None of this is new – so why is it so hard?

We ended the last blog with a question: you’re probably measuring more than ever, but is it really driving better decisions across the whole organisation?

We also promised to get to the heart of Metric-Driven Decision Making (MDDM), the way of working designed to carry those decisions all the way out into the organisation.

So let’s do that. But let’s also look at why so few actually manage to pull it off – and what’s now starting to change that.

MDDM IN BRIEF

MDDM is a way of working, not a platform. At its core are four questions:

• Which metrics to measure – and, just as importantly, which ones to leave alone. We start with the business’s North Stars and break them down into operational metrics that each function can actually influence. Before a metric earns a place in the tree, it has to pass three checks: can you define it and explain why it matters? Is there a clear owner? And do you know which data it rests on? If you can’t answer those questions, the metric isn’t ready.
• How they connect – the links from operational metrics all the way up to the strategic ones are made visible in a metric tree.
• Who owns them – every metric has an owner: a business leader accountable for both its definition and its outcome. The data team builds the metric; the owner interprets it and acts on it.
• How we follow up – through a recurring Weekly Business Review (WBR), where deviations are analysed and prioritised, with an owner and a next step assigned.

NOTHING NEW – AND STILL SO HARD

The interesting part here – none of what I’ve just described is new.

Metric trees, clear ownership, structured reviews – well-run businesses have worked this way for years. Which raises the obvious question: if the method is well known, why do so few manage to carry it through the whole organisation?

The answer is that it has always been time-consuming and complex – to build, to maintain, and to follow up. In our work, we see the same obstacles again and again.

When you build the tree and try to keep an overview:

• Too many metrics, no overview
  A few functions in, and you’re already looking at dozens of metrics across multiple branches. Zoom out far enough to see the whole tree, and you can no longer read the metrics.
• Hard to see what drives what
  Does this controllable metric actually move a North Star? The relationships are rarely obvious.
• Finding the right metrics is largely guesswork
  Knowing which metrics are worth testing requires a lot of business knowledge.

And when you run your WBR:

• The meeting drowns in noise
  Many metrics mean many deviations, but far from all of them are meaningful.
• Prepping the pack eats hours
  Pulling the numbers together and building the briefing before every meeting is painfully time-consuming.
• Spotting something isn’t the same as acting on it
  Identifying a deviation is one thing; agreeing on what to do next is where it often stalls.

It isn’t the method that’s the problem. It’s the threshold for actually living it, week after week.

WHAT IS NEW – AGENTIC AI

What’s new is AI. And when we say AI, we don’t mean a chatbot sitting on the side – we mean Agentic AI. Agents that can provide insights, take on tasks, and act, built to work on your data.

And that’s the key point.

Our clear recommendation is not to simply turn to a frontier model – Claude or ChatGPT – and start asking away. For insights and actions to be reliable, the agents need a data model, a semantic layer, and business context underneath. Without that foundation, they’re left guessing what your data actually means.

With that in place, the whole flow starts to change – and those same six obstacles become easier to overcome.

Within the metric tree, you can interact with your data in natural language: ask which deviations affect a given North Star, have an agent test whether the assumed cause-and-effect relationships actually hold, and identify where metrics may be missing from the tree.

In the WBR, AI can help separate real variation from noise, generate the briefing pack before the meeting, and suggest actions based on what the insights show, sometimes even as the meeting is happening.

On its own, each of these things might sound small. Together, they’re what lowers the threshold between knowing how you ought to work to actually being able to do it, week after week.

We don’t want to give the impression that this happens at the push of a button. But we do want to make clear that it’s achievable. We work with platforms such as Snowflake, Microsoft, and Qlik, all of which provide the capabilities needed to implement Agentic AI in the business, on top of a data foundation that holds.

WHAT HAPPENS NOW?

And that brings us straight to the next question.

Because everything we want AI to do rests on one thing: a data foundation, where every metric has a single definition and the right business context around it. Without that, the agents become unreliable – and the WBR gets stuck on: “But whose number actually counts?”

So in the next part, we’ll build that foundation: how to prepare and make the data available so that every metric rests on a shared, trustworthy truth, and how AI is changing the conditions there too.

But carry this thought with you until then: the problem has rarely been that you lack a method. It’s been making that way of working possible in practice.

And that barrier is coming down now.

/Jonas