cancel
Showing results for 
Search instead for 
Did you mean: 
Msutton
9 - Travel Pro
9 - Travel Pro

You didn’t buy Sisense just to answer one question. (Or, if you did, lucky you! You can stop reading right now.)

You bought Sisense because you understand that your company’s data holds the key to answering questions of all kinds and even evolving what you do and how you do it. Your data model is the foundation on which you’ll build your future success, so let’s talk about how you build a data model that answers today’s questions, tomorrow’s, and beyond — questions you could never foresee you’d have, but which may be instrumental for your company’s future success. 

In order to build a data model that scales, you need to do a few things: Get a handle on all the data at your disposal, then clean and prepare that data for analysis. Talk to your users about their needs and educate them about what analytics can do for them. Create a flexible data model that users of all skill levels can dig into, then you roll out your analytics and do your best to break whatever you’ve built — this will help you identify any gaps in your analyses and come up with better answers and a better data model that will scale to handle whatever your business throws at it!

 

Data is your second most valuable resource

Every business, regardless of industry, is becoming a data business. Companies are creating and collecting more data every day and storing it in a variety of places. While weaving all those datasets together is vital to answering deep questions and evolving a business, you can’t just hook a whole bunch of data warehouses into Sisense and hope for the best. 

Data needs to be cleaned and prepared before going into a data model so that it makes sense when users start asking questions. For instance, almost everyone wants “year over year” insights into some KPI, but picking the same date in two different years may not tell the story the user really wants to hear, say if one of those days is a Sunday and the other is a Monday. Little details like this can really undermine the success and accuracy of an analytics program. 

This is where involving your most valuable resource, your team, is critical. Talking to your users and understanding their starting point for business requirements and analysis will help guide your data modeling efforts.

 

Listening and educating are keys to adoption

Gathering basic business and user requirements is such an important step in building analytics, that sometimes people forget to talk about what makes that process work and how to improve upon it. For starters, you can’t build a halfway decent product for someone without listening to what they want out of it. 

Once you have a handle on the datasets at your disposal, learning what users want out of those datasets is a crucial first step. This is a great place to start asking questions that will lead to a data model that scales. Asking users to imagine “what’s next?” and other questions that might naturally follow on from whatever you’re initially trying to answer is a vital part of this process. 

Some user education here also goes a long way. Teaching your users a little about data architecture and how you’re structuring the data model that they’ll be using will help immensely. Of course you’ll identify more gaps later during the “try to break it” phase, but if your users are asking for a data attribute you haven’t identified during your dataset survey above, now’s a great time to identify that and find where it lives. Maybe there’s a strange naming convention you weren’t expecting or you didn’t know a dataset existed at all — whatever the circumstance, making your users your allies in this process can help ensure better results (and maybe even less stress for you). 

 

Make and break your data model

Now comes one of the most fun parts of making a data model that scales: Breaking your analytics! All the input you gathered in the previous phases should hopefully go a long way towards helping answer your users’ questions (at least at first!), but the whole point of this exercise is to answer future questions, including ones no one could have foreseen during the construction phase. 

Once you’ve got your flexible data model all set up, it’s time to sit down with your users and show them how to ask it questions. One of the great things about Sisense is that users of all technical skill levels can drill down into individual insights to learn more about the data behind them. With your data model set up, it’s time to let analysts and business users have at it and start clicking, asking questions, and hopefully breaking things. 

As mentioned in the last step: You can’t answer a question if the data to tell you what you want to know isn’t in the data model. There’s almost no way to perfectly ensure that you’ll have every scrap of data you need in your first data model, so keep pushing your users when they get their hands on Sisense. You want them to dig deep into all the questions they asked you to build this data model to answer, then you want them to go a little further. That won’t be the end of it (for you, them, or the data model), but it’s a really great start.

 

Constant refinement, constant growth

Hopefully this process serves as a great jumping-off point for you and your team when it comes to building flexible data models that scale. However, it’s just that: A jumping-off point. Datasets grow and change all the time. Cloud data storage space is so cheap that companies are adding more all the time, meaning new datasets that could enrich your model could be spun up without you ever knowing! Keep your finger on the pulse of what’s happening at your company and be sure to ask users “Is there something in this or that database that could help you do more?”

They may not initially say yes, but if you keep asking and keep encouraging users to ask more questions of your existing data model, you’re sure to get the answers you need to help them get the answers they need. It’s impossible to get this process 100% right the first time, but with a growth mindset and a few more rounds of exploratory questions, eventually you will build a data model that scales.

3 Comments