Advanced analytics

LAUREN FINNIS: Great to be here.

SCOTT GIBSON: And hello, Laura.

LAURA DODDINGTON: Hi, Scott. Great to be with you today.

SCOTT GIBSON: All right. As I said, we're going to be talking advanced analytics. And in particular, we're going to be discussing WTW's Advanced Analytics Survey. So we surveyed almost 50 insurers across the North American P&C landscape to understand how are they using advanced analytics, and how do they want to be using advanced analytics throughout their business.

This is a survey that we do every few years here at WTW. And I got to say, it was interesting, as I'm reading through, to see how similar the results were to the 2021 edition. Laura, can you help us understand what's going on with, maybe a lack of progress on, advanced analytics.

LAURA DODDINGTON: Yeah, absolutely. So like you say, every two or three years, we put the survey out and we see what progress we're seeing. One thing that is really clear through the survey is that there is huge leadership commitment across organizations to advancing analytics in their organizations.

So 86% of organizations said, yeah, we have leadership commitment to advancing analytics. That's great. And I suspect we'll be very different from if we were talking 5, 10 years ago where in some cases, there were still questions around how much analytics should be used. So we see the commitment there.

But then to your point, when we then look at actual progress around analytics, where are you using analytics? How much are you using it? How are you using it? The numbers are remarkably similar to where they were in 2021. And so as part of this, we say, hey, what's your number 1 challenge? What are your top 3 challenges around using analytics in your organization? And the number 1 challenge that comes out is IT bottlenecks.

Despite all the investment that we're putting into IT and the huge amount of effort that we're putting into IT, IT bottlenecks is coming out consistently as a big hurdle that companies and organizations are struggling with in order to advance analytics in their organizations. As we think about that, I think our IT teams are so busy with huge transformational projects in their organizations, things like policy admin system upgrades, for example, which we know are huge, and really important, and really beneficial.

Sometimes that is squeezing out the opportunity for us to be able to make advanced analytics progress. So, it can be really hard to prioritize the deployment of a claims triage model when your IT teams are so busy thinking about a new policy admin system as an example. And so we're hearing really consistently that that is a major challenge for organizations.

LAUREN FINNIS: And I would add on top of that, in the commercial line space, same issues apply. But then you also have the challenge of just a large amount of really diverse products and an attitude and an expectation across a lot of businesses that every product needs a bespoke solution, as opposed to trying to approach it with more of that CAPDAN: Common as possible, different as necessary approach, where 80% of the framework is consistent across product lines, and 20% is tailored.

So we see a lot of organizations just getting in their way, trying to come up with a custom solution, each different silo throughout the business.

LAURA DODDINGTON: The other interesting challenge that we're seeing, beyond technology and beyond data, is how organizations really bring together their analytics expertise with their business expertise. And so we consistently hear companies telling us that there is a gap between analytics and business.

And so, sometimes that gap might be around, for example, even just speaking the same language. Being able to understand what each team is saying and bringing that together. It might be around being able to prioritize the right business challenges in a way that data scientists and analysts and actuaries can then use to inform the work that they do. And so we do consistently see that there is a gap there that companies tell us about in terms of expertise, understanding across the business and training as well.

SCOTT GIBSON: Brings it back to the data. I mean, the data can be a big challenge. And that's a big hurdle for the IT teams to make sure they get right so that you can effectively use the advanced analytics.

LAURA DODDINGTON: And on that data point Scott, sorry to interrupt, but on the data point, I think it's so interesting to me that when I started my career 20 years ago or so, if we asked organizations what one of their biggest challenges would be, they would say, data. And people would say, I can't do analytics till I fix my data. I'm going to fix my data, and then we can build models.

We asked people today, and so I'd mentioned the top challenge was IT bottlenecks, the second challenge was data. And that was the second most frequently pointed out challenge. And so it's 20 years later, and we still haven't fixed our data. And to some extent, that's because data is hard and it's difficult to do. But to another extent, I think, actually data is just always a moving target.

I don't think we'll ever say like, now we've nailed data. Because there's always going to be more data we can get and more data we can use. And so I think it's a risk if we say, I'm not going to really invest in modeling until I've fixed my data. I think we have to do those two things in parallel. I think organizations, when they start accepting, my data's not perfect, I'm going to try and make it better, but also I'm going to do things with what I have. I think organizations can be surprised by how much you can already do with the data that you have, if it's used intelligently.

LAUREN FINNIS: Yeah. And I think that conversation that's been going on throughout our entire careers through a huge portion of this is the change management. And one of the things that came out in the survey was really about the vision. A lot of leaders said that they had commitment to this, but then being able to articulate to the team, what this transformational journey looks like and develop KPIs and tactical ways to move, we're still not progressing as much as organizations might had hoped.

And also in that change management piece, a lot of the things I found throughout my career with data are that if you're not just using the data you have, the data doesn't transform, because one of the best ways to improve your data quality is to have end users see the value of their data and understand why, it matters how I code my submission, it matters how I put this claim into the system, because I'm going to see the output of that. And if we don't start using these analytical tools, the data considered continues to stagnate.

SCOTT GIBSON: OK. So maybe some of these roadblocks that are stalling progress are a little bit self-imposed. We're waiting for our data to be perfect, but instead we just need to plow ahead and figure it out as we go, and improve it over time as we use it more and more. Now, not everything came up as a stall or a roadblock in this survey. It did highlight some areas where insurers are making progress. Lauren, so where are efforts being successful?

LAUREN FINNIS: So I think there's been a lot of success of late in the past couple of years on machine learning around document and submission intake. We're fresh off InsurTech insights in New York. And that was one of the biggest categories of vendors out there really creating some momentum.

And we're seeing insurers across the commercial line space, not just piloting, but actually putting some of those into implementation, transforming the amount of time and effort it takes to take a submission into the organization and put it in an underwriter's desk to start underwriting. And that's creating meaningful change. And that transformation, there are a lot more data points available to build analytics off of.

LAURA DODDINGTON: I completely agree with that, Lauren, and I think more broadly, I think we're seeing a lot of uses for machine learning across organizations. Machine learning has been around for a number of years now. I think we're really seeing an acceleration of that being put to use for real business uses.

So for example, when we said, are you using machine learning to better understand risk drivers? If we go back to 2021, in the survey, 31% of insurers said they were using machine learning to understand risk drivers. Today, that's 49%, with plans to grow that even more significantly over the coming years. And so I really think that ability to use machine learning to quickly understand and interpret data and see what's truly driving the business, whether that be through portfolio management, whether that be through understanding risk, I think it can add a lot of value.

SCOTT GIBSON: So we're making progress on machine learning and more advanced analytics of that nature. But thinking about the results, what's the biggest area of opportunity?

LAURA DODDINGTON: For me, I think, as I look at areas of opportunity, we've been doing analytics in pricing, for example, for decades now. And pretty advanced analytics in pricing. There's a world outside of pricing that I think is massively under using analytics. If you think about the use of analytics in the claims space, for example, to much better be able to triage your claims, or understand which claims have the potential to go too large, or where there's a potential for fraud -- huge amount of benefit that analytics can bring to the claims space.

And that can be benefit for the insurer and for the customer. And so that's massively beneficial. When we can find those things which help us help our bottom line and help customer experience, I think, those are real wins. And we still see huge opportunity for that to be rolled out in the claims space. Lots of people who have a lot of ambition to do it. But, as of yet, I think the rollout hasn't progressed as fast as we, or as insurers, thought it was going to.

LAUREN FINNIS: Definitely on the non-traditional areas. My background certainly is distribution. And there's so much more that can be done in the analytics space to support distribution strategy. Identifying agency fraud, everything across the gamut there. And we're not seeing a lot of those use cases prioritized.

I'd also say, just making general expense-related and operational decisions more analytically-based. For instance, it's very hard these days, it's so challenging to find people. And I don't find that a lot of our insurer clients have analytically-driven underwriting staffing models or operational support staffing models that can flex with market cycles and understand where they need to be based on key indicators today. What that looks like 3 months, 6 months, 12 months for staffing levels.

LAURA DODDINGTON: I think also, I know this is an advanced analytics survey, but also, I think there's a space for some fairly basic analytics in some of these spaces. So I think there's a lot of low-hanging fruit. You don't necessarily need to go and build like a super advanced GBM model to be able to improve in the area of distribution.

Actually, some really good lead indicator models where you can identify trends quickly, see what the problem is, work out solutions to it, because you've got, let's say, a dashboard that you're looking at on a monthly basis that shows you where some of those issues are. Those sorts of things can be massively beneficial and don't always exist in organizations.

I think as insurers, we spend a lot of our time looking backwards. I'm trying to think about what the indicators are that are going to help us look forward. It's really important. And I think, if I look in personal lines, we've just gone through a hard couple of years in terms of results and in terms of having to take action.

And the insurers, who I think have done best through that, are the ones who are able to identify the issues really quickly, make fast decisions, and respond to them fast and with agility. And I do think that all starts with identifying the issues fast, and that starts with having the right analytics. And that might be more basic descriptive analytics, but things which are measuring the right, things that are going to help you manage the business prospectively rather than just retrospectively.

LAUREN FINNIS: That's a great point. And another thing I'd add, especially in the commercial lines space, our challenges are going to be typically navigated by cross-functional teams across actuarial, underwriting, claims, distribution, operations and technology. And that maybe isn't unique to commercial lines, but definitely probably amplified.

And often, we find that some of those groups maybe only have qualitative data. And people react to an observation or a feeling, because there's not always data and analytical support right away. And so putting tools where, hey, we see this observation, how can I quickly look at what we have and do some analysis to support? Does this observation have legs? Or is it really just a feeling that we should maybe pay attention to but we don't have anything to back just yet?

SCOTT GIBSON: Really great stuff here because, you know, it's an advanced analytics survey, and everyone gets excited about machine learning and AI. But sometimes, maybe it's the broader adoption of analytics across the organization that insurers really need to make sure that they're accomplishing and where there's opportunity for them. So they can move towards the Holy Grail, the data-driven organization as a whole.

Now, Laura, you said that the leaders are committed. That's what the survey shows. So now that our leadership is committed, where are companies going to go next? What's their next step or next move to help increase the adoption of both advanced analytics and just the broader analytics use?

LAURA DODDINGTON: Yeah. Like I say, organizations are committed. And so the intentionality is there. And when we look at the things that people are saying are stopping them, then it's around technology and data and systems and it's around organization. And so addressing those two areas, I think, is really important. So if we think about technology, data, and systems, it's really important, of course, to have data that you can access.

And so while I said earlier, use the data that you have, absolutely do. But also be thinking about the data needs that you have for the future. And so organizations should be thinking about, what should my data warehouse look like in the future? How am I going to be thinking at customer level rather than policy level, for example? How am I going to use generative AI to be able to take my unstructured data and turn that into something meaningful?

There's definitely a value in spending time thinking about your data. When we think about technology, and we work with insurers all the time around technology solutions that enable analytics, and really that's in a couple of spaces, one of those is around being able to use technology to make faster, more joined up collaborative decisions across the organization. So technology can be used as an enabler.

When we think about the fact that it can be hard to translate between data science and state management, for example, actually technology can be used as an enabler to be able to show what the proposed changes are, for example.

And your pricing, and how that was going to impact your customers and your agents. Using that technology in that way can really help to bring people together to make faster, more collaborative decisions.

And then the other big trend we see around technology is externalizing rating. So we are seeing insurers thinking about, actually, I don't want my pricing and my underwriting to be right in the heart of my policy admin system necessarily, because that makes it harder to change and to have the agility and sophistication that I want. And so we're talking to a lot of insurers right now about externalizing that rating. Using technology to bring that out so that it's really put into the hands of the users so that they can make really fast changes to their models, to their analytics, and deploy those quickly.

LAUREN FINNIS: I think there's a lot that can be done on the automation of some of the things that our people that do deep technical analysis waste their time on today. Or waste is probably a overly harsh word, but we have people that are really brilliant at delivering actuarial models or predictive models outside of core actuarial, but they're doing manual data transformation. And I think that that's something that organizations need to prioritize so that people can spend more time doing the meaningful work.

LAURA DODDINGTON: Yeah. Use our smart people to do smart, brilliant things. That only they can do.

On the organizational piece, so, it's one thing to say that leaders are committed to analytics, and they are, but then when we actually start to look at, have you set up the processes that are going to help you to be successful in this? Often, those processes are not in place.

So things like, do you have a really clear vision around where you want analytics to be in your organization? And is everyone aligned to that? Do you have clear analytics goals and a roadmap? What are you going to do for the next three years around analytics? Often, that's not very well-defined.

Do you have a process for deciding, if there are hundreds different things that we could do with analytics, 100 business uses for analytics, which three are we going to pick to focus on? What's the process for deciding that? Often, that doesn't exist yet. Do you have really clear training so that your data scientists understand insurance and your business people understand data science? Often, that doesn't exist yet. And so some of those fundamentals that are going to ultimately help you be successful are not there.

And so what I would say is, for organizations to say, yes, we bought into analytics, but we're struggling to make this happen outside of the technology piece, I do think there is something around thinking about organizational foundations. The same as we would in anything that we're trying to bring into an organization. How are you going to manage change around this? How are you going to think about the new processes that you need to set up and the new ways of working that you need in order to bring this into the organization? I think then that's a gap for a lot of companies right now.

LAUREN FINNIS: Change management, as you mentioned, it's a thing that goes with absolutely any priority. But we see those low-hanging fruit of just talking to people at that individual level about their aspirations as well. I think a lot of our teams and people in the field have ideas of what they might see in optimal future but have no idea of how we might get there. So just some general discussions.

And then that top-down, like if a leader is going to talk about building an analytically-driven organization, that's got to be often and permeate throughout the organization at every single level and never really stop. It's not a point in time. We're going to do analytics tomorrow, and it's done. It's a long, consistent journey. And we find that organizations start to just create some process around that in terms of, if we have started to build some analytical solutions, we're going to reference them every time.

I'm going to send you back. If you bring me something on a sheet of paper that's not coming from the core source of truth for this area that we're building out, even if that might not be everything today. But we always have to start there. Just building some of that muscle memory around how we approach these things. I think too, on the data side, putting some real work around operationally understanding why you're not collecting the data that you want to collect.

If you're seeing gaps, I used to, my old life, I led the customer relationship management system and customer relationship management data, sometimes, isn't prioritized. And so we find that there were duplications in the customer names or people were not correctly coding the industry, which prevented us from doing deep analysis on industry segment trends that we really wanted to do, the business wanted.

But we had to go and sit down at underwriters desks and underwriting assistants and claims teams desks and say, OK. Why is it that this happened last week, and you coded this versus that? What were some of the bottlenecks and breakdowns? You've got to invest a lot of time in changing that behavior because you understand why it was so hard to do it right in the first place, and they're starting to take away some of the roadblocks to better collection.

SCOTT GIBSON: We want to be doing this. We realize there's some challenges. It sounds like the best way to make progress is to be intentional and get to work on it. And work your way through some of the difficulties and find the best approach but stick with it because it's not going to be easy and it's not going to happen overnight.

But what time frames are we talking about? If we're trying to reorient ourselves on how we're tackling advanced analytics or analytics in general, can we make short-term progress in three to six months, or are we looking at a multi-year journey here?

LAUREN FINNIS: Definitely can make short-term progress, but it is, I think, a multi-year journey at the same time. The short-term progress is what Laura mentioned of, what are your priorities? And what are some things that are achievable in the short-term? and setting that, if you don't already have that culture that is centered around this analytically-driven future, setting that and starting that transition. And that can start today. But you have to sit down and meaningfully decide where you want to go.

LAURA DODDINGTON: And what's nice is, if you get started, you can quickly add value. So if you go and build and deploy a claims triaging model, for example, that can really quickly improve your claims outcomes, therefore reducing liability, which is going to impact your bottom line right away. When it's claims, it earns immediately. You don't have to wait for it to earn out.

So, got some bottom line benefit immediately, which can fund the next thing I want to do in analytics, and the next thing I want to do in analytics. And so you can really get some quick wins there.

So I think, to Lauren's point, I think it's both right. I think you need to have a plan. I think it is helpful to know where you're going over the next few years. But then also just do something, get started. Because I think there can be a tendency to want to make things perfect. And you're not going to get it perfect to start with. It just won't be. And so much better to do something, try it, see how it does in the market, and then react and adjust to it. Than to spend all your time trying to get something perfect before you ever deploy anything.

LAUREN FINNIS: And also they're all very interconnected, and sometimes can have just cascading effects. If you're a business that doesn't have a strong, or even any predictive model in a certain area, and you build that predictive model, not only are you improving your outcomes in your business area, but you're also probably reducing your staffing need. Because when you have a poor model, typically, you're having to do double work because you're not only running said model, but then you're having a human deciding how we're going to go back and completely change the output because we don't trust the model. So you get benefits on both sides.

LAURA DODDINGTON: And actually, that's a really important point. And one of the reasons why having a plan is important, too. Because I remember being in a past life where I had a business insights team, for example, and they were generating lots and lots of insights very manually, and it was a time-consuming process.

But we never had time to automate and improve the process, because we were spending all our time manually adjusting things and manually running things. And so, by building a plan and being able to say, this is what the next two years are going to look like, and this is where you're going to get to. But, in order to do that, we're going to have to stop doing some stuff in the short term. And you won't get that report for the next few months that you kind of liked but we weren't really doing anything with. But, the reason we're doing that is because this is the long-term plan.

That then gives you the freedom to invest some time in automating some things so that you can make some other things better, so you can start doing the value-add. And so, do you think being able to have that plan and articulate it across the organization is so important?

LAUREN FINNIS: That makes me laugh! Of all the times you have that report that everybody thought they really needed. And then it goes away, and they realize nobody was doing anything with it. And having those conversations about, which information are we giving people that they actually act upon?

LAURA DODDINGTON: Yeah.

SCOTT GIBSON: That's the tact I take with my children's toys. I hide them in the closet and see if they ask for them.

As we wrap up, any parting words? Lauren, we'll come to you first.

LAUREN FINNIS: I think just remembering — Laura had mentioned this earlier — but it doesn't have to be really deep and crazy analytics. Even pretty standard things can drive meaningful change. I spend a lot of my time talking gen AI and all the latest, new, exciting things, and they will add value, but you can drive a ton of value with just meaningful, small, incremental change.

SCOTT GIBSON: Really insightful, Lauren. And, Laura, any words from you for the audience?

LAURA DODDINGTON: I agree with you, Lauren. There's so much value to be had. I think I would just encourage everyone to have a plan and then go and do something, doesn't need to be perfect on day one. Do something, learn from it, and keep going.

SCOTT GIBSON: Thank you for joining us, Laura.

LAURA DODDINGTON: Thanks for having me, Scott.

SCOTT GIBSON: And thank you, Lauren.

LAUREN FINNIS: It's been wonderful. Thank you.

SCOTT GIBSON: And thank you, everyone, for listening to this edition of the (Re)thinking Insurance podcast.

SPEAKER: Thank you for joining us for this WTW podcast featuring the latest perspectives on the intersection of people, capital, and risk. For more information, visit the Insights section of wtwco.com. This podcast is for general discussion and/or information only. It is not intended to be relied upon, and action based on or in connection with anything contained herein should not be taken without first obtaining specific advice from a suitably-qualified professional.