An AI-powered energy management app designed to give UK households real-time visibility into their electricity consumption — and the tools to act on it.
The April 2025 energy price cap rose by 6.4%, pushing average household costs to £1,849 annually — the third consecutive rise since September 2024. Electricity demand is projected to grow by 50% by 2035 as the country electrifies transport and heating.
Yet 36.4% of UK households now spend more than 10% of their income on energy. Apps from British Gas, OVO, and Scottish Power are locked to provider accounts, show only next-day data, and provide no personalised guidance. I experienced this myself. That frustration became EcoBuddy.
Average annual household energy cost — up 6.4% April 2025
Projected electricity demand increase by 2035
Households spending over 10% of income on energy
"Over half of UK homes have an EPC rating of D or below — the majority of users live in inefficient homes with limited upgrade options."
— UK Parliament Research Briefing, 2024
Six features. Each validated by a testable hypothesis.
EcoBuddy is a real-time energy management app that integrates with smart meters and IoT devices to deliver personalised AI recommendations, environmental impact metrics, and educational content — independent of any energy provider, in one place.
I led every stage — from scoping the research to shipping a tested, high-fidelity prototype. Lean UX was the right methodology for an MVP product: it keeps scope honest by validating assumptions before building, not after. Every design decision traces back to a tested assumption. Nothing was built without a reason.
Of household energy use is space heating — yet most apps treat all consumption identically
Of UK consumers say smart home technology costs too much — EcoBuddy had to work with what users already owned
I analysed five existing energy apps — British Gas, OVO Energy, Ivie, Energy Saver App, and Scottish Power. The gaps were consistent across all five.
| Gap | EcoBuddy's Response |
|---|---|
| Delayed data — next-day only | Real-time smart meter integration |
| Provider lock-in | Works independently of any supplier |
| No personalisation | AI recommendations based on individual usage patterns |
| No environmental metrics | CO₂ impact made tangible for users |
| Overwhelming notifications | Customisable, targeted alerts only |
| No government scheme guidance | Dedicated ECO4 and scheme content section |
Prefer mobile as their primary platform
Manage consumption via smart meter
Cited real-time insights as most desired feature
Spend £100–£200 monthly on energy bills
Semi-structured interviews with 4 participants gave depth — a homeowner in a three-storey townhouse, a London flat renter, an apartment owner, and a new property buyer. Seven themes emerged from the raw data: demographics, pain points, behavioural factors, awareness, preferences, suggestions, and dislikes.
Difficult to get real-time insights — existing apps only show next-day data
Lack of personalisation — generic tips don't account for house type or usage behaviour
Lack of knowledge — users want education alongside data, not just numbers
Fragmented solutions — managing energy across multiple apps creates friction
Interview data was synthesised into an affinity map across four participants, producing business and user assumptions before any design decisions were made.
Affinity map across 4 participants — 7 themes surfaced, including Pain Points, Behavioural Factors, and Preferences. Four key findings shaped the entire feature set.
Every feature decision was filtered through two distinct users.
IT professional · 3-floor London townhouse · Frustrated by delayed data and basic provider insights. Wants real-time visibility to act on his energy use.
Directly shaped H1 — the real-time dashboard built independent of any supplier.
Sales manager · New 2-bed apartment · Lacks consistent energy habits and knowledge. Wants simplicity and fewer notifications.
Directly shaped H2 and H3 — personalised recommendations and the EcoEducator.
Both personas were stress-tested against every feature decision — if a feature didn't solve a pain point for either user, it didn't make the MVP.
The full user experience was mapped across six stages to identify emotional low points and the design decisions they drove.
User journey map — 6 stages from Awareness to Advocacy. Emotional low points at Onboarding (trust gap) and Exploration (information overload) directly shaped the simplified dashboard and step-by-step device connection flow.
Rather than jumping to wireframes, I converted prioritised assumptions into 7 testable hypotheses using the Lean UX format — "We believe [business outcome] will be achieved if [user] attains [benefit] with [feature]." Every screen had a direct line back to a testable assumption. Design was not decorative — it was evidence-building.
Scope discipline is as important as feature design. These items were deliberately cut from the MVP to keep the core testable and coherent.
The IA was built directly from the seven hypotheses — every section of the app maps to a validated user need identified in research.
IA mapped directly to hypotheses. Home dashboard (H1) · Analytics (H1) · EcoEducator (H3 & H5) · Budget (H4) · EcoBuddy Chat (H7) — every nav item traces back to a validated assumption.
Before opening Figma, every screen was sketched by hand. Low-fidelity wireframes mapped six core screens — Home, Device Analytics, Smart Meter Analytics, Scan and Connect, Eco Educator, and EcoBot — directly from the information architecture and the seven hypotheses. The sketches confirmed which features belonged at each navigation level and which hypothesis each screen was serving before any visual design decisions were made.
Six screens sketched before Figma was opened. Home · Device Analytics · Smart Meter Analytics · Scan & Connect · Eco Educator · EcoBot — each mapped directly to a prioritised hypothesis.
A style guide was established before any hi-fi work began. Colour contrast ratios, text sizes, and button tap targets were verified against WCAG AA standards. Typography — DM Sans throughout — was chosen for legibility across all ages including the 55+ persona identified in research. Following H4 usability feedback, button sizing and component scale were revised before the final prototype was completed.
A single-hue green system — every shade chosen to reinforce the environmental and sustainability positioning without visual noise.
DM Sans across all levels — simple yet modern, chosen for legibility across all ages from 28 to 55+.
Rounded pills and soft cards — deliberately approachable for users who feel anxious or overwhelmed by energy bills.
Design system established before hi-fi began. Colour contrast, touch targets, and type scale verified against WCAG AA — updated after H4 usability feedback before the final prototype.
These decisions are directly traceable to research findings and hypothesis requirements — not aesthetic preferences.
Research showed users were frustrated by delayed, next-day data. The dashboard prioritises live consumption from the smart meter so users can act immediately — not retrospectively.
A key interview finding: users actively disliked generic tips. EcoBuddy's recommendations are tailored to the user's actual consumption patterns and home details.
Research found that connecting savings to tangible impacts — "saving plastic bottles" or "planting trees" — was motivating. The EcoEducator translates actions into visible outcomes.
Users flagged frustration with excessive notifications from existing apps. Weekly summaries were preferred by 57.1% of survey respondents.
Users had no visibility of ECO4 and government grants. EcoEducator's dedicated scheme section surfaces eligibility information directly — no external redirects.
Navigation was tested explicitly — users found their target screen without assistance. A flat five-tab structure replaced any layered menu approach.
Users wanted answers in context, not documentation. EcoBot uses LLM and RAG to answer energy questions conversationally — tested and validated at 100%.
7 hypotheses tested · 5 fully validated · H1 and H4 partially passed — both pointed clearly to the next iteration
Four of five participants passed. One participant failed the task. The result confirmed that real-time dashboards are a meaningful differentiator — and that the remaining gap is worth addressing in the next iteration.
Two of five participants failed the budget tracking task. Post-test feedback pointed clearly to the visual design — specifically the sizing of certain interface components like buttons — as the barrier. The feature concept is right. The execution needs refinement. The next iteration focuses on component sizing and visual hierarchy within the budget tool.
Users expressed strong appreciation for the app's functionality overall. Device-based analysis and personalised recommendations were a standout — participants called it out unprompted. Constructive feedback focused on refinements to the size of certain interface components to improve clarity and ease of use.
"H1 and H4 partially failed — and that was equally valuable. They showed exactly where the next iteration needs to go. That cycle — discover, frame, design, test, learn — is the work."
Every screen traces back to a user insight. Every decision traces back to a hypothesis. That is the work.