Data Trends

3 Tips for Crushing Your Next Analytics Project

I spend a lot of my time these days traveling to talk to customers about how to get the most value out of ThoughtSpot.  No solution or technology exists in a vacuum, and the conversation almost always comes around to their larger Business Intelligence and Data Warehousing initiatives, and in particular best practices.  

People are hungry for best practices.  I’ve had the fortune of learning from some of the best—AOL, Disney, Netezza, IBM—and I’ve seen the successes and failures of many different approaches.  Regardless of whether we’re talking about implementing a big data strategy, how to increase adoption of an analytics platform, or an initiative to increase customer retention through better information access, I’ve found that a few themes come up over and over again.

1. Embrace the journey

Gone are the days of waterfall projects.  I’ll admit it, there’s something reassuring about collecting thorough requirements, creating a detailed project plan, and building a glorious cathedral of knowledge.  But the problem is that we’re so focused on defining and building the final destination that we usually end up in the wrong place.  In reality there is no destination, it’s all a journey.  Give up the concept of getting it “right” and embrace the fact that frequent iterations of “better” will keep you focused on where your customers are going to be.

And speaking of customers...

2. Involve end users early and often

This seems self-evident—it just makes sense to involve the people who will use the platform, product or solution, doesn’t it?  Yet it always surprises me how end users are often not included in the design and build process.  Sometimes this happens because of organizational distance.  The end users are in a different building, a different functional group, sometimes even a different time zone.  Sometimes this happens because a process is mistaken for an interaction.  (We have a defined requirements process, so that’s how we’ll get input!)  In most cases though, the root cause is a focus on process instead of discussion.  I’m not recommending doing away with defined processes, but they’re not a substitute for bringing end users into the conversation.

At ThoughtSpot, we often recommend using our solution as a platform to bridge the gap between the end user vision and the reality of the information that’s available.  The best requirements are built when an analyst and an end user sit down together and look at the data that’s available. (Often this is the first time the end user has even had a chance to see their data up close.)  The result is always faster iterations, Involving the ultimate user in the conversation this way—by giving them direct access to the information they want using an interface that they’re all familiar with—makes them part of the ultimate solution and creates an opportunity for dialog that’s often overlooked when we’re all focused on process.

3. Building an MVP: Start small, but with value

Iterating and involving end users sounds good, but where do we start?  When I first joined ThoughtSpot, one of my biggest surprises was how much of what we’re used to in Business Intelligence is driven by a lack of access to information.  It’s normal for a typical BI report or dashboard to contain multiple charts, ten or more columns, and subtotals throughout.  Yet if you talk to business users, it quickly becomes evident that this format isnt actually how people wanted to consume information. It’s actually a response to the typical turnaround time for a new report request.  A typical end users dialog goes something like this:

User:  Can I find out how many units of product we sold last week, for the 3 highest volume regions?

Analyst:  Sure.  I should be able to schedule the work and have something for you in about three weeks.

User:  In that case, could you also add….

In other words, if you only get a new BI dashboard once a month, you're going to try and squeeze as many insights into that dashboard as you possibly can, creating bloated and confusing results.

Whether we’re talking about a report, a new system or a completely new analytic ecosystem, try to define the smallest unit of value, often called the “Minimum Viable Product” (MVP) and deliver it quickly.  A platform like ThoughtSpot that’s focused on getting valuable answers fast certainly helps, but it’s just as much about streamlined business process as it is about your solution.  The key is to define the smallest unit of work that has real business value, then to deliver it and iterate.  Sometimes you’ll even find that after you deliver that MVP, a lot of the other requirements the users had may actually change or even become obsolete.

At the end of the day it become a virtuous cycle. Start with your MVP, bring in users from the beginning, and make sure to always be iterating.  Happy searching!