Smart Assistant

Time: June 2017 - Oct 2017

Team: 1 designer (me), 1 product manager, 1 data scientist, VP of product, Engineers

Skill: User Research, UX/UI Design, Prototyping

Deliverables: Interactive prototype, High Fidelity Mocks, Specifications, Presentation

top.jpg

 

Overview

Smart Assistant is a new feature for CallidusCloud Lead to Money suite. It will use artificial intelligence technology to analyze data and provide suggestion / smart assistance to drive sales people's behavior and help them achieve their sales goal. It is a part of dashboard besides KPI information.

 

Research

What suggestion will be helpful for them?

A series of contextual user interviews were conducted. The interviews were intended to help the team understand what suggests / smart assistance we can provide. In design thinking sessions, stakeholders together analyzed research data. We used Elito method to help translate research data to ideas. At the same time, we discussed the feasibility of each possible idea. 

 

Ideation

Suggestions to Provide for this Release

We as a team voted for  8 kinds of suggestions we can provide for this release.

information architecture.jpg

Proposals

How to drive sales people's behavior?

Before I am going too far, I presented proposals to stakeholders as well as some internal sales people.

Key findings:

- While suggestion cards may update every day, KPI widgets information may stay the same for several weeks; the frequency will depend of how often the company run pipelines. 

- Suggestions can drive behaviors and it should be presented in a more obvious way.

- Users care about their attainment. To show the impact on attainments can motivate users to follow the suggestions. 

 

2nd Iteration

In this iteration, attainment information are exposed with the recommendations. When users hover on a card, it will show the impact on the attainment. After users mark done a card, the remaining card will slide in. 

Key findings:

- Users can not notice the animation when they hovered on the suggestion cards because they are visually too separated.  

- Users always have a main attainment which relates to their main commissions. 

- Users did not figure out there is an order from the layout of recommendation cards.

- The data visualization is confusing and it takes too much prominent space. 

- Users can not understand the animation after clicking "mark done" button. They wonder where the original card goes to and want to review that card later. 

- Users care about both their long term and short term goal. Some suggestions are only meaningful for long term goal.

Desktop HD Copy 17.png

Final Design

In the final design, users can choose their main attainment and review related suggestions. There are two sections, short term goal related suggestions and long term goal related suggestions. The potential deal value can motive users follow the instruction. After the suggestion has been marked done, it will show congratulations information and pop up the impact on the attainment. The original card will disappear and new cards will slide in to feed.

2 copy 24.png
 

Measure Success

- In the second round of usability tests, the EOU (Ease of Use) rating was improved to be 6.22 out of 7 while 5.25 is the passing score. This design has proven successful.

- Still in beta version but got good comments from customers.

 

Future

- Continue exploration and working with data scientists to identify other recommendations that users may need. 

 

Take Away

- While dealing with a complex product, technical constrains need to be considered at the beginning. Getting engineers and data scientists in the early design stage can help with better collaboration and save time. 

- Animation can help users understand what is going on. But complex animation may be confusing.