SDatta Dashbaord
Turning AI predictions into trusted decisions
Turning AI predictions into trusted decisions
Turning AI predictions into trusted decisions
match AI with daily supply chain operations
match AI with daily supply chain operations
My Role
My Role
My Role
Design Lead
Design Lead
Design Lead
Timeline
Timeline
Timeline
Q1 2024-Q2 2024
Q1 2024-Q2 2024
Q1 2024-Q2 2024
Team
Team
Team
1 Designer, 3 Data Scientist , 2 DevOps, CTO, CEO
1 Designer, 3 Data Scientist , 2 DevOps, CTO, CEO
1 Designer, 3 Data Scientist , 2 DevOps, CTO, CEO
Overview
Overview
My Role
Led product strategy and designed the UI/UX for the company's first product from
0→1, working with the founding team to transform static Excel exports into an interactive dashboard.
The Problem
Forecasts were based on outdated, unrelated calculations. The transformation to AI-powered predictions was crucial to catch expensive problems before they occurred.
The Challenge
How do we earn users' trust in AI-driven forecasts while making complex supply chain data simple enough for daily decision-making?
Impact
Served 10+ early customers
Reduced analysis time from days to minutes
Enabled daily decision-making with interactive forecasts
My Role
Led product strategy and designed the UI/UX for the company's first product from
0→1, working with the founding team to transform static Excel exports into an interactive dashboard.
The Problem
Forecasts were based on outdated, unrelated calculations. The transformation to AI-powered predictions was crucial to catch expensive problems before they occurred.
The Challenge
How do we earn users' trust in AI-driven forecasts while making complex supply chain data simple enough for daily decision-making?
Impact
Served 10+ early customers
Reduced analysis time from days to minutes
Enabled daily decision-making with interactive forecasts
My Role
Led product strategy and designed the UI/UX for the company's first product from
0→1, working with the founding team to transform static Excel exports into an interactive dashboard.
The Problem
Forecasts were based on outdated, unrelated calculations. The transformation to AI-powered predictions was crucial to catch expensive problems before they occurred.
The Challenge
How do we earn users' trust in AI-driven forecasts while making complex supply chain data simple enough for daily decision-making?
Impact
Served 10+ early customers
Reduced analysis time from days to minutes
Enabled daily decision-making with interactive forecasts
My Role
Led product strategy and designed the UI/UX for the company's first product from
0→1, working with the founding team to transform static Excel exports into an interactive dashboard.
The Problem
Forecasts were based on outdated, unrelated calculations. The transformation to AI-powered predictions was crucial to catch expensive problems before they occurred.
The Challenge
How do we earn users' trust in AI-driven forecasts while making complex supply chain data simple enough for daily decision-making?
Impact
Served 10+ early customers
Reduced analysis time from days to minutes
Enabled daily decision-making with interactive forecasts
Process
Process
As a solo designer in a fast-moving startup with no research budget, I had to find creative ways to understand users and validate decisions quickly.
As a solo designer in a fast-moving startup with no research budget, I had to find creative ways to understand users and validate decisions quickly.
What I Actually Did
What I Actually Did
Worked closely with founders to understand the problem space
Reviewed existing customer conversations
Analyzed customer files to understand their workflows
Conducted quick 15-minute validation calls with 3-4 users
Worked closely with founders to understand the problem space
Reviewed existing customer conversations
Analyzed customer files to understand their workflows
Conducted quick 15-minute validation calls with 3-4 users
Solutions
Solutions
Matching Product
to AI Readiness
Matching Product to AI Readiness
Matching Product to AI Readiness
Problem
AI forecasting was brand new in 2023. some clients needed proof of concept, others wanted full optimization. The company couldn't sell the same product to everyone when AI adoption itself was the barrier.
Problem
AI forecasting was brand new in 2023. some clients needed proof of concept, others wanted full optimization. The company couldn't sell the same product to everyone when AI adoption itself was the barrier.
My Solution
Tiered product architecture based on algorithm capabilities and client readiness.
My Solution
Tiered product architecture based on algorithm capabilities and client readiness.
Impact
Product-market fit improved—clients could start where they were comfortable.
Higher close rates and fewer feature objections in early sales cycles.
Impact
Product-market fit improved—clients could start where they were comfortable.
Higher close rates and fewer feature objections in early sales cycles.
AI transparacy
AI transparacy
Problem
Users rejected AI forecasts they couldn't validate
Problem
Users rejected AI forecasts they couldn't validate
My approach
Designed progressive disclosure: summary → drill-down into data sources, confidence levels, calculation logic
"Show me why" on any prediction
Color-coded confidence for quick scanning
My approach
Designed progressive disclosure: summary → drill-down into data sources, confidence levels, calculation logic
"Show me why" on any prediction
Color-coded confidence for quick scanning
Result
Users validated AI against domain expertise → trust in recommendations
Result
Users validated AI against domain expertise → trust in recommendations
Unified Filtering
System
Unified Filtering System
Unified Filtering System
Problem
Analyzing thousands of SKUs across dimensions = Excel chaos
Problem
Analyzing thousands of SKUs across dimensions = Excel chaos
My design
Hierarchical filters enabling multi-dimensional analysis (time / product / location / performance)
Multi-select combinations with visual chips and saved presets
Real-time result count showing impact of each filter
My design
Hierarchical filters enabling multi-dimensional analysis (time / product / location / performance)
Multi-select combinations with visual chips and saved presets
Real-time result count showing impact of each filter
Result
Enabled multi-dimensional analysis that was impossible in Excel
Result
Enabled multi-dimensional analysis that was impossible in Excel
Making AI
Recommendations
Useful
Making AI Recommendations Useful
Making AI
Recommendations Useful
Problem
AI suggestions were technically accurate but useless—they ignored business constraints, budget limits, and strategic priorities. Teams couldn't act on generic recommendations.
Problem
AI suggestions were technically accurate but useless—they ignored business constraints, budget limits, and strategic priorities. Teams couldn't act on generic recommendations.
My Solution
Strategy configuration layer that adapts AI outputs to business reality:
5-minute wizard to define objectives, constraints, and priorities
AI recommendations automatically filtered through user strategy
My Solution
Strategy configuration layer that adapts AI outputs to business reality:
5-minute wizard to define objectives, constraints, and priorities
AI recommendations automatically filtered through user strategy
Impact
Users shifted from ignoring suggestions to acting on them daily — recommendation adoption increased from ~10% to 70%+.
Impact
Users shifted from ignoring suggestions to acting on them daily — recommendation adoption increased from ~10% to 70%+.
Predictions Meet
Reality
Predictions Meet Reality
Predictions Meet Reality
Problem
The algorithm forecasted and recommended, but couldn't see what actually happened. Users missed critical signals when reality diverged from predictions.
Problem
The algorithm forecasted and recommended, but couldn't see what actually happened. Users missed critical signals when reality diverged from predictions.
Problem
The algorithm forecasted and recommended, but couldn't see what actually happened. Users missed critical signals when reality diverged from predictions.
Key Research Insight
Neither algorithm nor user has the full picture alone. The monitoring layer bridges both—algorithm can't see reality, users can't process thousands of SKUs.
Key Research Insight
Neither algorithm nor user has the full picture alone. The monitoring layer bridges both—algorithm can't see reality, users can't process thousands of SKUs.
My Solution
Monitoring layer that connects algorithm to ground truth
Alerts when predictions vs. reality diverge significantly
Tracks recommendation performance
Flags strategy conflicts with real-world data
My Solution
Monitoring layer that connects algorithm to ground truth
Alerts when predictions vs. reality diverge significantly
Tracks recommendation performance
Flags strategy conflicts with real-world data
Impact
Teams validated AI in real-time and caught when assumptions broke. The system learned from reality, not just history.
Impact
Teams validated AI in real-time and caught when assumptions broke. The system learned from reality, not just history.
Impact
Teams validated AI in real-time and caught when assumptions broke. The system learned from reality, not just history.
Impact
& Outcomes
Impact & Outcomes
Impact & Outcomes
Delivered
0→1 product in 5 months, 10 launch customers, contributed to seed funding
Business Results
Daily engagement - Teams shifted from weekly planning cycles to real-time dashboard monitoring
Decision velocity - Reduced analysis time from ~3-5 days (typical industry standard) to under 30 minutes
Product evolution - Design system scaled to support 3 post-launch features
Organic growth - Early adopters became active advocates, referring new customers
Delivered
0→1 product in 5 months, 10 launch customers, contributed to seed funding
Business Results
Daily engagement - Teams shifted from weekly planning cycles to real-time dashboard monitoring
Decision velocity - Reduced analysis time from ~3-5 days (typical industry standard) to under 30 minutes
Product evolution - Design system scaled to support 3 post-launch features
Organic growth - Early adopters became active advocates, referring new customers
What I Owned
& Learned
What I Owned & Learned
What I Owned & Learned
End-to-end product design
Strategy, UX/UI, and delivery for the company's first product—launched to 10 early customers in 5 months.
AI adoption is a design problem, not a data problem
The algorithm was accurate, but users didn't trust it. Building confidence required showing how it thinks, not just what it predicts.
Working backwards from constraints
Limited dev resources meant ruthless prioritization. Single-store focus over multi-network. Core features over nice-to-haves. Shipped fast, learned faster.
End-to-end product design
Strategy, UX/UI, and delivery for the company's first product—launched to 10 early customers in 5 months.
AI adoption is a design problem, not a data problem
The algorithm was accurate, but users didn't trust it. Building confidence required showing how it thinks, not just what it predicts.
Working backwards from constraints
Limited dev resources meant ruthless prioritization. Single-store focus over multi-network. Core features over nice-to-haves. Shipped fast, learned faster.
