strategy to action, and each one is a place where latency enters.
Translate: strategy that never becomes math
The first delay comes when a strategy never becomes something people can manage. “Win in grocery” quietly becomes “grow selection,” and “grow faster” isn’t an instruction anyone can act on.
In micromobility, the number of bikes in a city was relatively fixed, often set by city contracts. Growth couldn’t come from adding vehicles. It came from how many of those bikes were actually available and bookable at any moment, and that depended heavily on how fast broken ones came back from repair. Once we saw that, repair throughput stopped being a maintenance issue and became a growth driver. A lot of operating businesses work this way: break growth into capacity, utilization and rate, and the drivers people can actually move become visible.
A strategy that never becomes math stays a slogan.
Translate, part two: belief
Translation has a second half that rarely shows up in planning documents. People have to believe the strategy, not just hear it. Hailables didn’t stall because the plan was unclear. It stalled because the people who had to execute it didn’t think it would work, and nobody moves fast on something they privately expect to fail. Every skeptic in the chain adds latency: a resource request that waits, a launch that slips a quarter, a team that complies without committing.
What finally unlocked it was repetition. We explained the story and made the business case probably a dozen times, until it stopped sounding like another talking point and people understood we meant it. And we celebrated every early result loudly, to create momentum the skeptics could see. Conviction spread one proof point at a time.
The slowest handoff in a company is often belief, and no dashboard shows it.
Read: is it becoming true?
Once you know what has to become true, you need to know early whether it is. Companies drown in dashboards while missing the handful of numbers that tell them their assumptions are wrong. The job is to find the signals that move first. If the quarterly review is the first time you learn you’ll miss the quarter, the signal existed weeks earlier and nobody was watching the right number.
Assign: who owns it?
Every important outcome needs one person who wakes up knowing it’s theirs. Plenty of people will work on it, but one person owns it. When product, operations, marketing and finance all shape the same result, the metric belongs to all of them and effectively to none of them, so nothing happens until someone is told to act. Clear ownership is one of the cheapest ways to cut latency.
Check: where does reality meet the plan?
A good weekly business review makes the organization look at the few things that have to be true, see where reality is drifting, and decide what changes. The question that matters in that room is simple: what did we learn that changes what we do? Run that way, the review is how the company senses what’s happening. Run as a reporting ritual, it adds a week to every signal.
Execute: decisions that don’t move the machinery
This is the step companies underestimate most. Leadership makes a decision, the meeting ends, and nothing in the machinery moves: the roadmap, the headcount plan, the incentives and the metrics reviewed next week are all the same as before. Hailables had this problem too. Long after the decision was made, product and engineering resources still hadn’t followed it. A decision isn’t real until roadmaps, budgets and behavior move with it. With the Cornershop merchants, the decision and the change to the system were the same act.
Measure it like a production line
When we started repairing bikes, we ran it like a traditional bike shop: one mechanic, one bike, one repair at a time. As we scaled, that model couldn’t give us the throughput we needed, so we rebuilt it like an automotive line. Bikes were diagnosed upfront, routed through set stations and scanned at every step. The timestamps showed us how long each stage took, where work piled up and where to move mechanics when capacity was tight.
We learned to manage repair throughput only once we timestamped the work. Most companies still don’t timestamp their decisions. They don’t know how long it takes for a signal to reach an owner, for the owner to decide, or for the decision to change the work. Measure those three intervals, and the bottlenecks in your operating system become as visible as a backed-up repair station.
Luck doesn’t scale
Go back to the Osk call. It worked because the right person saw the problem and reached someone who could act. That night, the system depended too much on luck.
AI will make more of those weak signals visible earlier. The ratings, refunds and complaints that prompted Osk’s call were already in the data; a system reading them every day could have put the problem in front of its owner weeks sooner, with the evidence attached. Someone would still have had to decide whether 20,000 merchants was the right cut and what it would do to selection. The machine gets you to judgment faster. It doesn’t make the judgment.
Companies that still take weeks to believe, decide and act will end up competing against their own latency. As information latency collapses, organizational latency has nowhere left to hide.
Faster isn’t always better
I’ve spent most of my career in regulated, physical and supply-constrained businesses, where bad speed is expensive and some calls deserve more evidence and more challenge. The goal is not maximum speed. It is minimum accidental delay: the waiting caused by unclear ownership, stale data, slow approvals and decisions that never change the work.
The Monday test
You don’t need a transformation program to find out how much latency your company carries. On Monday morning, ask a few people two levels below the leadership team four questions. What are the company’s top three priorities? Who owns the metric that matters most? What changed last week that could make us miss? How long did the last important problem take to reach someone who could act?
If the answers come back as three strategies, a committee, a new analysis and a number measured in months, you’ve found where your strategy is dying. The job is getting the distance between reality and action back to what it was the night Osk called.