Trends & anomalies

To do their jobs efficiently, Sisu’s customers, data scientists, and business analysts, need to identify what drives and/or drags their businesses and then translating those findings into something more business friendly.

My challenge on this project was to find and create ways to encourage our users to try our new analysis offering and to prove to them that it would be valuable enough to warrant a role in their daily workflows.

Success also hinged on identifying and leveraging the following:

  • What data surrounding trends and anomalies would be helpful to surface to our users?

  • What kind of data visualization best tells the story of trends and anomalies in a way that is easy to glean insights from, easy to understand, and easy to share?

  • How can I use this part of the tool as a way to encourage customers to explore new areas of their dataset? Can this exploration drive Sisu’s business?

From a macro-level view, success for this project meant providing value to our customers. However, while providing that value, I also had to consider how this analysis type could increase adoption of our tool to ultimately drive business.

 

Step 1: Meet the user where they are in their workflow

The typical user’s workflow was coming into the tool and re-runing analyses they had already created with fresh data, or, uploading a new dataset and running a new analysis using the same analysis type.

This workflow gave me several areas of opportunity to introduce trends and anomalies to our customers. Some considerations were bringing trends and anomalies directly into other analysis types and giving the user the option to enable trends/anomalies within other analyses. Ultimately, the product team decided to make this offering a stand alone analysis type in an effort to encourage early adoption.

Step 2: Tell a compelling story, show value ASAP

Once we brought users in, it was important to demonstrate value immediately. If we could impress the user early with information that they found useful, compelling, and new, we’d be able to increase stickiness, ultimately, drive adoption, and earn a spot in the user’s workflow.

To approach the first goal of providing initial value, I had to consider:

  1. What information was needed vs. what made for an elegant visualization?

  2. What interaction patterns could keep users interested?

  3. Show the user that there’s more to their data than they may have realized (show them where they could explore outside of the trends analysis type)

Step 3: Start simple, add more if the user needs it

For the first iteration of this analysis, we decided to simplify the chart and only introduced trend lines, trend change points, and anomalies. It was important to first learn how users engaged (or didn’t engage) with this information, if they found it helpful, and if it encouraged them to explore their datasets more deeply. If not, we’d have to re-evaluate the approach.

We came away with some interesting finding:

  1. Surfacing trends and anomalies wasn’t always new information for customers and therefore, they didn’t engage with it

  2. Trend and anomaly insights didn’t warrant its own analysis type

While customers were using this analysis type, they weren’t spending as much time exploring their data as we had hoped. We recognized that trends and anomalies, while powerful, would have likely been adopted more quickly had we included it within existing analysis types as a an on/off switch.