The most useful thing I built at ZenCabs was a meeting. Every week the founder and I went through 15+ KPIs, and while it ran, driver-side cancellations fell from 40% to 6% and fleet utilisation rose from 40% to 75%.
The team made those moves happen. The review made sure each number was seen every week and owned by one person, and that a number moving the wrong way produced a decision while the problem was still small. This is how I would set one up again.
What a review does that a dashboard cannot
A dashboard shows information to whoever opens it. A weekly review guarantees that somebody looks, on a fixed day, and that a red number leaves the room with an owner and a date. Without the review, a problem surfaces when it can no longer be ignored. With it, the longest any number can drift unnoticed is a week.
Pick the number the business runs on
Every review needs one number at the top that the others explain, and choosing it is a strategic decision. At ZenCabs the obvious candidate was ride count. We chose utilisation, because on an owned electric fleet every idle car is capital sitting still, and revenue per car is what compounds. Ride count can grow while utilisation falls, if cars are added faster than demand.
The test I use: if this number rises and nothing else changes, is the business clearly better off? If the answer needs a caveat, the number belongs lower in the tree.
Build the tree underneath it
Break the top number into what drives it, break those down once more, and stop when every number has one obvious owner.
| Level | In a ride-hailing fleet | Owned by |
|---|---|---|
| Top number | Fleet utilisation | The founder |
| What drives it | Driver hours available, rides requested, rides completed | Supply and demand leads |
| Inputs | Driver partners onboarded, cancellation rate, repeat riders | Whoever runs each process |
| Guardrails | Rider complaints, driver churn | Someone with no incentive to trade them away |
Guardrails are the row most teams leave out. A team judged on completed rides can push that number up in ways that hurt the business, by overworking drivers or cutting corners with riders, and the guardrails are how you notice before the damage shows up in revenue.
Give every number a threshold
A measure with no number next to it is decoration. For each metric, write down the level that counts as on track and the level that forces a conversation, so the meeting can skip everything that is fine.
I hold every metric to the three tests in the measure step of my Push and Absorb framework. It has to move early, before the outcome it stands for. It has to survive someone chasing it directly, or come paired with a guardrail that catches the abuse. And it needs that threshold.
Run the meeting the same way every week
| Segment | Time | What happens |
|---|---|---|
| The numbers sheet | 10 min | Read against thresholds. Only red and amber items get discussed. |
| Red items | 25 min | For each one: what happened, why, what we will do, who owns it, and by when. |
| Stop rules | 10 min | Every change made with a written stop rule. If a rule has triggered, the change is reversed. |
| The slow list | 10 min | Numbers getting slightly worse every week without any single bad week. |
| Decisions | 5 min | Read back each decision and its owner. The list goes out within the hour. |
The slow list is the segment most reviews lack. A sudden problem gets attention on its own. A number that slips a little each week is never urgent in any particular week, so it needs a standing slot. My framework calls these slow, permanent forces, and they do the most damage because no alarm ever rings.
One page, same layout, every week
The written sheet matters as much as the meeting. Keep the same metrics in the same order, with this week, last week, the four-week trend and the threshold side by side. When the layout never changes, the founder can read it in two minutes and spot the one thing that moved, and every redesign costs a week of people relearning where to look.
How weekly reviews usually fail
The most common failure is trying to discuss every number. Track as many as the tree needs and talk about the red ones. Close behind is reviewing only lagging numbers like monthly revenue, which report what already happened when the meeting exists to change what happens next.
Actions without owners kill reviews slowly. "We should look into this" has no name and no date attached, so nothing happens and the same item returns next week. Skipping the review in good weeks does similar damage, because good weeks are when slow problems grow. And a review that turns into people presenting what they did has become a status meeting, which is a different and much less useful thing.
Who should run it
In an early startup, the natural chair is the person who sits closest to the founder without owning a single number of their own: a Founder's Office or Chief of Staff hire, or a strategy role like mine at ZenCabs. That person touches every function and has the founder's time, and having no number to defend makes them the most neutral person in the room.
What the review surfaced at ZenCabs, and what utilisation means for the economics of an electric fleet, is in the unit economics breakdown.
