Podcast: Data Driven Intelligence

September 1, 2020 14:49

About Us, Tech

Forbes features on this week's prestigious Instech London Podcast, discussing our history, plans and how we became one of their 'companies to watch'.

"MIS has become one of the companies to watch as it combines aerial images, sensors on the ground and human intelligence to build detailed pictures of risks"

You moved into insurance after a career in the military. How did McKenzie Intelligence Services get started?

I had a wonderful career in the British Army Intelligence Corps. I signed up on 9 September 2001 and then two days later, everything changed. I did a lot of work with the US intelligence architecture, which is beyond anything in the UK or Europe, and worked with some very interesting space and drone technologies that provided real-time intelligence. After 10 years, it was time to settle down, so I set up a niche consulting business covering the UK and hotspots around the world. During the Arab Spring, some colleagues from the military were working in the Lloyd’s marketplace and required information from Libya, Benghazi in particular. Those questions could be answered by satellite technology. We still went to Benghazi but could do a lot more remotely. We worked with Aon and other managing agents, which led on to working with start-ups in North America, including Skybox Imaging which became Planet via Google. We worked with their data, pixels, automation and the Google stack, to understand what the scale and the potential would be.

Essentially, you're identifying the best sources of data and converting that into actual insights and intelligence?

We're very agile and use lots of different data sources that are either flown in space, or non-air breathing as we call them technically, or air-breathing which is a drone or a plane. In the last 12 months, we’ve added the ability to draw data from fixed points on the ground using the Internet of things (IoT). Clients give us their risk and we fix points on the ground to answer their questions. We combine everything to make a detailed picture of what's taking place.

When MIS started working with Lloyd’s there were very few organisations doing this type of work. How did you persuade Lloyd’s to spend money when it seemed like no one else could provide these services?

My military background is in human intelligence and part of doing that is to have an access point to an organisation of interest. I took that approach to it. As a small business who spent all its money on developing a product, how could we gain access to an organisation that offered scale? Being London-based, we had the Lloyd’s marketplace on our doorstep, and it made sense to work with them from a commercial penetration point. A lot of time was spent looking for the right people and how to get alongside them. That led into the inner workings of a number of managing agents, who then passed our details into Lloyd’s and the Lloyd’s Market Association.

How did the relationship with Lloyd’s develop?

We started picking up small pieces of work related to the aftermath of the Arab Spring. Our satellite imagery analysis of Tripoli airport, Saana airport in Yemen, and some pharmaceutical plants in Mosul raised our profile and we were invited to a conversation with Lloyd’s. The Strategic Market Claims Group in Lloyd’s was set up to prove we could get data after an event, and the event we worked on was Hurricane Matthew. We had nine satellite images with 24 hours of the event taking place, and our imagery analytics was applied on top to extract exactly what had taken place at hotels in the Caribbean. We built a portal to deliver the intelligence to clients in a way they could understand it. That one action achieved market penetration at scale with 53 managing agents in the London marketplace.

What are the mains reasons why clients want to use your data?

Initially, the claims community liked the potential of seeing a property just after an event had taken place. Typically, a hurricane or flood event is not going to destroy a building. There is an adjusting cost and we can get an assessment of what that cost might be. We didn't know at the time what a novel way of applying the data this would be for exposure teams. Our exposure estimates are regularly 93% more accurate than using traditional models because we collect data from the ground in real-time. Typically, that data is through IoT, sensors, CCTV, toll data from roads, and we aggregate that information together to give a detailed understanding of whether a risk sits inside or outside the event footprint. That’s a binary yes or no answer against an insurer’s risk. We also have a good understanding of whether the risk sits in the primary layer or excess. We can tell the insurer whether the risk sits in or out geographically and also whether the impact is likely to hit their layer.

Do you have to develop a relationship with somebody with access to CCTV data? Or is that publicly available?

Different countries around the world have different solutions. In the UK, we have to make Freedom of Information requests so it’s quite difficult. In the US, there are federal data sources available so we’re able to plug into those and run the data through an algorithm. Moving a stage further, we’ve recently received a significant grant from the European Space Agency enabling us to provide CCTV feeds per property. Taking Amazon Ring as an example, if the homeowner gives us permission to view their CCTV system, we can make an assessment based upon that intelligence. Is their front doorstep wet, yes or no? Three years ago, it was a picture from space. Now it’s a picture from the property itself.

What about crowdsourcing? Is there value in photos of flooding on Twitter for example?

Absolutely. We talk about signals intelligence from IoT, imagery intelligence from CCTV and aerial imagery, but we also talk about human intelligence. We have smart humans, maybe someone at the fire service who has written a report, and we have what we call unsmart humans, passers-by who post on Twitter. We scrape that information within a geo ringfence and apply an algorithm to what is being said. We can specifically look within social media feeds for images that are geo-located and apply them in the intelligence process. It's full-spectrum intelligence capability.

Are you offering those services worldwide?

We’re not restricted to North America, or where we can fly a plane with the correct licences and budget. We look everywhere. We’re contracted into the World Food Programme and the UN. I'm proud that we are a company with a conscience, and we'll work particularly hard at taking what we would do for the insured world, to other parts of the world that we can add real value to. It’s not just about replacing a roof. It’s about replacing a whole community which may have lost absolutely everything. Our staff come from vocational backgrounds of service. Many of us went to Sandhurst where the motto is ‘Serve to Lead’. Having a social consciousness is important. It makes us sleep well at night.

How does an insurer integrate what they know about their clients with the data MIS can provide?

That's one of our selling points to the marketplace. In the beginning, we told insurers that we didn't want their secrets. They could keep their risk and we would supply the intelligence they needed. As the global market has got to know that MIS is a trustworthy company with integrity, we now get access to risks via an API feed.

The claims piece is all about automating. We layer intelligence on top of the risk until we have what we call our confirmation layer, which tells us what has happened to that risk. We give it a grading system based on the impact and offer the grading system based on the functionality. There might be a power station that looks fine because of its concrete structure, but if there is seawater engrossed in the freshwater coolant system, it’s not working functionally. That causes downstream effects for all the industries expecting to get electricity from that location.

Does that API allow clients to plug MIS data into their existing systems?

We've been told quite clearly that clients don’t want another logon. They want an API call that speaks to their system. Our data is agnostic in terms of what that system is. There are lots of different ways to write down the intelligence, or replicate the intelligence, and we do all of them to fit into our clients’ systems. Some have very advanced systems, and some have problems with historic tech legacy. We have to cater for all.

What are you doing with companies outside of the UK?

We now have a sales team in Denver in the US. It's a team we worked with eight years ago from Skybox. We know them well; they know our product inside out and are taking us to market in the US. North American domestics are data-hungry, and I was surprised to find they look to buy all data sources, rather than stick with one company. We thought we would move into the London market first then go to North America, but US domestic owners started asking for our data through the London market. A lot of introductions are also happening because we work with a subsidiary in London.

As part of our sales pitch, we ask potential clients to run us against an event. We will have the data on their system in 24 hours, give them an answer, then ask them to compare it with what they have historically.

Another recent project is a Covid-19 dashboard. What can you tell us about that?

We don’t just work with natural catastrophes; we do all things geopolitical. We've covered the Black Lives Matter protests and how they impacted property insurance all over North America. We recognised the impact of Covid-19 very early on, thought about business interruption, supply chains, and started looking at the Far East and the impact on distribution networks. We pointed our technologies to collect against that and produced a layer which Lloyd’s liked a lot.

The feedback from our US clients was they needed to understand what was happening at a county level. There are over 3000 counties in the US, so we created a database with an automated capability to collect and aggregate what was being said. Humans must come into that and we have human machine teaming as well. We can tell you how each county has treated, from a legislative perspective, the Covid-19 impact on school closures, business closures and restrictions of social movements.

We’re taking the understanding that insurance companies need for natural catastrophes and pointing it towards Covid-19. It’s a very dynamic data set and is available via an API feed subscription model.

What has MIS got coming up that you can share with us?

It was always an ambition for the company to replicate what we did for Hurricane Harvey over Houston, but for the whole world at the same time. That’s what our Global Events Observer now delivers for us. We have spent the last 24 months crafting a technological roadmap with the European Space Agency, and we’re now able to deliver a detailed understanding at sub-one metre resolution anywhere in the world.

The applications are faster for everything that we’re doing, for the scale the Future at Lloyd’s programme requires, and for working with the US domestics at the scale they require. It also futureproofs us against new and novel intelligence sources and the database can be used for underwriting decisions in the future.

We are also speaking to parametric providers about how our intelligence can act as the trigger for their solutions, whether there is a risk to buildings, to business interruption, or agrarian risks wherever they are in the world.

MIS is now a corporate member of InsTech London. What made you decide to join us?

What InsTech London has done is quite amazing. Matthew and Robin have a unified vision for what insurtech should be and where it fits in. There is a very mature view of the insurtech community and a great environment at the events. They bring young companies like ours and more mature companies together in a forum where we can engage with each other.

The larger players are used to slower product cycles, but the small companies have a burn rate that is typically 12 to 18 months at best. They need to be doing things in tighter cycles to be able to prove their capability. From our side, we’re chuffed to bits to be in the community.

...

Published on September 1, 2020 14:49

© 2013 - 2020 McKenzie Intelligence Services LTD