About This Case Study
About This Case Study
About This Case Study
Part 1: How a simple client question led us from dashboards to whiteboards and the foundation of RCA.
Part 1: How a simple client question led us from dashboards to whiteboards and the foundation of RCA.
Part 2: How we transformed that business logic into flows, wireframes, and the final RCA experience.
Part 2: How we transformed that business logic into flows, wireframes, and the final RCA experience.
My Role
My Role
My Role
I worked as the Product Designer for Clevrr AI, leading the end-to-end design process for the RCA feature.
I worked as the Product Designer for Clevrr AI, leading the end-to-end design process for the RCA feature.
My responsibilities included problem discovery, user flow design, business logic mapping, wireframing, interaction design, high-fidelity UI design, and collaborating closely with founders and engineers to bring the experience to life.
My responsibilities included problem discovery, user flow design, business logic mapping, wireframing, interaction design, high-fidelity UI design, and collaborating closely with founders and engineers to bring the experience to life.
Part 1: Dashboards to whiteboards and the foundation of RCA.
Part 1: Dashboards to whiteboards and the foundation of RCA.
Part 1: Dashboards to whiteboards and the foundation of RCA.
When we started building Clevrr AI, our goal was simple: help D2C brands make faster and better decisions using their business data.
When we started building Clevrr AI, our goal was simple: help D2C brands make faster and better decisions using their business data.
Naturally, we built dashboards — lots of them. We developed sales dashboards, marketing dashboards, product analytics, campaign tracking, and operational metrics. Packed with filters, comparisons, trend analyses, and performance breakdowns, we genuinely believed we had covered almost every important business metric a brand could ever need.
Naturally, we built dashboards — lots of them. We developed sales dashboards, marketing dashboards, product analytics, campaign tracking, and operational metrics. Packed with filters, comparisons, trend analyses, and performance breakdowns, we genuinely believed we had covered almost every important business metric a brand could ever need.
The product looked complete. Then, one conversation changed everything.
The product looked complete. Then, one conversation changed everything.
The Question That Broke the Illusion
The Question That Broke the Illusion
A client came to us with a straightforward but critical problem:
A client came to us with a straightforward but critical problem:
“Our sales are down 30% compared to last week. How do we figure out why?”
“Our sales are down 30% compared to last week. How do we figure out why?”
At first, we thought finding the answer would be easy. We built the product ourselves, understood the data structure, and intimately knew the ecosystems of Shopify, Meta Ads, Google Analytics, fulfillment metrics, attribution systems, and campaign performance. We assumed we could find the answer in a flash.
At first, we thought finding the answer would be easy. We built the product ourselves, understood the data structure, and intimately knew the ecosystems of Shopify, Meta Ads, Google Analytics, fulfillment metrics, attribution systems, and campaign performance. We assumed we could find the answer in a flash.
But reality hit differently.
But reality hit differently.
We opened Clevrr and began jumping from the sales dashboard to marketing, then to product, customer, and fulfillment metrics. We started manually cross-referencing and comparing a massive web of variables:
We opened Clevrr and began jumping from the sales dashboard to marketing, then to product, customer, and fulfillment metrics. We started manually cross-referencing and comparing a massive web of variables:
•
ROAS
ROAS
•
Conversion rates
Conversion rates
•
Average Order Value (AOV)
Average Order Value (AOV)
•
Discount trends
Discount trends
•
Product performance
Product performance
•
Delivery issues
Delivery issues
•
Returning customer behavior
Returning customer behavior
•
Spend allocation
Spend allocation
Even with all this data readily available, locating the actual root cause behind the sales drop was painfully difficult.
Even with all this data readily available, locating the actual root cause behind the sales drop was painfully difficult.
An Uncomfortable Realization
An Uncomfortable Realization
We eventually found the answer, but it left us with an uncomfortable truth: The time it took us to find the root cause inside our own advanced analytics product was almost equal to the time the client’s team spent manually switching between spreadsheets, Meta, Google Analytics, and Shopify.
We eventually found the answer, but it left us with an uncomfortable truth: The time it took us to find the root cause inside our own advanced analytics product was almost equal to the time the client’s team spent manually switching between spreadsheets, Meta, Google Analytics, and Shopify.
That single moment completely transformed how we thought about analytics. We realized a fundamental truth: Businesses don’t struggle with seeing numbers; they struggle with understanding cause and action. Surprisingly, most analytics tools are terrible at answering the “why”.
That single moment completely transformed how we thought about analytics. We realized a fundamental truth: Businesses don’t struggle with seeing numbers; they struggle with understanding cause and action. Surprisingly, most analytics tools are terrible at answering the “why”.
That painful realization became the starting point for building our Root Cause Analysis (RCA) framework.
That painful realization became the starting point for building our Root Cause Analysis (RCA) framework.
The Whiteboard Sessions: Deconstructing the Revenue Engine
The Whiteboard Sessions: Deconstructing the Revenue Engine
The next few days looked completely different for our team. We completely stopped designing UI screens. Instead, we gathered around a whiteboard to manually map out how revenue actually moves inside a D2C business so we could systematically answer our client’s question.
The next few days looked completely different for our team. We completely stopped designing UI screens. Instead, we gathered around a whiteboard to manually map out how revenue actually moves inside a D2C business so we could systematically answer our client’s question.
We started with one core question: “Why does revenue go up or down?”
We started with one core question: “Why does revenue go up or down?”
Initially, we identified four major business drivers:
Initially, we identified four major business drivers:
•
Orders
Orders
•
AOV
AOV
•
Fulfillment
Fulfillment
•
Discounts
Discounts
But once we started digging deeper, we realized that every single factor opened up an entirely new layer of questions.
But once we started digging deeper, we realized that every single factor opened up an entirely new layer of questions.
The Branching Effect
The Branching Effect
For example, if Orders are down, you have to ask why? That question immediately forces you down several deep branches:
For example, if Orders are down, you have to ask why? That question immediately forces you down several deep branches:
•
Lower traffic on the website
Lower traffic on the website
•
Poor conversion rates
Poor conversion rates
•
Weak ad creatives
Weak ad creatives
•
Inaccurate targeting
Inaccurate targeting
•
Checkout friction
Checkout friction
•
Underperforming Meta campaigns
Underperforming Meta campaigns
If you follow the Meta campaign performance branch down further, the next question becomes: “Why is Meta performance dropping?” Once again, more sub-branches appear:
If you follow the Meta campaign performance branch down further, the next question becomes: “Why is Meta performance dropping?” Once again, more sub-branches appear:
•
Top of Funnel (TOF) campaigns are failing
Top of Funnel (TOF) campaigns are failing
•
Creative fatigue is setting in
Creative fatigue is setting in
•
Wrong audience targeting
Wrong audience targeting
This process kept repeating. Every single answer generated another “why,” and every metric revealed a hidden layer underneath it. Very quickly, our whiteboard stopped looking like a clean dashboard flow and started looking like a massive decision tree.
This process kept repeating. Every single answer generated another “why,” and every metric revealed a hidden layer underneath it. Very quickly, our whiteboard stopped looking like a clean dashboard flow and started looking like a massive decision tree.
That’s when the real insight clicked: Root cause analysis is entirely non-linear.
That’s when the real insight clicked: Root cause analysis is entirely non-linear.
Uncovering the Hidden Layers of Causation
Uncovering the Hidden Layers of Causation
Business metrics don’t live in isolation; they behave like a highly connected chain of business events. The deeper we investigated, the more we noticed a fascinating pattern: no matter where an investigation started — whether with Orders, AOV, Fulfillment, or Discounts — almost all paths eventually connected back to customer acquisition behavior, operational inefficiencies, or campaign quality.
Business metrics don’t live in isolation; they behave like a highly connected chain of business events. The deeper we investigated, the more we noticed a fascinating pattern: no matter where an investigation started — whether with Orders, AOV, Fulfillment, or Discounts — almost all paths eventually connected back to customer acquisition behavior, operational inefficiencies, or campaign quality.
Behind a simple drop in sales, the real culprit was often a complex, multi-layered issue, such as:
Behind a simple drop in sales, the real culprit was often a complex, multi-layered issue, such as:
•
Marketing Missteps: Scaling weak campaigns or suffering from creative fatigue.
Marketing Missteps: Scaling weak campaigns or suffering from creative fatigue.
•
Operational Friction: Inventory mismatches with active ads or delayed deliveries are hurting repeat purchase rates.
Operational Friction: Inventory mismatches with active ads or delayed deliveries are hurting repeat purchase rates.
•
Financial Erosion: Aggressive discounting reduces overall profitability or low-quality traffic, inflating ad spend.
Financial Erosion: Aggressive discounting reduces overall profitability or low-quality traffic, inflating ad spend.
Between the final sales drop and the actual root issue at the campaign or operational level, there were often 4 to 5 layers of contributing factors. This hidden complexity was exactly what businesses were struggling to untangle every day.
Between the final sales drop and the actual root issue at the campaign or operational level, there were often 4 to 5 layers of contributing factors. This hidden complexity was exactly what businesses were struggling to untangle every day.
The Paradigm Shift
The Paradigm Shift
That was the exact moment RCA stopped feeling like just another software feature; it became something much bigger.
That was the exact moment RCA stopped feeling like just another software feature; it became something much bigger.
We completely stopped thinking about building:
We completely stopped thinking about building:
•
More charts
More charts
•
More dashboards
More dashboards
•
More filters
More filters
•
More reports
More reports
Instead, I collaborated with the Engineering team around a single question: How can we build a system that thinks through business problems the exact same way an experienced human operator would?
Instead, I collaborated with the Engineering team around a single question: How can we build a system that thinks through business problems the exact same way an experienced human operator would?
We shifted from simply displaying data to actively connecting events, explaining causation, and drastically reducing investigation time. Our goal was to help teams seamlessly transition from asking “What happened?” to instantly knowing “Here’s why it happened.”
We shifted from simply displaying data to actively connecting events, explaining causation, and drastically reducing investigation time. Our goal was to help teams seamlessly transition from asking “What happened?” to instantly knowing “Here’s why it happened.”
This shift completely redefined the trajectory of Clevrr AI. Because at the end of the day, the future of analytics isn’t more dashboards — it’s intelligent systems that empower businesses to truly understand their decisions.
This shift completely redefined the trajectory of Clevrr AI. Because at the end of the day, the future of analytics isn’t more dashboards — it’s intelligent systems that empower businesses to truly understand their decisions.
Part 2: Wireframes and the final RCA experience.
Part 2: Wireframes and the final RCA experience.
Part 2: Wireframes and the final RCA experience.
1. We Had the Logic. Now We Needed an Experience.
1. We Had the Logic. Now We Needed an Experience.
After weeks of mapping revenue drivers and complex business relationships, we had built something massive — it looked like a giant, sprawling decision tree. While the underlying business logic was rock solid, it brought us face-to-face with a looming challenge:
After weeks of mapping revenue drivers and complex business relationships, we had built something massive — it looked like a giant, sprawling decision tree. While the underlying business logic was rock solid, it brought us face-to-face with a looming challenge:
How do we turn a complex backend decision tree into a product that human beings can actually use?
How do we turn a complex backend decision tree into a product that human beings can actually use?
2. Ground Zero: The Whiteboard
2. Ground Zero: The Whiteboard
Before we even dreamed of opening Figma, we lived on whiteboards. We drew relentlessly, mapping out:
Before we even dreamed of opening Figma, we lived on whiteboards. We drew relentlessly, mapping out:
•
Revenue trees
Revenue trees
•
Cause-and-effect branches
Cause-and-effect branches
•
Investigation paths
Investigation paths
•
User journeys
User journeys
The goal wasn’t to design pretty screens. The goal was to deeply understand the psychology of an operator — how a real user thinks and moves when they are frantically trying to uncover the reason behind a sudden sales dip.
The goal wasn’t to design pretty screens. The goal was to deeply understand the psychology of an operator — how a real user thinks and moves when they are frantically trying to uncover the reason behind a sudden sales dip.
3. Sketching Different User Flows
3. Sketching Different User Flows
Once the business logic crystallized, we transitioned to notebooks for rough, rapid-fire sketching. We forced ourselves to answer tough, structural questions:
Once the business logic crystallized, we transitioned to notebooks for rough, rapid-fire sketching. We forced ourselves to answer tough, structural questions:
•
Where exactly should a user start their investigation?
Where exactly should a user start their investigation?
•
How many levels deep should we allow them to go?
How many levels deep should we allow them to go?
•
How do users seamlessly transition between different root causes?
How do users seamlessly transition between different root causes?
•
What happens when multiple distinct factors contribute to the same problem?
What happens when multiple distinct factors contribute to the same problem?
•
How do we prevent users from getting utterly lost in a sea of data?
How do we prevent users from getting utterly lost in a sea of data?
At this stage, speed mattered infinitely more than visual aesthetics. We weren’t designing pixels yet; we were entirely focused on the flow.
At this stage, speed mattered infinitely more than visual aesthetics. We weren’t designing pixels yet; we were entirely focused on the flow.
4. Defining the Core Functionalities
4. Defining the Core Functionalities
Only after the foundational flows were completely locked in did we begin listing actual features. We approached feature definition by asking operational questions:
Only after the foundational flows were completely locked in did we begin listing actual features. We approached feature definition by asking operational questions:
•
Should users be able to zoom in and out of a massive canvas freely?
Should users be able to zoom in and out of a massive canvas freely?
•
How should interactive drill-downs work mechanically?
How should interactive drill-downs work mechanically?
•
Should there be an automated AI summary to synthesize the findings?
Should there be an automated AI summary to synthesize the findings?
•
How do users compare different metrics side-by-side?
How do users compare different metrics side-by-side?
•
How do teams navigate large, complex RCA trees without losing their minds?
How do teams navigate large, complex RCA trees without losing their minds?
•
How do they backtrack to a previous node without losing their current context?
How do they backtrack to a previous node without losing their current context?
At this stage, we were still strictly designing behavior, not visuals.
At this stage, we were still strictly designing behavior, not visuals.
5. Moving into Wireframes
5. Moving into Wireframes
With the behavioral architecture finalized, we started wireframing. The structural layout of wireframes allowed us to aggressively validate:
With the behavioral architecture finalized, we started wireframing. The structural layout of wireframes allowed us to aggressively validate:
•
Navigation paths
Navigation paths
•
Visual hierarchy
Visual hierarchy
•
Core interactions
Core interactions
•
Information architecture
Information architecture
This low-fidelity validation helped us catch and kill critical UX issues long before we spent time on visual design.
This low-fidelity validation helped us catch and kill critical UX issues long before we spent time on visual design.
6. The First True RCA Experience
6. The First True RCA Experience
Once the wireframes proved stable, we stepped into high-fidelity UI exploration. This was the exciting moment where Root Cause Analysis finally started feeling like a real, living product.
Once the wireframes proved stable, we stepped into high-fidelity UI exploration. This was the exciting moment where Root Cause Analysis finally started feeling like a real, living product.
Everything began coming together beautifully:
Everything began coming together beautifully:
•
The interactive canvas
The interactive canvas
•
The data nodes
The data nodes
•
The step-by-step investigation flow
The step-by-step investigation flow
•
The contextual side panels
The contextual side panels
•
The deep drill-down experience
The deep drill-down experience
7. From Sketches to a Living RCA Product
7. From Sketches to a Living RCA Product
Looking back, the polished final screens were only a tiny fraction of the actual journey. The real heavy lifting happened on dusty whiteboards, torn notebook pages, complex flow diagrams, and during endless, heated debates about how human beings solve business problems.
Looking back, the polished final screens were only a tiny fraction of the actual journey. The real heavy lifting happened on dusty whiteboards, torn notebook pages, complex flow diagrams, and during endless, heated debates about how human beings solve business problems.
Ultimately, we didn’t design RCA to be just another passive dashboard.
Ultimately, we didn’t design RCA to be just another passive dashboard.
We designed it to replicate a rigorous thinking process.
We designed it to replicate a rigorous thinking process.
Key Learnings
Key Learnings
Key Learnings
•
Users don’t need more dashboards. They need answers.
Users don’t need more dashboards. They need answers.
•
Revenue changes are rarely caused by a single metric. Every business outcome has multiple contributing factors.
Revenue changes are rarely caused by a single metric. Every business outcome has multiple contributing factors.
•
Root cause analysis is a journey of asking “Why?” until the actual issue is uncovered.
Root cause analysis is a journey of asking “Why?” until the actual issue is uncovered.
•
Most of the RCA work happened before Figma, through whiteboards, business logic mapping, and flow design.
Most of the RCA work happened before Figma, through whiteboards, business logic mapping, and flow design.
•
A great analytics experience guides users toward insights instead of making them manually connect the dots.
A great analytics experience guides users toward insights instead of making them manually connect the dots.
•
Every branch in the RCA flow should answer a question and lead users closer to the root cause.
Every branch in the RCA flow should answer a question and lead users closer to the root cause.
•
The best product ideas often come from real customer problems, not feature roadmaps.
The best product ideas often come from real customer problems, not feature roadmaps.
•
RCA taught us that analytics should explain causation, not just report numbers.
RCA taught us that analytics should explain causation, not just report numbers.
•
One simple client question, “Why are sales down?”, completely changed the direction of our product.
One simple client question, “Why are sales down?”, completely changed the direction of our product.
