top of page
  • Writer's pictureSteven Doherty

Why I'm not a fan of Power BI's Smart Narratives

There’s been a lot of excitement over the release of the Smart Narratives visual for Power BI this month. I think its an excellent tool but, I’m not sure it’s as groundbreaking as being made out.

Firstly, it doesn’t work for all data (see the screenshot below). In this report, I categorise Customers into an A, B, C category using their current MAT sales. This measure is a calculated dynamically, and all other visuals use a derivative of it to obtain their values.


However, on a positive note, you can manually add comments and supplement them with measures, including those with dynamic categorisation in the commentary.

Nonetheless, its the nature of the commentary that concerns me more. My experience in management taught me that it’s good to identify a trend, but it’s better to know why it’s happening.

And therein lies the difference between human intelligence and AI. AI is fantastic at identifying connections, correlations and possibly causation with the data provided to it.

However, consider a narrative going to a CEO stating, “sales have increased x% in Channel A” or “GL Item X is overspent against budget by y.” The statements are all valid, but it’s not insight. The insight comes from the statement that sales are up in channel A because of the promotion of product x, new distribution with retailer “Z”, the launch of product “abc” or the overspend in GL is due to emergency maintenance requirements for plant “A”.

I could be wrong, and maybe AI can be trained to derive these associations with data. But, what if the causative events for sales is a periodic event such as an early winter or late summer or a catastrophic event (fire, floods, coronavirus lockdown?). I would suggest AI may be able to pick up that a KPI has increased or decreased for a sector or period but, it won’t know why.

So, before sending off your report to a CEO using Smart Narratives consider if you are answering the three questions management always asks of data and they are;

1. What are we doing well and should do more of?

2. What needs attention, what should we do less of? And;

3. What does the future look like?

and, I think you will find the Smart Narrative is not ticking these boxes and you're still required to identify this detail.

129 views0 comments
bottom of page