Nandhini SelvarajProduct Designer
← Back to all work
Case Study

From Reactive Inventory Tracking to Intelligent Decision-Making

Designing an AI-assisted decision system for supply chain operations — turning static inventory data into proactive, explainable recommendations.

RoleUX Lead
TimelineFlexible
TeamProduct Team
Year2026

Project Summary

Inventory teams were struggling to identify and act on stock risks before they impacted customers.
High-demand products frequently went out of stock, resulting in:

  • Order cancellations
  • Lost revenue
  • Increased support workload
  • Potential Amazon marketplace penalties
  • Poor customer experience

I redesigned the inventory monitoring experience into an intelligent decision-support system that helps teams identify high-risk products, understand business impact, and take action before stockouts occur.

The Business Problem

A growing number of customers were placing orders on Amazon for products that appeared available but were actually unavailable internally.
As a result:

  • Orders had to be cancelled
  • Customers received refund notifications
  • Support teams handled avoidable tickets
  • Revenue opportunities were lost
  • Marketplace performance was impacted

Although the issue initially appeared to be an inventory synchronization problem, deeper investigation revealed a broader operational challenge.

Research Approach

To understand the root cause, I conducted:

Support Ticket Analysis

Reviewed cancellation-related tickets to identify recurring patterns and common failure points.

Interviews & Discussions: Spoke with 10–15 stakeholders across Customer Support, Operations, Inventory Management, Marketplace Operations, and Product Teams to understand their workflows, challenges, and perspectives on inventory-related issues.

Key Insight

  • Different teams viewed the problem from different perspectives, but a common pattern emerged across all groups.
  • Teams were spending significant time analyzing inventory reports and reacting to stock issues after they occurred.
  • Although large amounts of inventory data were available, teams struggled to identify which products required immediate attention.
  • Users found it difficult to understand the business impact of inventory risks and determine the appropriate next steps.
  • The core challenge was not a lack of data, but the need for faster, clearer, and more confident decision-making.

Workflow Observation

Observed how inventory and support teams identified, investigated, and resolved stockout issues.

Customer Journey Mapping

Research Findings

Finding 1 :

A Few Number of Products Created Most Problems

Analysis revealed that a small set of high-selling products generated a disproportionate number of stockout-related issues.

Impact : Prioritizing these products could significantly reduce operational and business risk.

Finding 2 :

Teams Discovered Problems Too Late

Inventory discrepancies were often detected only after customers had placed orders.

Impact : The organization operated reactively instead of proactively.

Finding 3 :

Data Was Available, Action Was Difficult

Teams had access to inventory reports but struggled to determine:

  • What action should be taken next
  • Which products required immediate attention
  • Which risks were business-critical

Impact : Decision-making relied heavily on manual analysis and experience.

Finding 4 :

Business Impact Extended Beyond Inventory

Stockouts created:

  • Lost Revenue
  • Customer Dissatisfaction
  • Support Costs
  • Amazon Penalties
  • Operational Inefficiencies

Impact : The organization operated reactively instead of proactively.

Key Insight

  • Teams did not need more inventory data.
  • They needed help identifying what mattered most and what action to take before customers were affected.
  • The challenge was not visibility.
  • The challenge was decision-making.

Existing Workflow

Problems

  • No prioritization — All SKUs looked equal in the table.
  • No risk visibility — Urgency was unclear without a forward-looking signal.
  • No business impact — Decisions lacked financial context.

Design Opportunity

How might we help inventory teams identify and resolve stock risks before they impact customers?

Design Principles

Prioritize What Matters

Surface the most critical risks first.

Reduce Analysis Effort

Replace manual investigation with actionable insights.

Build Confidence

Provide transparent reasoning behind recommendations.

Support Fast Decisions

Enable action directly from inventory monitoring workflows.

Reframing the Workflow

What if the system could not only show inventory status, but also predict risk, quantify impact, and recommend actions?

BEFORE
1

OOS List

2

Manual Check

3

Guess

4

Action

Reactive · Manual
AFTER
1

OOS List

2

Predictive Agent

Forecasts demand and assesses risk

3

Risk + Insight

Quantifies impact and prioritizes action

4

Confident Action

Recommended actions you can trust

Proactive · Confident

The Solution

A SKU-level predictive agent embedded directly into the inventory screen — turning a static table into a contextual decision surface. Selecting a SKU updates the right-hand panel with risk, impact, recovery, and supplier signals.

How It Works

📊

1. Inputs · Signals

What the system observes
  • Sales Velocity

    Trailing 30/60/90-day demand signal.

  • Current Stock

    On-hand units across warehouses.

  • Incoming ETA

    Open POs and expected arrivals.

  • Supplier Reliability

    Historical delivery performance.

🧠

2. Predictive Agent

What the system does
  • Demand Forecasting

    Predict future demand at SKU level.

  • Risk Scoring

    Identify stockout probability.

  • Recovery Modeling

    Estimate time to recovery.

  • Reliability Index

    Evaluate supplier performance.

📈

3. Outputs · Insights

What the system tells you
  • Days Out of Stock

    How urgent is this issue?

  • Revenue Risk

    What is the financial impact?

  • Recovery Time

    How long will recovery take?

  • Confidence Score

    How reliable is the prediction?

⚡

4. User Actions

What users can do
  • Prioritize SKU

    Focus on highest-risk items.

  • Trigger PO

    Create or expedite orders.

  • Notify Vendor

    Collaborate before delays occur.

  • Plan Recovery

    Monitor and manage mitigation.

Final Design

Inventory screen, augmented with a predictive agent.

Left holds the data. Right holds the intelligence. Selecting a SKU updates the panel — try clicking a row.

Key Features

Contextual intelligence

Selecting a SKU updates the predictive panel — focused, not overwhelming.

Revenue risk visibility

Quantifies business impact in dollars. Converts data into urgency.

Recovery time prediction

Shows how long stock will take to stabilize so teams can plan.

Supplier reliability score

Highlights vendor risk inline, enabling better procurement.

Explainability layer

Helper text, formulas, and reasoning behind every prediction.

System Flow

01 · INPUTS

  • Sales velocity
  • Current stock
  • Incoming POs
  • Supplier history
→

02 · PROCESSING

  • Demand forecast
  • Risk scoring
  • Recovery model
  • Reliability index
→

03 · OUTPUTS

  • Days OOS
  • Revenue risk
  • Recovery time
  • Supplier score
→

04 · USER ACTION

  • Prioritize SKU
  • Trigger PO
  • Notify vendor
  • Plan recovery

Outcome

Even when measured directionally, the impact was clear: faster, more confident decisions.

  • Faster decision-making for inventory operators
  • Reduced manual analysis per SKU review
  • Better prioritization of critical SKUs by revenue risk
  • Reduced risk of marketplace order cancellations

Key Learnings

  • Intelligence without explainability fails adoption.
  • Insights must be actionable, not informational.
  • Embedding intelligence beats creating dashboards.
  • Trust is a UX problem, not just a data problem.