Product Management · ENGIE Impact · 2024
ENERGY STAR Benchmarking
Replaced a 12-year-old manual process with an automated platform. Two analysts. Hundreds of corporate buildings. A monthly data load that used to take hours — now takes minutes.
Timeline
12–18 Months
Role
PM + Designer
Team
~4 Engineers
Type
0→1 Internal Platform
97%
reduction in monthly upload time
125
clients migrated at launch
10×
data capacity vs. legacy system
~0
client-reported errors post-launch
Brief
ENGIE Impact runs energy benchmarking for hundreds of corporate buildings — tracking consumption, generating ENERGY STAR scores, and helping clients meet sustainability targets. The analysts doing it had been running the same cobbled-together process for 12 years. My job: rebuild it, inside a platform that didn't support it yet. I was the PM and the only product/design person on the team.
Problem
The old stack: a 2012 transaction tool, SQL scripts, Excel templates, and an Access database. A single monthly data load took ~2 hours — and that was just runtime.
Analysts would open 30+ parallel instances of the tool because it maxed out at 50,000 rows. Duplicate submissions to the EPA were common. The whole thing was held together with duct tape, and demand was growing.
Research
I sat with the analysts and watched.
Trent and Andrew had been running this service for years. Before touching any requirements, I mapped their full workflow — from receiving client data to delivering an ENERGY STAR score — step by step. That meant reviewing their Excel templates, watching them use the old tool, and cataloguing every workaround. Three things kept coming up:
“Errors didn't surface until a client reported a discrepancy. Sometimes months later.”
“Duplicate submissions happened when analysts ran multiple instances to speed things up.”
“Onboarding a new client's buildings took 2–3 days of back-and-forth with outdated Excel templates.”
Solution
One platform. One action. No more duct tape.
We built a new ENERGY STAR module inside Ellipse — ENGIE's internal platform — to replace every piece of the legacy stack. The core of it: analysts pick sites, pick a date range, and hit one button. Everything else happens automatically.
Site Onboarding
Standardized templates instead of 50 Excel files.
Analysts can now create or import client sites directly in Ellipse using standardized property-type templates. The system automatically links each site to the client's ENERGY STAR Portfolio Manager account. We also built a 'grandfathering' flow for clients who already had ESPM accounts — no re-entering historical data from scratch.
Send to ESPM
Select sites. Choose a date range. Done.
The new 'Send to ESPM' workflow replaced dozens of manual steps. The module batches the data, calls EPA's API, creates building entries and meters as needed, and handles retry logic automatically. What used to be a 2-hour chore is now a few clicks.
Automated Monthly Load
The 15th runs itself.
Beyond on-demand uploads, the system runs a scheduled monthly job — automatically pushing the latest energy data for all active client sites on the 15th. Analysts monitor job statuses. They only intervene when something goes wrong.
Error Dashboard
You know exactly what to fix.
Every job shows a real-time status per site: In Progress, Successful, or With Errors. Failed jobs include a downloadable error report that pinpoints the issue and tells you what to do about it. Before this, figuring out what went wrong meant digging through logs and the ESPM website manually.
Key Moments
Three decisions that shaped it.
Ship in phases, not all at once.
I pushed to split the project into two releases instead of one big launch. Phase 1 — just the site onboarding piece — shipped mid-2024. It wasn't glamorous, but it gave us real integration feedback from ESPM before we built the automated upload engine. That feedback shaped Phase 2 in ways we couldn't have anticipated from a whiteboard.
Automate the output of good data, not bad data.
About 20% of sites had some kind of data mismatch in the old process. When I pushed to build pre-upload validation, some people questioned the effort — it wasn't adding new customer-visible features. I used that 20% number to make the case. Without it, we'd just be running the old errors faster. Leadership agreed. After launch, client-reported data issues dropped to essentially zero.
Sometimes the right call is messier.
We planned to fully retire the legacy system. Then we discovered an old Access config file with historical mapping data that hadn't fully migrated — edge cases like campus-level aggregations and 10+ years of solar production data. We kept two legacy components in read-only 'cold storage' instead of cutting them. It wasn't a clean break. It was the right call.
Outcome
The VP called it a template for the rest of the org.
At launch, we migrated all 125 existing ENERGY STAR clients into Ellipse — including roughly 60 who had been managed entirely off-platform. Monthly data loads that used to run for hours now run in minutes. The system can now handle 600,000 building sites; we had about 60,000 in the old one.
The VP of Product and the head of the Energy Star service highlighted it at an all-hands. Our work was included in the company's H2 digital product launches communication as an example of successful modernization. Trent, one of the analysts, told me having everything in one place “makes our lives so much easier — I can't imagine going back.”
Reflection
A better product doesn't sell itself to its own users.
Phase 1 launched just before the annual reporting season — the analysts' busiest time of year. They didn't use it. Not because it was bad. Because they were underwater, and switching tools mid-sprint felt riskier than grinding through the old process one more time.
For Phase 2, I coordinated training away from peak periods, wrote the user guides myself, and got department leadership involved in the rollout. That's what actually drove adoption. I used to assume the product did the selling. This project taught me it doesn't.