Numbers are useful once you know what they're hiding

Data is very good at telling you where something is happening. A drop between two steps, a segment behaving unlike the rest, a page that everyone reaches and nobody leaves in the direction you expected.

It's much worse at telling you why. I've watched a lot of time go into constructing explanations from funnel charts that turned out to be completely wrong the moment someone sat with a customer for twenty minutes.

So in my experience the useful pattern is fairly simple. Use the numbers to find the place that matters, then use research to find out what's going on there. Doing it the other way round, starting with a hunch and looking for a chart to support it, is easy and mostly produces confirmation.

For a long time I found it uncomfortable to be asked what the design team had delivered in numbers. It felt reductive, and I think some of that discomfort was fair.

But I also came to realise that if I couldn't answer it, someone else would answer it for me, usually with a worse measure. Teams that can't describe their contribution tend to get resourced as a cost.

At Selfridges I put in financial tracking and resource planning partly for this reason. Knowing what the team cost, what it was working on and what came out of it made every conversation about headcount and investment considerably easier than it had been.

The financial elements of the L7 programme were the furthest outside my comfort zone, which is exactly why they turned out to be the most valuable part.

I'd been comfortable with customer metrics for years. Reading a set of accounts, understanding how investment decisions actually get made, seeing where a design initiative sits in a wider P&L conversation: those were things I could follow but not really speak.

That has changed how I frame work. I still care about the customer problem, but I now expect to be able to explain what it's worth and what it costs, and I'm much less willing to bring a case forward without that.

The failure mode I see most often isn't a shortage of data. It's a large amount of insight sitting in places nobody looks at while decisions get made from memory.

Part of the fix is editorial. Somebody has to decide what the three things worth knowing this quarter are, and be willing to leave the rest out. Dashboards that show everything usually mean nobody has done that work.

The other part is timing, which I keep coming back to. Insight arriving a week after a decision is just an argument. I'd rather have a rougher answer available when the choice is being made than a precise one afterwards.