Shipped
100+ Facilities
Designing the Interface Behind Renewable Energy’s Hidden Data Problem
How I turned renewable energy’s most tedious task into intelligent collaboration.
DURATION
Sep 2022- Dec 2022
ROLE
Product Design Intern
TOOLS
Figma, AI
TEAM
Head of Product, Engineers, Interns, CEO
Ren’s mission to scale sustainability was stalling at the starting line.
Ren Energy helps retailers reduce their environmental footprint through precise data. But before impact could begin, every new client had to upload a “perfect” data file—one small error could crash the system.
A single column could block the mission
One broken upload stopped the mission cold.
Clients didn’t understand how to prepare their files, and engineers had to fix them by hand. The process demanded perfection from people who were just trying to participate.
I learned this by talking to both sides of the problem.
I interviewed five client users and three Ren engineers to uncover where uploads failed. Clients described feeling “stuck and embarrassed,” while engineers described it as “the same fire drill every week.” These conversations formed the basis of two personas—one seeking clarity, the other seeking time.
My challenge: protect precision, remove pain.
Ren couldn’t scale because engineers were trapped hand-holding clients through uploads. I needed to design a process that preserved accuracy but guided real people toward success on their own.
“My goal wasn’t to lower the bar for accuracy—it was to raise understanding.”
I mapped where understanding broke down.
The upload flow was simple on the surface: clients submitted a Higg data file, and engineers handled the rest. But inside that black box, the system often failed silently. When imports broke, clients couldn’t see why, and engineers had to reopen and re-map the files by hand.
The system demanded precision—but never taught it.
Each file had to match an internal schema exactly: column names, spellings, and formats. Clients had no way to check their data or understand the rules. When one field name was off, the entire import failed—and every fix landed on the engineering team’s desk.
Store_ID
≠
Store Number
Not a Match
My Early Insight: Give Control Back to Clients
Instead of trying to automate everything, I wanted to give clients visibility—show them what the system saw, and let them help fix it. The goal wasn’t speed; it was shared understanding.
My Design Goals
To make complexity visible without overwhelming users, I defined three goals:
Analyzing Higg File…
Give Clients Transparency
Let clients see how their Higg file was interpreted.
19 of 20 columns matched
Confidence: 92%
Build Trust Through Control
Empower clients to participate in the process by matching fields directly, believing that visibility would create confidence.
Energy Use (kWh)
Needs Review
Water (m³)
Confirmed
Reduce Engineering Bottlenecks
Shift responsibility earlier in the process so clients could fix simple mapping issues themselves before engineers had to step in.
When I Gave Users Control, I Created Confusion
I built a manual mapping tool so clients could match Higg columns themselves, thinking transparency would build trust. But engineers showed me it only recreated the same chaos: clients didn’t know what the fields meant.
Then I Let Automation Take Over, Breaking the System
My second solution auto-filled columns based on name matching, but when even one label differed (“Power_kWh” vs “Total_Energy_Use_kWh”), the system failed silently. Instead of trust, automation created blank mappings, broken imports, and slower reviews.
The Breakthrough was finally Collaboration
I introduced AI matching that analyzed both column names and content patterns, offering suggestions with confidence scores. Instead of forcing users to fix errors, it invited them to verify likely matches—turning a one-sided process into collaboration.
Trust Came From Validation, Not Perfection
The mapping interface let users match their column names to Ren’s schema, see what didn’t align, and fix it directly. It transformed a technical alignment process into a transparent, guided experience—so users could do what engineers used to.
Turning Precision Into Partnership
By quantifying confidence and surfacing uncertainty, we improved both speed and accuracy—turning AI precision into shared accountability.
90%
Matched Automatically
AI matched the majority of fields on upload. Engineers only reviewed low-confidence mappings (<80%).
60%
faster imports
Average file mapping time dropped from 18 minutes to 7.
35%
Fewer errors
Schema inconsistencies decreased across client submissions.
100%
faster imports
All engineers switched to the AI-assisted workflow.
REACH OUT!
If you're building for education, care, or real-life complexity, reach out: naoboru@sas.upenn.edu — I’d love to collaborate.