Aishwarya UX

Back to home

Reimagining ML Workflows: A UX Case Study on EngageML by SuperAlign

Simplifying Machine Learning for Business Users

Overview:

EngageML is a no-code machine learning (ML) data prediction platform designed to help businesses make intelligent, AI-powered decisions. While ML offers powerful insights, traditional workflows can be complex and overwhelming for non-technical users.

My Role: UX Designer
Key Focus Areas: User Research, UX Strategy, Interaction Design, UI Design
Duration: 3 Weeks

Client

SuperAlign - EngageML

Client

SuperAlign - EngageML

Client

SuperAlign - EngageML

Type

UX Case Study

Type

UX Case Study

Type

UX Case Study

Date

Nov, 2021

Date

Nov, 2021

Date

Nov, 2021

📝 Understanding the Problem

Challenges with Traditional ML Workflows many businesses struggle with:
✅ Complex ML setup requiring technical expertise
✅ Lack of clear insights into model performance
✅ Tedious, time-consuming processes leading to inefficiencies
These issues led to a frustrating user experience and lower adoption rates.

Problem Statement

The world is changing with emerging technology. Not only large businesses but also small scale businesses can seek an advantage of data in growing their business. Because of the high demand and high cost, not everyone can afford to hire a data science team. Hence small-scale enterprises are being pushed away from taking advantage of data.

Design Statement

The current data analysis process demands a huge investment on teams of professionals. Having a no-code intuitive platform for data prediction will ease the process as anybody can use it. This would help in accurately predicting future trends rather than guesswork.

📝 User Research & Insights

To understand user pain points, I conducted:
📌 6+ in-depth interviews with ML practitioners and business analysts
📌 Usability testing sessions on existing ML tools
📌 Task flow analysis to identify friction points

We performed one-on-one interview with the users. The majority of interview questions are open-ended, allowing for the collection of detailed information.

Respondents worked in bank and tech based MNC.

There are roughly 5-6 steps in data analysis process which might take 2 weeks to 6 months

Duplicate entries, spelling errors, data quality can be hampered and reduced by these errors.

Spends a lot of time cleaning the data

Factors like age, credit score, background history impact a lot on the customer’s personality and his likely behaviour in the future

Key Insights

Users feel overwhelmed by ML setup. Complex configurations made onboarding difficult.
Lack of guidance led to errors. Users struggled to interpret model performance.
Businesses need clear, actionable insights. Users wanted simplified dashboards with key metrics.

Personas

We mapped out user personas to align design decisions with real-world needs.

Meet Jay Sethi – A Credit Risk Manager in Need of Smarter Tools

Jay Sethi, a Credit Risk Manager in Hyderabad, ensures loans align with risk appetite and profitability. He evaluates hundreds of applications daily, but manual tracking and limited data access slow him down.

With the right solution, Jay can approve loans with confidence, minimize risks, and focus on strategic decision-making.

Storyboarding – A Day in Jay’s Workflow

This storyboard highlights the inefficiencies of manual processes and the need for an intelligent, automated solution to transform Jay’s workflow.

Jay would like to dig deep into new technologies to facilitate the completion of report analysis.

Jay would like to dig deep into new technologies to facilitate the completion of report analysis.

Jay would like to dig deep into new technologies to facilitate the completion of report analysis.

‘What if’ Scenario to find quick individual results is the key feature of EngageML.

EngagML allows to upload bulk data and the output at once.

One can also focus on getting comprehensive insights into machine learning models.

Users can delve into detailed information, including accuracy, precision, recall, and other essential metrics, empowering them to make informed decisions.

📝 Task Flows & IA

From task analysis to Information Architecture

Apart from the ease of use, the simplicity of the model training process set up was crucial. Analyzing the ML process, we were able to identify users’ biggest pain points and streamline the flow for the least amount of steps and highest completion ratio.