Data should change what happens next.

The most useful measure of an analytics function is not how many dashboards it ships. It is whether the business makes better decisions because that function exists.

That means understanding the decisions that matter, connecting evidence to commercial outcomes and giving analysts permission to challenge the question before answering it.

Natural-language access can take the journey to insight from days to minutes. But faster answers are only an advantage when definitions, permissions and underlying data are dependable.

The opportunity is not simply to add a chatbot to the warehouse. It is to combine governed metrics, good analytical judgement and thoughtful automation so more people can act with confidence.

Marketing can optimise acquisition cost while trading optimises gross margin—and the business can still lose money. Local targets do not automatically add up to a healthy commercial system.

Shared definitions, contribution economics and honest conversations about trade-offs turn competing functional incentives into a common view of performance.

At Babylon, a costly AI improvement seemed like the obvious answer to product drop-off. The evidence pointed somewhere else: customers were leaving before they reached the AI at all.

The job of experimentation is not to provide scientific-looking cover for intuition. It is to isolate the real constraint, test what matters and help product teams spend their energy in the right place.