top of page

Marshmallow Miles

A market disrupting insurance offering that rewards customers for driving well

MacBook Air - 1.png


Marshmallow Miles is one of the first app based insurance products in the UK market, its core aim is to help offer accessible car insurance with great customer benefits. This product plan allows Marshmallow to reward customers that show safe driving behaviour with renewal discounts whilst removing dangerous risk from our books.

Leading a small product team from an initial design sprint to our beta market launch within 7 months. We successfully delivered an new type of insurance product that has helped us increase our market share of 21-30 year customers whilst positively impacting our Loss Ratio (In insurance this is the number one factor of success: Policies sold vs claims paid out).


- Lead product designer

- Product strategy

- User research

- Service design


12 months+

The challenge

Marshmallow were moving into their fifth year as an insurer but were finding it difficult to branch out successfully into younger customer segments where traditionally their higher risk of claims results in negative business impact. 

The Risk Mitigation product team's brief was essentially to find ways to help reduce risk for this segment of customers whilst building a product that meet customer needs in a competitive market. We needed to prove out to the business that our chosen approach could help meet the following goals:

  • Increase market share for 21-30 year olds

  • Increase click to paid quote (Main conversion metric)

  • Remove a higher percentage of high risk customers from our books

  • Reduce overall claims frequency and Loss Ratio

Kick-off & Discovery

We begin with a week design sprint to map out possible strategies that could help mitigate risk for the business. We decided to focus on the app based cover concept as this was a differentiator to other insurers in the industry. If done successfully this would open up to new segments and increase our market share with younger drivers.

The early customer proposition 

App based cover that rewards you for driving well. 

Project vision


Discovery approach

Stage 1

  • Market and competitor analysis into telematic product

  • Researching engagement approaches to build safer driving behaviours 

  • Stakeholder discovery workshop to gather requirements & explore problem space

  • Ran 12 initial user interviews with existing and potential customers

  • Concept testing the Marshmallow landing page

Stage 2

  • Feature journey testing

  • Diary study with competitor app with the new customer segment of 17-25 year olds

  • Design workshops 

Industry research: Building safer driving habits

Early findings

We concept tested a future Marshmallow Miles landing page to understand users expectations and potential challenges.

  • General comprehension of the product and how it works

  • Any concerns around privacy or sharing data with their insurer

  • If the rewards proposition was really a true influence on their decision to purchase

Programme overview

To setup our new insurance offering multiple areas of the business needed to work closely across the following touch points.

MacBook Air - 14.png

Building core functionality

As the product research continued we started to build out the core features of the product that would be required for MVP. These features needed to be validated with users whilst we looked into who we could partner with from a data perspective to help create a reliable driving data model. This model was crucial in accurately estimating the risk of drivers.

Main user journeys 

Web landing page

Customer   acquisition


Onboarding in the app

Driving score & factors

Engagement features & rewards

Step 1: Customer sign-up

This is the journey a new customer will go through to purchase miles insurance 

Key metrics within the growth team tracked across the sign-up funnel 


Click through rate

(From aggregators)


Net conversion rate 


Cancellation rate

Step 2: App onboarding

One of the most complex journeys to design and optimise was for setting users up in our app to share their necessary permissions to ensure we can track their drives accurately. 



Successful onboarding rate

(Under 24 hours)

Start a live chat to get assistance


Avg time to complete

Step 3: Driving and rewards

After their first drive the user will receive a score out of 100 based on three parameters speed, smoothness and focus (phone use). Depending on the drivers score at the point of renewal will qualify them for a renewal discount.

Improvements made post launch included:

  • Recency bias to help accurately reflect their current driving risk level

  • Ability for users to see drives we have detected as a passenger 

  • Visualising their drives and improvements on a map route 

  • Weekly round-ups to increase engagement and mindfulness of improving

Rewards & Loyalty approach

We believed by providing a 5% discount upfront for a telematics based plan would increase the traffic through comparison sites due to ranking in the cheapest positions near the top. We also advertised the opportunity to earn up to £100 off at renewal. It was important to look at how this could scale as a longer term loyalty programme for all our customers.

  • What was our budget spend on miles based customers per year and how to best utilise this through different incentives (Personalised treats, discounts, prize draws etc)

  • How can we ensure loyalty becomes more compelling over time to increase customer LTV

  • How can we ensure we are providing real value to customers

Evolving the design system

A host of new components and styles were created for this new feature, more than doubling the current app features we had in place. Documentation of components was key when briefing engineers and other designers to utilise these components for more consistency across the product suite.

Service design (Operations)

Working closely with operations teams to setup workflows for our main driving alerts to understand how we would prioritise these and setup flags/systems to monitor their progress. Through multiple workshops we prioritised and created eight workflows within the operations, fraud and underwriting teams to ensure all aspects of these new customers were being review and measured accordingly.

Our main achievements

  • 1000 policy to 1 full time agent was the initial KPI target, after our first beta month this read 10 agents per 1000 policies to review and needed improvement

  • Reducing sensitivity of alerts helped reduce their numbers and allowed agents to increase the number of dangerous driving events they could review

  • Improved product and email comms to customers to improve self service within the app and improve comprehension of the product plan

Product launch strategy

We wanted to de-risk our launch to mass market and learn as much as possible to improve the product and processes. Our primary metrics to monitor were:

  • Driving score weighting

  • Accuracy of drives recorded

  • Flagging thresholds for dangerous driving

  • Comprehension of their insurance plan (Cancelation rate)

  • Operational workflows and SLA's

Staged release plan

Testing and design reviews

Beta launch to 1000 new customer (21-30)

Review claims impact & customer mix

Internal alpha launch (40 users)

Release to full market (21-35) 


MacBook Air - 2.png

Driving score weighting 

  • Increased the weighting of smoothness as we found out this had the highest effect on their driving score along with phone handling (0.4,0.4,0.2)

  • Introduced a recency bias to better reflect their current level of risk


Accuracy of drives recorded & alerts thresholds

  • Provided features to allow customers to send a request if they felt their drive was scored incorrectly to help reduce negative reviews

Comprehension of their insurance plan (Cancelation rate)

  • Improved content within the sign-up flow and intentional friction to ensure customers were clear on the type of insurance they were purchasing to reduce cancelation rate 

Final product overview 

Sign-up & Purchase

Core app features

Driving insights widget

We built out a driving score widget as a way to increase awareness of their driving score to try and drive more awareness and engagement with safer driving.

  • 23% of customers downloaded the widget within 4 weeks

  • Of those customers 71% improved their driving score post download

Customer feedback (Month 3 survey)

We sent out feedback surveys at month 1 and 3 to our initial customers to understand first hand how they were finding their insurance cover. 

Ways of working improvements

Over the months post launch we adapted our ways of working in order to ensure we were meeting our objectives and continuing to build feedback loop with our early customers to improve the product. Multiple working groups were established to cover the following:

Customer feedback squad

  • Review miles based trustpilot review

  • Cancelation follow up and questionnaire 

  • Month 1 and 3 customer surveys

Product improvement

  • Review customer sign-up iterations and A/B tests

  • Setup an in app feedback feature to gather app/play store ratings and qualitative feedback

  • New engagement feature development 

Operations & Score performance

  • Dangerous driving cases and summary of driving alerts

  • Score performance and overall distribution

  • Cancelations & complaints reviews

Overall impact 

We managed to successfully find product market fit and establish a scoring model that accurately reflects the customers risk profile, however there were many learnings along the way.

Business impact (After 6 months)





Increase in % customer 21-30

Improved loss ratio

Increased click through

Increased in cost to serve

Product impact (After 6 months)



Accept push notifications

Weekly active users


Increased driving score post feature engagement 

Comprehension & Cancelation challenges

Underestimated the difficulty of explaining a greenfield type of insurance to a majority of migrant based customers that were not familiar with this. Therefore leading to higher initial cancelation rates which we later managed to reduce by switching customers over to standard plans.

Lack of operations oversight

Our flagging criteria for dangerous driving was fair to strict which create thousands of alerts that needed to be manually reviewed by the operations team. We slowly reduced the sensitivity of these and built automated responses and filtering into our agent portal.

Driving score errors & edge cases

There were drives that were incorrect and rightly customers complained and lost trust in the product. We begin to automatically remove passenger drives and give users the chance to reach out to update their profile.

Key learnings

bottom of page