Seven dashboards, and only then did I see the leak

I once spent a single long day building seven operating dashboards. What came out of it was uncomfortable, and in hindsight entirely predictable.

At Cadenza, where I'm COO, nothing was broken. That's how it felt, anyway. Revenue was growing, the team was delivering, customers were coming in. There was just that vague sense that you had to look in too many places at once to know how things were going, and that every question anyone asked cost half a day of digging.

So I sat down and built seven dashboards on the data that already existed. A cockpit, plus commercial, onboarding, customer health, revenue and retention, cash conversion and engineering delivery. No new system, no migration, no three-month data warehouse project. Just reading what was there and putting it side by side.

What surfaced

Four things, none of them a secret. They simply weren't visible in one place.

• Phantom deals. There were deals in the pipeline that effectively didn't exist. Not out of bad faith, but because nobody had ever agreed when something has to leave the pipeline. The forecast was structurally too optimistic as a result, and so was everything planned on top of it.
• A leak in billing. Credit notes were being issued systematically, and the cause sat neither with a customer nor an employee but in a step of the process itself. Everyone resolved their own case. Nobody added them up.
• Re-billable receivables. Costs that could contractually be passed on, and that nobody passed on, because it wasn't in anyone's job description.
• A drop in product engagement. After April, engagement had fallen sharply. That had happened weeks earlier and nobody had noticed, because there was no screen where you would see it.

The common thread wasn't in the numbers

When I lined those four up, the pattern was immediately visible, and it wasn't a data pattern. All four were symptoms of the same structural problem: after the sale, nothing had an owner.

Before the signature, everything is clearly assigned. A deal has an account executive, a demo has an owner, a quote has a deadline. After it, that evaporates. The customer belongs to everyone and therefore to nobody. Billing is a process, not a person. Usage is something the product does, not something someone follows up.

Every leak I found sat in a place where no name was written.

That's the kind of problem you never find by measuring better within one function. Sales measures its pipeline, finance measures its revenue, product measures its usage. All three are correct, and yet there's a gap between them where money and customers fall through.

Why one day was enough

The reflex for questions like this is a programme. A data warehouse, a BI tool, a consultant, a quarter. That is almost always too heavy for the question underneath it.

What I needed wasn't perfect data. It was roughly correct data on one screen. A number that's ten percent off but that you see every week is infinitely more useful than a number that's exact and that nobody requests. Precision can come later. Visibility has to come first.

In practice that meant: read-only access to the CRM, billing, product analytics and project tooling, some Python to pull it together, and HTML to show it. No new licences, no integration project, no vendor dependency.

What I'd do differently

Two things.

One: I'd have started earlier. I felt I had to understand the operation before I could measure it. It's the other way round. Measuring is understanding. The dashboards taught me more about the company in a day than the six weeks of conversations before them.

Two: I'd have attached each finding to a name rather than to a dashboard. A leak you show is interesting. A leak you show together with who follows it up from tomorrow is solved. The first version only carried the number, and for the first few weeks not much happened with it.

If this sounds familiar

The test is simple. Can you say, within five minutes and without asking anyone, how much revenue lands this month, which customers are stuck in onboarding, and what work is open with no owner? If the answer is no, your operation probably carries the same kind of leak. Not because your team works badly, but because nobody was ever asked to see the whole.

It takes a day to find. Not knowing costs you every month.