Company

Bulb

Position

Lead designer

Driving a 14% increase in app usage with smart insights and behavioural design

A 198k user experiment driving engagement through smart energy insights

The image featured at the top of the about us page #1

The image featured at the top of the about us page #1

The image featured at the top of the about us page #1

Context

Context

In 2022, following the onset of the UK energy crisis, Bulb entered special administration. With the government seeking a buyer and wholesale prices exceeding the trade cap, members were facing 54% increases in their bills.

Our team focused on improving engagement to help members manage costs, raise the product’s perceived value, and support potential operational savings by encouraging smart meter data sharing.

We rolled this out to over 140,000 members through a staged experiment, with a 14% uplift in engagement among users who received the feature.

My role

My role

As lead designer on Smart Insights, I owned the experience from early discovery through delivery — balancing user research, behavioural strategy, system integration, and cross-functional alignment.

Facilitated cross-functional workshops with data science, engineering, and product to align on behavioural strategy.

Shaped experimentation structure with PM and data lead.

Drove iterative design and evaluation, using both user feedback and behavioural data to refine how we surfaced insights and set goals

Discovery insights

Discovery insights

Users wanted to feel in control

Past research showed peace of mind came from knowing what’s going on. This was especially true during the energy crisis.

Usage breakdowns were wanted

During concept testing, showing usage breakdown by categories like heating and cooking was one of the most popular ideas with a test group.

Comparisons drove action

Studies showed that comparing energy use to similar homes reduced consumption by 3-12%, motivating members to adopt energy-saving behaviours

Usage breakdown and comparisons were consistently top-performing ideas in concept tests — so we prioritised these for MVP.

Direction

A behaviour-led design strategy

A behaviour-led design strategy

To design data that can help members reduce their bills, I proposed to use Fogg’s Behaviour Model—focusing main 3 items:

Build members ability to understand and manage their usage by giving members tools—like usage breakdowns, personalised tips, and clear benchmarks.

Boost motivation with savings insights, comparisons to similar homes, and progress through achievable goals.

Prompt action with timely nudges, usage insights, and contextual comparisons that encourage immediate decisions.

To design data that can help members reduce their bills, I proposed to use Fogg’s Behaviour Model—focusing main 3 items:

Build members ability to understand and manage their usage by giving members tools—like usage breakdowns, personalised tips, and clear benchmarks.

Boost motivation with savings insights, comparisons to similar homes, and progress through achievable goals.

Prompt action with timely nudges, usage insights, and contextual comparisons that encourage immediate decisions.

To design data that can help members reduce their bills, I proposed to use Fogg’s Behaviour Model—focusing main 3 items:

Build members ability to understand and manage their usage by giving members tools—like usage breakdowns, personalised tips, and clear benchmarks.

Boost motivation with savings insights, comparisons to similar homes, and progress through achievable goals.

Prompt action with timely nudges, usage insights, and contextual comparisons that encourage immediate decisions.

To design data that can help members reduce their bills, I proposed to use Fogg’s Behaviour Model—focusing main 3 items:

Build members ability to understand and manage their usage by giving members tools—like usage breakdowns, personalised tips, and clear benchmarks.

Boost motivation with savings insights, comparisons to similar homes, and progress through achievable goals.

Prompt action with timely nudges, usage insights, and contextual comparisons that encourage immediate decisions.

These elements worked together in a continuous loop of feedback and progress, designed to create engagemnet and build new energy habits.

First release

Helping members make sense of rising bills

Helping members make sense of rising bills

What we introduced

A new home screen module showing where members spent the most energy
Usage categories to provide a clear breakdown of members' energy bills
Comparisons to similar households

Who used it

First released to 18 Bulb employees and 12 members for early testing

What we learned

✅

Members were drawn to everyday categories — they felt personally relevant`

✅

Transparency around the data improved trust

⚠

Many were curious about how the figures were calculated`

⚠

Some questioned the accuracy without clear explanations

What we learned

✅

Members were drawn to everyday categories — they felt personally relevant`

✅

Transparency around the data improved trust

⚠

Many were curious about how the figures were calculated`

⚠

Some questioned the accuracy without clear explanations

Who used it

First released to 18 Bulb employees and 12 members for early testing

What we improved next

Added simple definitions of ‘similar homes’
Introduced tooltips and explanations to build confidence in the data

Released in stages across ~113,000 members

Building trust and accuracy through member profiles

Building trust and accuracy through member profiles

What we introduced

A short survey to collect home profile details and improve data accuracy
A progress bar and gentle reminders to encourage completion
Fixes for earlier bugs and issues raised in testing

What we learned

✅

41% completed the survey — members were motivated to improve accuracy

✅

Transparency around the data improved trust

⚠

Some found the onboarding slightly cumbersome

What we improved next

→ Made the survey easier to access
→ Highlighted the benefits of completion throughout the experience

Rolled out to ~46,000 members

Goal-setting — giving members a sense of control

Goal-setting — giving members a sense of control

What we introduced

• A feature to set a monthly usage goal, based on forecasts
• Suggestions to reduce usage by 5–10%
• On-track/off-track nudges, plus links to energy-saving advice

A feature to set a monthly usage goal, based on forecasts.
Suggestions to reduce usage by 5–10%
On-track/off-track nudges, plus links to energy-saving advice

What we learned

✅

The goals page was visited in over half of all sessions

✅

Around 25% became regular users who checked their progress

⚠

Some wanted to compare their goals with similar homes

Feature roll-out: staged release with experimental control

Feature roll-out: staged release with experimental control

To measure real-world impact, we rolled out Smart Insights gradually using a randomised control trial. Members were grouped by feature type and release phase:

To measure real-world impact, we rolled out Smart Insights gradually using a randomised control trial. Members were grouped by feature type and release phase:

To measure real-world impact, we rolled out Smart Insights gradually using a randomised control trial.

Members were grouped by feature type and release phase:

To measure real-world impact, we rolled out Smart Insights gradually using a randomised control trial. Members were grouped by feature type and release phase:

Usage breakdown + Similar homes data

Released in stages across ~113,000 members

Goals

Rolled out to ~46,000 members

Combined

Largest group to test the full experience

Control group

33,000 member who didn't receive Smart Insights

Results

Results

App views increased by 14% among users who received the new features, with the highest engagement seen in users who had access to both features.

41% completed profile survey - indicating high level of interest, even though long survey

25% became regular goal users

Future vision

Future vision

We saw significant potential in providing members with better understanding and control over their energy use. Based on the results of our experiment I suggested the next steps for future iteration:

We saw significant potential in providing members with better understanding and control over their energy use. Based on the results of our experiment I suggested the next steps for future iteration:

To measure real-world impact, we rolled out Smart Insights gradually using a randomised control trial.

Members were grouped by feature type and release phase:

We saw significant potential in providing members with better understanding and control over their energy use. Based on the results of our experiment I suggested the next steps for future iteration:

Refine Eliq’s data quality

Continue to identify and fix inconsistencies in Eliq’s data to build trust and resolve member concerns.

Complete the behaviour model

Add personalised recommendations to not only foster behaviour change but also provide important, tailored content that resonates with users.

Boost goals adoption

Increase the uptake of goals by combining them with similar household comparisons (e.g., "Most similar homes use less energy—set this as an achievable goal").

Increase frequency of usage engagement

Utilise 10-second data to present usage insights on a daily basis instead of monthly, encouraging more frequent interaction and proactive energy management.