Video Transcript

Hello everyone! Welcome back, my name is Hobbs from Valorem Reply and this is This Not That, where we talk about BI best practices and interactions with clients where we've really been able to direct them to help make good decisions. Today we are going to be talking about when it is you should actually build a report and when it is you should build an engine for answering questions.

Welcome back everyone my name is Hobbs. Let me tell you another story. I was working with a client here recently and I had designed some reporting for them. [I] built out a variety of things about different topics [however], most of it was accounting based information. Usually for accounting things have to be very precise and organized and set up in ways that encourage auditors to be able to find things easily. But I introduced them to a capability which most modern BI engines are going to make available to you, just the idea of natural language querying. So, natural language querying allows someone to type in a question in English and get the results of that from the data. The computational engine on the background basically is taking that English, finding keywords like sum or average or the name of particular column or the name of a particular table, and translating that into something that the background data engine can give you a result for. So, you can say ‘what were my Q4 profits’, right? ‘Which of my companies are making the most money?’ ‘List my employees by their employee satisfaction.’ ‘Show me the truck with the most shipments moved.’ Whatever that might be, you're asking a question in English and retrieving a result.

Now, there's definitely a need for both of these things. You need reports. You need predefined questions, that's what a report represents. Someone ideally needs to make a good decision, they're going to go and find the data they need to make that decision and you present it to them in a way that enables them to quickly find, the best answer to whatever decision they need to make. Sometimes though, it's easier to give people access to all of the data and say, ‘now ask whatever questions you have.’ And I found this is particularly useful for managers. Managers, often they're interacting with a manager underneath them or an employee, an expert, a developer or whatever it might be, and in that interaction, they come away with a question in their mind. They go, ‘I wonder what this is?’ ‘I wonder if this is true?’ Now in the past, what that would mean is you would log a ticket with IT, or you would go to whatever analyst reports to you, they would go out to the database, they would create a report for you, and they would return the report to you. You would then filter through the report to find the thing that you were interested in. And that, right there, that's the sweet spot for natural language processing or natural language querying. Natural language querying enables that person, ad hoc, on the fly, to ask whatever question they want to ask. I'll give one caveat to this suggestion, and I do strongly suggest it, especially for managers, give them the capability to ask a question of their data without the need to go through you as the analyst or as the IT department to fulfill that need for them. Let them feed themselves if you will. And the only caveat is usually, in order for this to work smoothly, you're going to have to do some work on the back end. You’re going to have to work within your data model. Again, [Microsoft] Power Platform is where I spend most of my time. So, from a [Microsoft] Power BI perspective, you’re going to need to make sure that you’re setting up aliases for things and making proper relationships and all of that. Do your background work. Do your homework and then present to your manager the ability to say, ‘here is this vast connected network of tables and information, all I need you to do is type in the question you want the answer to and you can have that question right there in front of you.’

So, to recap, I recognize the importance of reports, they’re necessary, they’re a regular part of life for many people. But in some cases, what someone really wants to do is just be able to ask a question to the data and get the answer. They don't want to go through the process of building out an entire report or having other people build out the report. And in those instances, let them just ask questions. Do the background work necessary to build out that natural language processing capability and then set it loose. Let people ask whatever questions they want to ask in the simplest possible, you don't have to learn hardly anything, there's no need to code anything, environment.

Thank you all for joining me. I hope you enjoyed it this topic. Natural language processing is something I find really fascinating and very useful once you get people understanding how it works. If you're interested in bringing Valorem Reply in to help you with a project, to set up some natural language processing at your company, we'd love to do that. We would love to get involved in any way we can and partner with you to help you be successful. Until then, check out our website, leave some comments here on the video, and we will see you next time.