What we used to call the Smart Narrative visual, is now just the Narrative visual, even though I do feel like it is a "Smarter" Narrative visual 😊 This always offered us a quick text summary of our reports and visuals. However, it now includes Copilot and it pulls data from the visuals on the report or those that are selected using the "Reference visuals" option, not the underlying semantic model. Below are the results of me exploring this visual to derive some insights, using a Power BI report I regulary use for blogs and demos.
When I started to use this, I immediately appreciated the reference tags provided in the text. They offer end users clarity on the narrative's origins, enabling them to verify and explore the data further. These tags are also helpful because they point to the relevant pages where the data for the narrative is located.
With this visual, three useful prompts are provided, “Give an executive summary,” “Answer likely questions from leadership,” and “Create a bullet list of insights.” I was genuinely impressed by its functionality. It seems particularly beneficial for users who prefer a brief overview of events over browsing through multiple visuals. Although the narrative might not always reveal new insights, it excels at compiling the main headlines into an easily readable format.
I expanded my exploration of this visual by requesting suggestions for improving the performance of the underperforming ATV. The suggestions I received, while somewhat generic, were nonetheless relevant. I'd also like to highlight the visual's compatibility with cross-highlighting and filtering features, even though I believe the official MS documentation says it is not compatible with cross-highlighting/filtering. For example, I might be the regional manager of the Midlands, I can concentrate on this specific area by selecting the relevant columns. Refreshing the Narrative visual after would provide me with a tailored summary and suggestions for the Midlands region.
I then continued my exploration of this visual as the regional manager for the Midlands. I sought the Narrative visual's assistance in identifying stores with high footfall, but poor conversion and sales figures. These are stores potentially missing out on sales and might benefit from enhanced customer engagement and sales closing skills training. However, notice the Narrative visual did offer some information, but mentioned "the data provided does not offer specifics about store footfall". This aligns with my earlier note about the visual using data from report visuals, rather than from the semantic model.
Due to the above, I expanded the scope of the Reference visuals from just the "Current page" to include other pages that contain a detailed analysis of footfall, conversion, and sales by store. Notice, that the reference visuals option showcased in the screenshot below now displays "23 visuals" instead of "current page." Despite this adjustment, the insights I anticipated did not materialise, even though the selected pages/visuals contain the relevant information.
I then added the Narrative visual into a Power BI page that contains the necessary detailed breakdown. Utilising the same prompt, it succeeded in highlighting the insightful information I was seeking, with Wolverhampton being notably prominent. Given that this report was constructed with dummy data, Wolverhampton's significance was anticipated. My objective was also to find out whether the Narrative visual would recognise and emphasise this point of interest.
Before moving on, one thing that I found interesting, in the previous section I asked the same prompt but from the home page and I did not receive the insight I expected. However, I retried running the same prompt after applying a filter on the region for the Midlands, and it did work. Do keep in mind, this feature of the Narrative visual with Copilot is still in Preview.
Continuing my exploration, I prompted the Narrative visual with a question about trending. Again, as the regional manager, I made recent team changes at the store. However, I did not get an answer to this from the Narrative visual, and it wasn't like the above case where I would have expected a result. Here, it's quite clear that the issue is the visual only works with data that is available in the visuals, regardless if it exists in the semantic model. Well, we don't have any visuals that show store performance over time and also, this report page is for a single week.
To change this, I added two additional visuals, "Sales by Weeks..." and Footfall by Weeks..." and I also selected Wolverhampton as a filter on the "Trending Analysis" page.
After the above change, I can see the Narrative visual did offer some helpful information to validate my assumption that this may have started due to the rota change in the Wolverhampton store.
To finish up, I wanted to share that the Narrative Visual with Copilot is only available if you are connecting to a Semantic Model that is hosted within a Premium or F64 and above capacity. All in all, I can see much potential with this visual and look forward to what else is to come once its out of preview.