This one I think is at the crux of a lot of the challenges: "50% of the use case of a dashboard is to download data back and analyse in Excel”
There are two core challenges that I've seen:
- The HUMAN challenge is that people often have different habits/styles of interacting with data. If you match that style, it's easy for them to consume. If you change up the style, it takes energy for them to adapt and so that threshold is often too much for an exec to bother with. Much easier to tell someone to cook their data just how they like it and email it to them. And if you please one person, you're displeasing someone else. A way out here is effective education and creating new habits that are SHARED in the org. That's not always easy and can take time, but I don't know a way around it yet. Perhaps with AI, the customized cooking can be done with some simple voice prompts to iterate quickly? ... but I'm jumping to solutions here.
- The BUSINESS challenge is that there are often many nuanced ways to slice a problem with data. And when we try to answer too many questions with a dashboard, it gets complicated to use for anyone but the power users. A dashboard may provide one default view, but it might be too blunt to answer a specific question. So then we create dashboards with a lot of bells and whistles, and 30 filters at the top -- which then makes it overwhelming and complex to use. Not just to choose the right filters to have the right data, but perhaps someone needs to see a trend vs a snapshot for the most recent week, etc.
this "customised cooking" is what has caught my attention over the last few months / year. The challenge, of course, is in figuring out exactly what to cook for whom.
and maybe this can help with the business challenge as well?
Yeah. But rather than having to guess what to cook, I wonder if we're now at a place where the user can do the customization with a few simple prompts. Like:
"I'd like to see sales as a monthly trend over the last 24 months, broken down by channel. Show it to me as a 100% stacked bar graph. Show data labels."
And perhaps they go through a couple of iterations, the way one may iterate with Midjourney until they get the image they want. And then the user saves that view as their custom view and they're set.
Feasible?
And if we can do this with typed prompts, then it shouldn't be that much of a stretch to do it with direct voice prompts.
Or another option would be for the user to sketch out their idea on paper, take a pic of it and upload it as supporting material for their prompt.
I come from an technical (ETL) background dealing with setting up the data for dashboards. Few comments-
1) Many dashboards go unused because of data quality issues. The ETL has failed to ensure the data quality and in some cases even the ETL cannot do much. There are other ways now to dead with data quality issues.
2) There is the concept of "self-service" dashboards. Some tools (like OBIEE) provide a logical layer exposing all data points, business users can pull in whatever data points they need in the report. Of course this works only for table style reports. Also, such reports are used a lot by middle-level managers and not the higher ups.
3) If the nightly ETL jobs get delayed, dashboards do not have the latest information. This is one of the main frustrations of business.
4) Business are usually happy with summary reports. I don't think they use drill-downs much.
As I said, I come from a tech background, but if you think there is value talking to me, then please reach out to me at mukundhan@gmail.com.
The data quality issues was something I'd missed from my list. Also I realise that the more the dashboards you have, the harder it is to keep all of them "fresh" and the bigger this dat quality probelem.
On the self service dashboard thing - do business users actually have the willingness / enthu to make their own dashboards?
I have seen mid-level managers in Operations team use the self-service capability. Many times they are looking for specific data points and are happy to set them up as a report and get the report sometimes emailed to them everyday in the morning. The email report becomes the starting point of their work every day.
This is a really great list!!
This one I think is at the crux of a lot of the challenges: "50% of the use case of a dashboard is to download data back and analyse in Excel”
There are two core challenges that I've seen:
- The HUMAN challenge is that people often have different habits/styles of interacting with data. If you match that style, it's easy for them to consume. If you change up the style, it takes energy for them to adapt and so that threshold is often too much for an exec to bother with. Much easier to tell someone to cook their data just how they like it and email it to them. And if you please one person, you're displeasing someone else. A way out here is effective education and creating new habits that are SHARED in the org. That's not always easy and can take time, but I don't know a way around it yet. Perhaps with AI, the customized cooking can be done with some simple voice prompts to iterate quickly? ... but I'm jumping to solutions here.
- The BUSINESS challenge is that there are often many nuanced ways to slice a problem with data. And when we try to answer too many questions with a dashboard, it gets complicated to use for anyone but the power users. A dashboard may provide one default view, but it might be too blunt to answer a specific question. So then we create dashboards with a lot of bells and whistles, and 30 filters at the top -- which then makes it overwhelming and complex to use. Not just to choose the right filters to have the right data, but perhaps someone needs to see a trend vs a snapshot for the most recent week, etc.
this "customised cooking" is what has caught my attention over the last few months / year. The challenge, of course, is in figuring out exactly what to cook for whom.
and maybe this can help with the business challenge as well?
Yeah. But rather than having to guess what to cook, I wonder if we're now at a place where the user can do the customization with a few simple prompts. Like:
"I'd like to see sales as a monthly trend over the last 24 months, broken down by channel. Show it to me as a 100% stacked bar graph. Show data labels."
And perhaps they go through a couple of iterations, the way one may iterate with Midjourney until they get the image they want. And then the user saves that view as their custom view and they're set.
Feasible?
And if we can do this with typed prompts, then it shouldn't be that much of a stretch to do it with direct voice prompts.
Or another option would be for the user to sketch out their idea on paper, take a pic of it and upload it as supporting material for their prompt.
Just spitballing here...
I come from an technical (ETL) background dealing with setting up the data for dashboards. Few comments-
1) Many dashboards go unused because of data quality issues. The ETL has failed to ensure the data quality and in some cases even the ETL cannot do much. There are other ways now to dead with data quality issues.
2) There is the concept of "self-service" dashboards. Some tools (like OBIEE) provide a logical layer exposing all data points, business users can pull in whatever data points they need in the report. Of course this works only for table style reports. Also, such reports are used a lot by middle-level managers and not the higher ups.
3) If the nightly ETL jobs get delayed, dashboards do not have the latest information. This is one of the main frustrations of business.
4) Business are usually happy with summary reports. I don't think they use drill-downs much.
As I said, I come from a tech background, but if you think there is value talking to me, then please reach out to me at mukundhan@gmail.com.
Thanks Mukundhan.
The data quality issues was something I'd missed from my list. Also I realise that the more the dashboards you have, the harder it is to keep all of them "fresh" and the bigger this dat quality probelem.
On the self service dashboard thing - do business users actually have the willingness / enthu to make their own dashboards?
I have seen mid-level managers in Operations team use the self-service capability. Many times they are looking for specific data points and are happy to set them up as a report and get the report sometimes emailed to them everyday in the morning. The email report becomes the starting point of their work every day.