To provide the right insurance for homes and workplaces, insurers have to be sure of the exact location of the assets.
But 500 years on from the creation of the UK’s Royal Mail, the challenge of knowing where something is and what surrounds it remains.
Addresscloud is one company rising to that challenge. Five years on from launch, it has processed 32-million addresses and 25-million buildings, with the data being used to help insurers quote and bind, fight wildfires in the US and even for organic vegetable deliveries.
Founder and CEO Mark Varley joins Matthew to discuss where the idea came from and how the company has secured a strong list of clients in the UK, US and Europe.
Talking points include:
- Identifying and processing useful data sources
- Using analytics for higher resolution modelling
- Why big tech isn’t always best on geocoding
- Bootstrapping vs external funding
- The benefits of working with clients outside insurance
If you like what you're hearing, please leave us a review on whichever platform you use, or contact Matthew Grant on LinkedIn.
Sign up to our newsletter for a fresh view on the world every Wednesday morning.
Hitting the spot - getting location accuracy right - Episode 114 highlights
Matthew: Mark, you’ve got an intriguing founder’s story. You started at Accenture and were working for RSA before setting up Addresscloud. What was the motivation for starting your own business?
Mark: RSA were quite pioneering as they took Geographic Information Systems (GIS), which was often a sort of specialist back-office function, and put it as part of the quote and bind journey in the direct channels.
We were buying high-quality address data and it worked well, but when we dealt with broker submissions where addresses had complex postcodes or typos, the challenge of getting the location right was hard. We were seeing addresses that were going to the wrong location and some that we couldn't match at all. There had to be a better way, which is why I created Addresscloud.
Matthew: What’s your advice to other early-stage entrepreneurs who have spotted a problem?
Mark: I aimed to build the solution first, rather than aiming to build a business. I spent my evening and weekends working to solve the problem, then pitched it back to the organisation.
Don't be scared of the challenge, go out there and build it. You never know, it might work out and become successful.
Matthew: It might surprise people to know there’s still a challenge in getting accurate address data. Why is that and what problem does Addresscloud solve?
Mark: We’re blessed with a very good addressing system in the UK but getting beyond postcode level to address levels is challenging. The data is closed source, so a list of all UK addresses would have to be bought from Royal Mail and information on coordinates would have to be bought from the Ordnance Survey (OS). We are an official OS partner and work closely with their mapping experts and that data gives us a really good location with an address associated with it.
The big search engine providers don’t pay for those data sets, so using one of those solutions won't necessarily get to rooftop level in the right location. Addresscloud brings the OS and Royal Mail data together in one system that matches an address to a set of coordinates.
Matthew: Analytics are getting more high resolution. For example, people are looking at modelling flood at a higher resolution. Can Addresscloud offer a full building footprint definition through the OS data?
Mark: We’ve seen that shift from postcode level assessments to address level assessments. The optimal level for property risks assessment is the building itself as addresses can often have too narrow a focus. For something like a large, high-net-worth property, we take into account the footprint of the building or the site. For example, if a commercial site has a warehouse at the back, taking that information into account is a better way to work for a high-resolution peril like flood.
Matthew: You’re bootstrapped as a business, so you’re relying on revenue and must have a strong customer base. Who are your customers?
Mark: RSA is still our biggest name customer. They use us in their quote and bind journey for direct business, both through their branded partners and their MORE THAN brands, and for commercial risk. Any addressing transaction that takes place comes through us and we do around 10 million a month.
We’re working with Brit on its UK property and US business. The UK work is focused on perils like flood subsidence, while the US focus is on wildfire and flood. We also work with some smaller insurers and MGAs like C-Quence and Unicorn, and we recently signed a three-year contract with Protector in Norway for geocoding.
Matthew: The original focus was on UK data. Are you now expanding into other geographies?
Mark: Absolutely. Our address matching solution is very data-intensive, so we focus that on the UK and Ireland. We're partnered with HERE in the US and use their geocoding service, and we have our own global perils platform.
We’re also working with another US partner called Anchor Point, which is a team of firefighters and data scientists who’ve created a fantastic wildfire model. We provide that data through APIs and our maps visualisation platform. We’ve got around 250 individual users of that application in the US and another 60 cover holders, mainly through the Brit relationship.
Matthew: Firefighters and data scientists is a really interesting combination.
Mark: They’ve taken a similar approach to what JBA did in the UK with floods. Going back 10-15 years, flood assessments were quite broad and the public data from the Environment Agency was around a 50-metre resolution. JBA took that down to a five-metre resolution and really pushed forward individual property flood assessments.
Anchor Point is seeking to do the same in the US with wildfires. Rather than saying a whole city block is at risk, they focus on the periphery and go down to that higher level of resolution.
Matthew: You mentioned it is labour intensive to create data for the UK. What does your team do with the data from different sources? Do they need to review data manually or is it fully automated now?
Mark: It’s semi-automated at the moment. The data that comes from JBA, the British Geological Survey, or Cranfield University is huge when it arrives. It’s the same for US wildfires. The data contains highly complex, geographic formats and working with it requires specialist skills and tooling.
In the UK, we process 32-million addresses and 25-million buildings. They’re pre-processed with some quite complex rules that require a lot of compute. The result is a score or a set of scores that we push out via our API.
Matthew: There are always lessons to be learned from different areas. Can you share some examples of clients Addresscloud has outside insurance?
Mark: Our two pillars are our Match API, which is our address matching, and our Intelligence API that describes properties. We’ve gone from working purely with insurers in the intelligence space to working with lenders and surveying companies as well. The survey part is born out of surveyors not being able to visit sites, so they are looking for external data sets and our property and risk data has been really helpful.
On the lending side, there’s a lot of pressure on banks to understand climate change and take a longer-term view. We’re used to working with 12-month insurance policies, so taking a 20-30 year view of the risk around a mortgage is a big challenge for us. Providers we work with like JBA are also creating Environment and Climate Change specific products, which have been interesting and different to contribute to.
We've also worked with Dixons Carphone on logistics in Ireland, where a third of the country doesn't have a unique address. More recently, we started working with Riverford Organic Farmers to solve last-mile logistics problems. What's important to delivery drivers isn’t the rooftop, but where they can pull up outside a property. We’ve created curbside locations and are assigning those to every address.
Matthew: How easy is it for people to use the APIs you mentioned, and what do they integrate within their systems?
Mark: The Match API takes in either an address or just a postcode and returns an exact match or list of addresses for customers to pick from. We return a clean, structured postal address, geographic coordinates and assign a unique identifier which we call our address key.
We also provide the unique property reference number (UPRN) from the OS. Some clients like RSA just use that as it gives them everything they need. They use it in their internal systems to look up by coordinates or by the UPRN to find out more information.
The second Intelligence service takes in coordinates or our address identifier and returns everything we know about a property. That could be the geographic risk profile for perils like flood, or property attributes like the number of bedrooms, the height of the building, etc. That might be our data, curated open data, or third-party data that we've linked to.
Matthew: Several organisations have been trying to come up with an ID number for every UK building, but it sounds like the OS is already doing it?
Mark: They launched their address base products in 2011 so, for example, each flat in a block would have a UPRN. They've done that in partnership with local authorities and the UPRN is often used on planning applications. From the point at which construction starts through to demolition, each property should have a consistent UPRN.
Matthew: Does it track changes to the building? That's a big challenge for insurance and producing a correct valuation. Is there a central source?
Mark: The datasets we get will take in the land registry price, EPC data, and a few of them use the UPRN, but it’s not that one identifier to rule them all. It’s getting better all the time and it’s great in the UK. It won’t solve the problems elsewhere in the world, but we’ve got to start somewhere.
Matthew: That seems like a big opportunity. Can you explain EPC data for anyone not familiar with the acronym?
Mark: It’s the energy performance certificate, which is an open data set that is available to download. Anyone selling a rental property is required to get an EPC certificate and it’s really good data. There's information around construction type and age which we use in our property data. It gives us a good open-source for that information.
Matthew: Addresscloud has a lot of clients for a small organisation. What is your market strategy for signing them up?
Mark: We tend to get new business through existing customers. People either hear about us through recommendations, or they go to one of our data partners to procure a service and that partner recommends us. There are five of us in the current team, but we’re looking to bring on more people next year. The steady pace seems to have worked for us.
Matthew: Has it been a conscious choice to work with people in different industries?
Mark: We probably followed the money in the early days, but it’s led us to some good places. We’ve recently put in an internal triage system and we do say no. Essentially, we’d do anything focused on finding or describing addresses, but we wouldn't do anything to do with individuals or something like a credit risk.
Matthew: Does Addresscloud have anything coming up that you’d like to mention?
Mark: The property data we've touched on a couple of times is something we've worked hard on for the last six months. We've now got 32 attributes across 32 million addresses.
There are others out there providing that data through AI, but we see it as a geographic problem. We work with property listings data, warranties, surveys, sales data, and we link it to addresses and buildings. We model out a very carefully curated data set and assign either an A record, an address level record, through to a B record, where we've modelled where the addresses or data correlates.
That might be a block of 20 flats where three of them have had a survey undertaken and they're all saying the same results. We can then predict with a degree of confidence that the rest of the 16 will be the same. That data is now live in our API and the data model is available to view at docs.addresscloud.com.
Matthew: Thank you for your support for InsTech London. Why has Addresscloud signed up as a corporate member?
Mark: I follow the events and I'm a big fan of the podcast. It's now my go-to and I'm catching up on the back episodes as well. We think InsTech London is a great forum with lots of like-minded people, and we've been fortunate to meet some of our data partners through it.
We’re keen to work with more of the ecosystems and insurance platforms who are looking for third-party feeds. We're happy to discuss that with anyone interested. It's a great community and we're excited to be a part of it.
Continuing Professional Development - Learning Objectives
InsTech London is accredited by The Chartered Insurance Institute (CII). By listening to an InsTech London podcast, or reading the accompanying transcript, you can claim up to 0.5 CPD hours towards the CII member CPD scheme.
- Claim 0.5 hours for listening to Episode 114 of the InsTech London Podcast