MACHINE LEARNING PLATFORM
The Problem
This application is a personalized research assistant that combines machine learning with human expertise to provide key information about markets, prospects and investments. These keywords could be used to search information about the prospects which will help in decision making for better investments.
Here the problem was, the users were not able to teach the system in easy manner which increased the user abandonment of this product. The system was not intuitive enough to what to do, where am i, how to do and what's next
The Approach:
To understand the real problem, summative evaluation was performed with the existing system. Users were provided with key tasks to perform with the existing system.
Identified Key Tasks:
- Create learning and publish it against your prospects
- Edit learning
- Map keywords to the Prospects you wish to know about
During each task, the facilitator (myself) did remind the participant to think aloud, but will refrain from assisting with the task unless something requires further clarification. The observer will use a prepared notetaking form for each task to record the participants’ path, time spent per task, outcome, and relevant comments made during the think aloud. After all tasks have been completed, the facilitator will follow up asking the post-test questions identified for the sessions. The usability problems of the existing system were identified.
- The system was not intuitive enough to what to do, where I am, how to do and what's next.
- Too much of data presented. “I wish to see what I wanted to see”
- Map keywords to the Prospects you wish to know about
- Mental model mapping of the real world scenarios “I would like to teach my system like how I would do to my children”
- Help me with suggestions to make my system learn faster
The Solution
Re-designed the user flow, how the users will interact with the system. Have designed a conversational user interface where the system talks to the user and guides the user to finish their task. I have validated the user flow with the respective users (Refer below figure) using Focus group method.
Started building wireframes based on the user flow. Usability testing was performed with users. The metrics used where, objective measures were task efficiency (time taken to complete the task) and errors count; and subjective preferences on the satisfaction of the revised system.
Overall usability of the revised system was good. Users were satisfied with the proposed system. Especially they were happy with keyword suggestion analysis which made their tasks much easier.
