Sumitomo Spyglass
A data-driven AI application that analyzes, monitors and predicts actions across global business vectors of impact.
Lead UX Designer | MVP Design Sprint | 9 Weeks
Overview
Spyglass is an AI-driven data visualization platform designed for Sumitomo’s advanced materials division. It enables predictive monitoring of production data and real-time detection of material stress and inefficiencies.
Problem
Capabilities for managing logistics, inventory and supply chain management were confined to an antiquated database with manual processes. Executives needed an optimal way to analyze data, monitor investments, and identify business opportunities to manage the companies they oversee.
Solution
Spyglass is a centralized business intelligence platform that ingests contextual real-time data based on user preferences. It leverages predictive data modeling to provide a financial data summary, facilitate supply chain root cause analytics, and present cost-saving and revenue-generating recommendations.
Process
• Learn
• Discover
• Define
• Design
• Test
• Discover
• Define
• Design
• Test
Deliverables
• Use Case
• Info Architecture
• User Flows
• Wireframes
• Interactions
• Info Architecture
• User Flows
• Wireframes
• Interactions
Team
• Data Engineer
• Data Scientist
• Web Developer
• Senior Designer
• Ul Designer
• Data Scientist
• Web Developer
• Senior Designer
• Ul Designer
Tools
• Confluence
• Jira
• PowerPoint
• Sketch
• Slack
• Jira
• PowerPoint
• Sketch
• Slack
Research
Stakeholder Interviews
To learn business and technical challenges, interviews and focus groups were conducted via Zoom with a Chief Financial Officer (CFO) and a Managing Director, who also acts as Chief Operations Officer (COO) to a subsidiary Sumitomo company. Both executives are strategic partners in conducting day to day business operations. Due to the large scale of the companies under the Sumitomo umbrella, an executive can hold more than one executive position. High-level pain points included:
To learn business and technical challenges, interviews and focus groups were conducted via Zoom with a Chief Financial Officer (CFO) and a Managing Director, who also acts as Chief Operations Officer (COO) to a subsidiary Sumitomo company. Both executives are strategic partners in conducting day to day business operations. Due to the large scale of the companies under the Sumitomo umbrella, an executive can hold more than one executive position. High-level pain points included:
• Current record keeping methods are not cost effective
• Unable to see global to granular data in real-time
• No way to leverage AI in critical decision-making
Competitive Analysis
Product capability, review count, collaboration, administrative features visualization, and cloud capabilities were considered when reviewing other products like Looker, Sisense, Tableau, Domo and Qlik.
Product capability, review count, collaboration, administrative features visualization, and cloud capabilities were considered when reviewing other products like Looker, Sisense, Tableau, Domo and Qlik.
Industry Research
To increase my domain expertise, I researched supply chain and corporate business management best practices and KPI quantifiers that had the biggest impact on global supply chains and corporate investment to correlate with external market data via the product design.
To increase my domain expertise, I researched supply chain and corporate business management best practices and KPI quantifiers that had the biggest impact on global supply chains and corporate investment to correlate with external market data via the product design.
Business Goals
• Maximize business processes cost efficiency
• Leverage AI to monitor and predict market volatility
• See holistic and granular views of business operations
Tech Goals
• Establish a corporate data foundation for AI insights
• Determine build out of capabilities
• Transform how businesses run via data supply chains
User Stories
"As a user, I want to see only the data relevant to my role and responsibilities."
"As a user, I want to see data on potentially critical events, so I can determine the best course of action ahead of time."
"As a user, I want to receive AI business recommendations, so I can know how a decision would affect the company."
"As a user, I want to monitor specific quantitative data margins, and have the system alert me when those margins are reached."
Features
AI Recommendations
Learns over time based on user KPIs and role-oriented actions
Learns over time based on user KPIs and role-oriented actions
KPI Customization
Streamlines the data pipeline based on role specific KPIs
Streamlines the data pipeline based on role specific KPIs
Watchlists
Monitors the performance of user-specific quantitative datapoints
Monitors the performance of user-specific quantitative datapoints
Industry News Feed
Aggregates articles and reports based on user preferences
Aggregates articles and reports based on user preferences
Anomaly Detection
Monitors KPI metrics of businesses, business groups and subsidiaries. Below is a demo of the Long Short-Term Memory (LSTM) Recurrent Auto Encoder Neural Network presented by Hypergiant Data Scientist Sanjeev Namjoshi to stakeholders that showcases the work that went into building, training and monitoring based on KPIs.
Monitors KPI metrics of businesses, business groups and subsidiaries. Below is a demo of the Long Short-Term Memory (LSTM) Recurrent Auto Encoder Neural Network presented by Hypergiant Data Scientist Sanjeev Namjoshi to stakeholders that showcases the work that went into building, training and monitoring based on KPIs.
User Flows
KPI Customization
Allows the user to select the indicators most relevant to their role and responsibilities. This process optimizes the data supply chain by delivering only the relevant data and information necessary.
Allows the user to select the indicators most relevant to their role and responsibilities. This process optimizes the data supply chain by delivering only the relevant data and information necessary.
Watchlists
Allows the user to select specific percentage (%) or price ($) limits based on qualitative data visualizations for monitoring by the system in order to gain foresight on potential upturns or downturns based on KPIs.
Allows the user to select specific percentage (%) or price ($) limits based on qualitative data visualizations for monitoring by the system in order to gain foresight on potential upturns or downturns based on KPIs.
Data Schema
Overall Business Performance
After intensive research on corporate finance, I created a data schema for a data model for monitoring and predicting business performance based on aggregated analytics (KPI data, financial and non-financial indicators, etc). Below is the data schema for the Failure Prediction Model, which scores businesses from 0% to 100% based on financial and non-financial indicators. A percentage threshold is set by the user, with 100% indicating total business failure.
After intensive research on corporate finance, I created a data schema for a data model for monitoring and predicting business performance based on aggregated analytics (KPI data, financial and non-financial indicators, etc). Below is the data schema for the Failure Prediction Model, which scores businesses from 0% to 100% based on financial and non-financial indicators. A percentage threshold is set by the user, with 100% indicating total business failure.
During stakeholder interviews, it was highlighted that financial executives would rather know when a business is underperforming than when it is successful, thus the name for the data model.
Wireflows
Below is the first iteration of the happy path experience for adding a datapoint to the Watchlist. Left side navigation, as seen in the final design, was created in later iterations.
I advocated for Sumitomo executives to become involved in the iterative design process, and annotated interactions of the experience for technical review.
General Symptoms of Failure
Metrics and Analytics: Business Failure
Modular Style Guide
Our Product Team also communicated with the Hypergiant Design Team creating a design system called the Sumitomo Standardized Modular Design System (SMDS).
The system specifies visual components and provides comprehensive documentation explaining how they should be used. SDMS improves design efficiency of software development at scale for Sumitomo.
Visual design components for Spyglass were based on the SDMS Style Guide. Components created specifically for Spyglass were added to SDMS as needed.
Visual Design
Based on my most simplified use case, below is an example of the user flow in high-fidelity for predictive data analytics as it applies to the Failure Prediction Model. This feature allows the user to instruct the system to monitor business performance based on a sliding scale from 0% to 100%, with 100% equaling total business failure.
In the example, the user sees that the Tubular Group has crossed a predetermined threshold of 50% to 52% (in red) due to factors correlating with a drop in the company's valuation.
The user views more details about the issue and is given an analysis of the cause. Upon review, the user decides that the best course of action is to increase the threshold to 54%, and continue monitoring business performance.
Product Impact
Spyglass enables a well-integrated user experience, and allows more efficient interaction between users from multiple companies and business functions. Consolidating data into a single repository, it provides the building blocks for machine learning and AI innovation. Some long term machine learning benefits include:
Evolved Analytics: Models predict of global events and help users to stay proactive
Action Recommendations: AI integration focuses on user actions and not just analytics
Unique Perspective: Granular data for users from the executive level to the subsidiary level
In Retrospect
This project was challenging due to the fact that it occurred during the beginning of the pandemic, and required collaboration by a team located in several areas of the world. My first project with Hypergiant, it was also a challenge to get onboarded quickly and get involved with the Spyglass project immediately. Some of the key takeaways from this project are:
Dive deep into the unknown with determination.
Ambiguity is normal, but the level is rarely predictable. Be ready to take a deep dive into a subject matter and learn as much as possible. Although stakeholder interviews are valuable, sometimes they provide insight to business problems and not business processes. In-depth research in a short time frame can be very challenging, so a sincere willingness to learn is an asset.
Ambiguity is normal, but the level is rarely predictable. Be ready to take a deep dive into a subject matter and learn as much as possible. Although stakeholder interviews are valuable, sometimes they provide insight to business problems and not business processes. In-depth research in a short time frame can be very challenging, so a sincere willingness to learn is an asset.
Demand design reviews at the testing phase.
Although I involved the engineers and developers early on in my design process, there were aspects of my designs that were overlooked due to time restraints. Design reviews are essential during the building and testing phase, but are challenging to uphold when working remotely.
Although I involved the engineers and developers early on in my design process, there were aspects of my designs that were overlooked due to time restraints. Design reviews are essential during the building and testing phase, but are challenging to uphold when working remotely.
Observations
Evaluation
Suggestions
