There’s a rush to point AI at growth — personalised recommendations, lifecycle prediction, smart merchandising. All good ideas. All built on a foundation most teams haven’t checked in years: their measurement.
Here’s the uncomfortable version. A recommendation model is only as good as the events that tell it what people did. A churn model is only as good as your definition of “active.” An attribution-driven budget decision is only as good as the tracking that fed it. Point sophisticated AI at a broken event layer and you don’t get insight — you get confident, expensive guessing.
The symptoms are familiar
- Two dashboards disagree, and which one you believe depends on which team you’re sitting with.
- A conversion gets claimed by three channels at once.
- An event fires twice on one platform and not at all on another.
- The “single source of truth” is a spreadsheet someone maintains by hand on Fridays.
If any of those made you wince, the problem isn’t your model. It’s the data underneath it.
What good looks like
I spent years on this at Nykaa, through the messiest, fastest part of the growth curve. The teams that compounded weren’t the ones with the fanciest tools — they were the ones whose numbers reconciled. A clean event taxonomy everyone could read. Funnels where activation, retention and revenue tied out. A small set of dashboards wired to decisions, not vanity.
That foundation does two things. It makes your current growth work trustworthy. And it’s the precondition for any AI on top of it to be worth the spend.
Where to start
You don’t need a year-long data programme. You need an honest audit of how your most important events are defined and fired, a taxonomy your whole org can trust, and the handful of dashboards that actually change what you do this week.
Get that right and the AI opportunities tend to announce themselves — because for the first time you can see where the money is leaking.
That’s exactly the engagement we run. Tell us where your numbers stop being trustworthy, and we’ll start there.