AWS Supply Chain

Defining UX for scalable, data-driven demand forecasting

Demo / Prototype
View final designs
Role
Founding UX Designer
& strategy lead
Duration
15 months
Outcome
Successfully launched with 71% forecast adoption, 68% planner engagement, and 1 approved patent.
AWS Supply Chain

Business goal: improve supply chain efficiency through data-driven insights

Business Goals
  • Improve overall supply chain efficiency
  • Enable demand and supply managers to enhance demand forecasting
  • Optimize inventory management and resource planning
  • Empower informed, data-driven decisions that drive cost savings and performance

Demand planners lacked intelligent forecasting tools

To boost supply chain efficiency, I interviewed six demand planners from companies like Philips and Novartis to uncover forecasting and inventory challenges. The insights shaped a detailed persona capturing their needs and goals.

Customer Persona

Research revealed key data and forecasting challenges

The insights gathered from customer interviews helped identify opportunities to enhance their processes and support better decision-making.

Key Findings

Design thinking workshop aligned teams on information architecture

I facilitated a design thinking workshop with 3 PMs, 2 Principal Engineers, 2 Senior Engineers, and 4 Engineering Managers across 4 work streams - data lake, forecasting, collaboration, and setup. This helped bring customer insights, AWS legacy systems, and canonical data model together, forming the foundation for both UX and technical architecture.

Information Architecture

Workflow visualized onboarding and demand planning

I created early wireframes to visualize how customers would onboard, set up a demand plan, add their data for models to predict forecasted demand, analyze forecasts on visually rich dashboards, and manually override demand when needed.

Enable Jane to easily upload and map her existing data files.

Enable Jane to easily upload and map her existing data files.

Let Jane quickly explore forecast predictions across products and override forecasts manually during known demand spikes like holidays.

Let Jane quickly explore forecast predictions across products and override forecasts manually during known demand spikes like holidays.

Help Jane configure forecasts accurately through guided,
user-friendly controls.

Help Jane configure forecasts accurately through guided, user-friendly controls.

Highlight data quality issues to be fixed.

Highlight data quality issues to be fixed.

Added explainability factors to earn trust with Jane

Added explainability factors to earn trust with Jane

Cross-functional collaboration accelerated development

The team worked on their respective work streams while I developed high-fidelity designs. I presented the narrative to Peter De Santis, SVP of AWS, and his leadership team, addressed feedback, and advocated to expand the design team to meet launch timelines.

Collaboration

Measuring success for adoption and engagement

I partnered with my PM to brainstorm Forecast Adoption Rate and Planner Engagement Rate as a key signal of how seamlessly our UX enables users to trust, understand, and consistently use the forecast. Apart from these, Forecast Accuracy metrics (MAPE, WAPE and Bias) ensured forecast quality from an engineering persepctive.

Success Metrics
Success Metrics 2

Aligning with design systems enabled scalable patterns

I collaborated closely with Amazon Connect and Central AWS UX teams to align on design language and patterns. I built on foundational patterns from Cloudscape Design System while iterating on new ones.

Design System

Finalizing high fidelity designs to start development

While the individual work streams started functional development with the wireframes, the final designs and specs helped them execute the front end for the workflows.

Settings
Forecast

Customer feedback refined forecasting capabilities

While the sales team presented demos to companies under NDA, I facilitated 1:1 task-based studies with Amazon.com, Fabric.com, Philips, and Novartis. Key enhancements included locking periods for configuration, forecasting method selection, and bulk overrides with cascading across product hierarchies.

Locking Period Configuration

Feedback 1

Forecasting Method Selection

Feedback 2

Bulk Overrides

Feedback 3

AWS re:Invent demo

Successful launch achieved planned goals

Outcome

Hierarchical overrides pattern earned 1 approved patent

The hierarchical overrides concept, which emerged from our information architecture work, earned an approved patent. This innovation enhances flexibility and control in complex supply chain systems.

Patent

Reflections

Through this experience, I learned how to break down complex problems into manageable parts, design scalable solutions, and think beyond immediate challenges. Embracing a systems-thinking approach helped me understand how individual components interact within a larger framework, ensuring sustainable and impactful outcomes.

Other Case Studies

Let's Connect

I'm always open to discussing new opportunities, collaborations, or just chatting about design. Feel free to reach out!

divs.hariharan@gmail.com