Storage Isn't Storage: The Telemetry We Rebuilt Around Physics
Energy · VPP + AIM platform | Hybrid Greentech | Systems + data-viz design
Summary
A platform where money and machines share one screen
HGT operates a dual-purpose platform — a Virtual Power Plant that turns flexible green-energy assets into revenue, fused with an Asset Integrity Management layer that watches their health. One product serves four roles and six physically different asset types.
This case study follows a single decision: splitting one "unified" storage view into two — and how that one correction reset the way I design every asset screen on the platform.
Product
Web · Energy VPP + Asset Integrity Management
Timeline
2025 · ~4 months
My Role
Sole UX/UI designer, end-to-end. I mapped each asset's physics with the engineering team, defined the telemetry architecture, and owned the decision to split the component.
Skills
Information Architecture, Data Visualization, Systems Design , Usability Validation
Impact
Reset how every asset screen is modeled on the platform — two truthful telemetry architectures, zero ambiguous storage states, faster operator decisions under load. BESS shipped to production; Loadbank approved.
Roles & Research
The platform serves four roles, and I mapped all four (below). But a focused redesign needs a protagonist — so this case study zooms into the one where the hardest problem lived: the Operator.
Four roles mapped — one protagonist. The Operator, reading live asset telemetry hour to hour, is who this redesign solves for. Here's their day:
My research wasn't a survey — it was co-design. I sat with HGT's EPC Project Manager to map how each asset physically behaves, then stress-tested high-density layouts against those engineering constraints.
The Problem
"Storage is storage."
I designed one elegant, unified "Energy Storage View" and pushed both BESS and Loadbanks through it — the same bi-directional charge/discharge chart, the same State-of-Charge cycles, the same battery-rack hardware array. To a designer, storage is storage. Reuse felt like good systems thinking.
It was truthful for one asset — and a quiet lie for the other.
How I Worked
A five-step loop, run with the engineers
The same method runs through everything I take on — diagnose from reality, then design against it. On HGT it looked like this:
The Wrong Assumption
One sentence rearranged the architecture
In a technical review, Michael — HGT's EPC Project Manager — stopped the unified template cold. His one-line verdict rearranged the entire architecture.
A BESS flows both ways: it pulls power from the grid and pushes it back when market prices peak. A Loadbank flows one way: it absorbs excess energy and dissipates it as heat into the atmosphere. My "unified" widget was rendering phantom bi-directional flows and battery cells that physically don't exist on a Loadbank — an operator tracking a live grid interaction would have been reading a fiction.
What was at risk: On a live grid trade, a phantom bi-directional flow could trigger a wrong dispatch — a blown aFRR commitment, grid-compliance penalties, and lost revenue. The "lie" wasn't cosmetic; it was operational and financial.
Loadbanks are not the same as BESS.
How I Decided
Optimising for truth over reuse
Splitting the component looks obvious in hindsight — but I weighed three ways to handle two different assets inside one product before committing. The point wasn't to find a clever UI; it was to pick the option least likely to mislead an operator mid-trade.
Approach | Truthful | Learnable | Misread risk |
|---|---|---|---|
A · Mode toggle | No | Yes | High |
B · Tabbed views | No | Yes | High |
C · Two parallel views | Yes | Medium | Low |
I optimised for truth first, reuse second. In an operational tool, a wrong-but-elegant screen isn't a smaller win — it's a liability.
The Redesign
Wireframes & Structure
Before any hi-fi, I diverged the two architectures in low fidelity — proving the split worked structurally before a single visual decision. The wireframes encode the physics: bi-directional for BESS, one-directional for the Loadbank.
AI-assisted: I used Claude and Gemini to stress-test high-density layouts against Michael's engineering constraints, then hand-structured the winning direction before moving to hi-fi.
Two architectures, each true to its hardware
I tore the unified widget apart and built two specialized telemetry architectures — knowingly trading the elegance of one reusable component for two that each tell the physical truth.
Designed as a system
Two separate screens do not make a system. To prevent this from becoming a messy, one-off patch, I extracted a repeatable, parameterized telemetry framework. Every asset screen on the platform now inherits a shared layout foundation that flexes dynamically based on the asset's underlying physics.
Shared DNA (Every Asset): KPI tile rail (efficiency, health, cycles), date-range controls, a hardware-module region, and a time-series trend frame.
Flexes by Energy Flow: Dynamically shifts flow visualizations (bi-directional, one-directional, generation), hardware modules (battery rack, fan bank, inverter, turbine), compliance metrics (aFRR, thermal dissipation, curtailment), and degradation models (cycles, run-hours, thermal stress).
By establishing this framework, we unlocked immediate coverage for the platform’s other asset types: PV Solar, Wind, Hydro Dams, and industrial UPS networks.
A component that's elegant and wrong is worth less than two that are specific and true.
Built for the Worst Moment
Designed for 2am, not the demo
Operational dashboards are judged at 2am during an incident, not in a demo. So I designed for the failure modes first — the moments when an unreadable screen costs a real dispatch decision.
Severity never relies on colour alone — Normal / Warning / Critical each carry an icon + label, so a colour-blind operator reads severity instantly, not just a green or red dot.
Degraded data has a face — When a real-time feed drops, the tile shows a stale / disconnected state with a last-known timestamp — never a frozen number pretending to be live.
Alarms are allowed to be loud — Critical states deliberately break the calm layout, because silence during a fault is the real failure. Calm is the default, not the rule.
Glanceable under stress — The numbers an operator needs mid-incident sit top-left and largest, ahead of any exploration affordance. Decisions first, analysis second.
I treated empty, loading, error and alarm states as first-class screens — not afterthoughts. In an operational tool, the edge case is the job.
Impact & Learnings
Outcomes (qualitative)
Immediate Stakeholder Buy-In: The updated designs were approved on the spot with zero rework loops. When engineering saw their physical operations mirrored perfectly in the UI, friction vanished.
Shipped to Production: The BESS interface is live and actively used by operators for real-time grid balancing. Engineering is currently finalizing the thermal pipelines to launch the Loadbank interface.
A Culture Shift: This project redefined the role of design within the organization—shifting the team’s reputation from "pixel pushers" to strategic systems architects.
Learnings
Reuse isn't automatically good systems thinking. A shared component that contradicts the physics of the hardware is a liability, not an efficiency.
Research isn't a survey step — for technical products it's co-design. The engineer's mental model set the entire information architecture.
Design for energy flow, not labels. "Storage" is a word; bi-directional vs one-directional is a fact — and the fact wins.
Earning engineering trust compounds. Once they saw the UI respected the hardware, approval loops shortened for everything after.
How we'll know it worked: ↓ time to interpret a live trade · ↓ support tickets about confusing charts · ↑ operator confidence (pulse survey). Not measured yet — this platform is early. But naming the targets up front is how I'd keep the redesign honest once data exists.
Technical validation
There were no formal usability labs here — validation came from the people who run the hardware. The moment the technical team saw their physical operations mirrored exactly, the friction vanished and they approved the screens on the spot.
Interrogate the physics — Directional flow, degradation, TSO vs DSO grid constraints, mapped with the engineer first.
Ideate against constraints — Sketch and AI-assisted exploration of high-density layouts that fit those hard limits.
Sanity-check with engineering — Validate the architecture against the real data pipeline before a single hi-fi pixel.
When engineers see a designer who respects the laws of their hardware, the approval loops shrink.
What's Next
Loadbank goes live — engineering is finishing the dedicated thermal-telemetry pipeline that powers the cooling and dissipation views.
Apply the principle across the fleet — run the same physics-first loop for PV, Wind, Dams and UPS; each gets a model true to how its energy actually moves.
Deepen role-specific views — tailor the same telemetry to the Asset Owner, Operator, Service Partner and Admin mental models.
Designing correct models of reality — one asset at a time.