Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Max Wicks on the Price Writer Podcast Episode 26
23 Plays1 month ago
In this episode, we’re joined by Max Wicks, a pricing professional whose journey from professional boxing to general insurance is anything but typical. Max talks about falling into pricing via a graduate scheme, how his motor trade background gave him a head start, and building a career without following the traditional actuarial exam route.
We discuss his move from GLMs to GBMs, and how data science is transforming pricing teams. Max shares practical tips on building models more efficiently, navigating legacy systems, and improving transparency in modern methods. He also highlights the value of curiosity, subject matter expertise, and what insurance can learn from tech.
Perfect for anyone in pricing or data science looking to modernise their approach, question legacy thinking, and rethink career paths.
Recommended
Transcript
Introduction and Guest Welcome
00:00:00
Speaker
Hello everyone, welcome to the show. We've got Max Wicks with us today. It's brilliant of him to be with us. Very kind and we're looking forward to hearing from Max. Hello Max.
00:00:11
Speaker
Hi. How are you? Brilliant. It's great to see you Max. It's so kind of you to come along and join us.
Journey to Pricing Analyst
00:00:17
Speaker
So the first thing i wanted to ask you is, how did you get where you are today?
00:00:22
Speaker
ah So a lot it was almost by accident, maybe as a lot of other pricing analysts ends up. I knew at school, sort of maths was sort one of my strong subjects.
00:00:32
Speaker
So I went to uni, did maths, and then in that summer time, straight after, trying to hunt around, look for things, talk to different recruiters, get an idea of what to do.
00:00:44
Speaker
And then had one call up and talk about a graduate opening day at an insurance company for the pricing analyst position. And so I went along and found some of the sort of tests and things that we went through interesting.
00:00:57
Speaker
And I did reasonably well at them as well. So I got offered a job quite quickly and sort of fell into it, really. And I think my background a little bit. So I have family members in sort of like the car trade. and also that do racing so yeah already had a bit of ah interesting cars not quite as deep as them of course and sort of gave me a little bit of an advantage as the first insurance job i had was in the motor insurance industry yeah i think most of us have a favorite product so is car insurance your the one you find most most interesting
00:01:38
Speaker
ah Yeah, most likely. I think the one that i definitely have the most experience in these days I've done touched on sort of home quite a bit too. i Some know other niche products here and there are NFUM, the vast majority in car.
Balancing Boxing and Analytics
00:01:54
Speaker
So Max, I know you've sort of come from a different background to a lot of people. ah Yeah, so obviously when I first started, a lot of people were talking about the actuarial exams um and a lot of people that I worked with took them.
00:02:11
Speaker
ah However, for me, at the time, I was actually boxing professionally as well. So to fit in work plus training and then find some time to study for the actuarial exams would have been...
00:02:26
Speaker
I think nigh on impossible because they're not the UC exams after exam software So rather than taking those, I still managed to do some sort data science-y kind of studying here and there, but it wasn't quite much of a rush and I could do some more application of that on the job.
00:02:46
Speaker
Yeah, being able to train for boxing and studying for actuaries, I don't think would have gone well. So what happened with boxing? I was doing quite well, but I ended up being a little bit injury prone and and a little bit. I had some injuries that occurred not long before fights.
00:03:07
Speaker
I had to pull out of one or two and then went into another one carrying an injury because it was a big fight. and So I just made it a whole lot worse. i oh oh no did you feel pressured uh well it was for a southern area eliminator okay it was 10 round fight if i won that i'd have been boxing for the southern area title right um and the week before in the very last sparring session
00:03:38
Speaker
ah I re-injured my hand from an old injury that I had. But because it was a big opportunity, I still went in and took the fight, ah but had to pull out in the up, like ah think it was in the seventh round in
Advice on Dual Careers
00:03:53
Speaker
the end. just So on the whole then, what do you what do you think is the, if somebody was thinking about, if someone's a boxer or something similar, but they also like sort of the general insurance pricing side, what would you give them as advice for their career?
00:04:08
Speaker
and Get used to a lot cold, wet morning runs. and
00:04:17
Speaker
Yeah, just when opportunities come up during the workplace to look at those new technologies and they develop your skill sets, utilize that as much as you can and because you do have that much more restricted outside side of work time to be able to focus on that. So new projects, new ah methods that do sort of pop up here and there, or even ask, ah can I create a GBM for this and see how it performs against it.
00:04:50
Speaker
and We've got this new data coming in rather than as an into our GLM models, can I have a go at seeing how well it would perform with GBM? Just sort of do that kind of side of thing u and so that you can most you to like learn on the job because it's more difficult to do so outside.
00:05:12
Speaker
Yeah, I definitely think that's right. I mean, PriceWrite is trying a lot to get people what they need in order to actually learn these things. But it is quite hard just to learn on the job if there's no no support particularly. And you're actually right, people being curious and saying, well, can I try this is the is the first step.
00:05:30
Speaker
That is actually how we got GLMs originally adopted. So that is the right way to go.
Transitioning to GBMs
00:05:36
Speaker
Awesome. OK, so what would you say then your mission is for insurance pricing?
00:05:44
Speaker
it's ah At the moment, it's a lot of sort of working on moving over from GLMs to the GBM side. Sort of just improving that efficiency, not having to sit there for every factor, can create some more models, test more data, just so much faster, just getting it out, getting it done, automate, as well as having similar levels or better levels of accuracy with the price. And what would you say is because sort of the biggest reluctance? What's kind of the barrier to people making that switch over?
00:06:17
Speaker
I think it's largely due to just legacy software. and So many ah insurance companies are on old software that they've been using for tens, twenties of years.
00:06:32
Speaker
And to change over, to completely retrain current staff, as well as IT departments having to bring in new software, new licensing agreements, paying, all of those kinds of things, essentially building from the ground up again.
00:06:50
Speaker
um sort of stops a lot of them in that sort point. yeah Yeah, I think that's definitely right. like And what do you see as being the biggest kind of benefits of switching over to the GBM?
00:07:02
Speaker
That speed of being able to create models, my automating a lot more easily than sort of the previous. emblem and so on of fitting.
00:07:14
Speaker
Yeah. Rather than taking a team of analysts to to four weeks to do get out some parallel-level models, it can take one or two data scientists but a week or less kind of side of things.
00:07:30
Speaker
Mm-hmm. No, think that's right. And what about the sort of statistical fits? do you feel the GBNs outperform the JLNs? In the testings that I've generally done, they've either outperformed or been very similar.
00:07:45
Speaker
ah For the most part, they outperform in some cases where with JLNs, do have that little bit extra flexibility, those low exposure sort of niche areas.
00:07:56
Speaker
where subject matter experts can have a bit more influence that can help improve on that side, but you can have those same sorts of adjustments after the fact with GBMs.
00:08:10
Speaker
What would you say for people about the explainability side?
Challenges with GBM Adoption
00:08:16
Speaker
Previously, was definitely quite a contentious factor. um These days with partial dependence and Shapley clots and so on, and feels like every so few months a new Python package is up trying to improve the explainability side.
00:08:36
Speaker
It's not perfect and it'll never be as completely crystal clear as a GLM because of for each level of the tree, there will be some interaction, but it's so much better than it was and you can still get those general trends.
00:08:51
Speaker
Yeah, I think I think there's a tendency to think about the explainability in terms of comparing it to how did we explain GLMs? But really, it's quite different. It's it's just as explainable, but but in a different way.
00:09:07
Speaker
And to be honest, GLM isn't necessarily the gold standard. I think we've all seen times where the people the linear predictors going in a different way to the one way so straight away that's quite hard to explain in a GLM isn't it because you're going to start talking about the correlations and the things going on in other factors and crossings over in the exposure so to be honest It's not we all got used to after a long time of working with GLMs of being able to explain them.
00:09:37
Speaker
And I don't think it's necessarily something intrinsic that it's something that we all learn skill to actually do ah because you certainly can get linear predictors. They're a different shape to the one way.
00:09:50
Speaker
mean, one of the things for me is that explanations should be narrative, narrative based and evidence based. And you can certainly still do that with a GBM in a similar way to the GLM.
00:10:01
Speaker
You touched on as well. There are some perils and situations where I do think the GLM still is probably a good a good tool. I don't think it's an either or thing. I think they both feel different needs in different ways what would you what advice then would you give to people that are thinking about kind of testing the gbm and they've got sort of the problems you mentioned like there's a skills gap or there's a sort of concern over the it side and things like that what what advice would you give to people
00:10:32
Speaker
Yeah, I think it's a lot of newer faces within businesses or like coming straight out of uni will have started learning this stuff already. So allowing them to explore more freely and just sort of taking the leap. um Once you start having a guy, it becomes a lot less daunting than it seems.
00:10:54
Speaker
h I imagine when they first started using things like Emblem, when their senior analysts loaded up in front of them, they were probably a bit scared too.
00:11:05
Speaker
But now, five, 10 years later, it's all they want to do. Yeah, I mean, I'm going to show my age, but yeah, we it's not like we all adopted emblem and glm's in a great big rush of enthusiasm kind of 15 20 years ago that was an uphill struggle in itself um which you see actually there's still lines of business that are still resisting sort of scientific based pricing anyway plenty of places on excel rating tables that an underwriter or actuary drew up together
00:11:38
Speaker
Yeah, definitely. So that's it's sort of a continuum and it's a journey that people are on in different lines of business or in in different
Future of Pricing in Industry
00:11:48
Speaker
places. um But now, absolutely. I can remember showing people chairlands and being sort of given a hard a hard time about it because people didn't understand that.
00:11:57
Speaker
And in fact, having just very. very similar questions of well can you explain that to me that doesn't match up to the one way actually or uh it's the one way steeper than that why you're not charging people enough and things like things so not dissimilar conversation and in fact it's not like we also didn't have conversations about how prices had moved in ways that would not have matched up to the now i say bias or kind of assumptions made about how the prices will move and i'm going to remember one conversation now that was to do with rated area where i was asked to move a post very specific postcode bear in mind there's 1.6 million just because this the yeah particular person knew that postcode and felt that it was high risk
00:12:46
Speaker
So could we put it in a different higher band? And, you know, like having knowing a postcode as a person and understanding how that is relative to 1.6 million other postcodes is not a very easily done thing. So what do you see as your kind of vision for the future? Where do you want us to go or where do you want yourself to go in pricing?
00:13:17
Speaker
Um, just sort of getting it more into that data science software engineering, almost a level of skill. And you think the level of some of the data science within pricing teams is slowly catching up to that, that we would have seen in the just general technology industry.
00:13:39
Speaker
Yeah. So like based on Google and Facebook and things like that, obviously they've had world-class data sciences. creating things for them for a long time.
00:13:51
Speaker
And and they their tools are slowly making its way out to the wider industries as well. So yeah, just that thing of being able to be so much more rapid with deploying new models, testing new data, and things like that, rather than having of six month to year long test or remodeling cycles ah and of some places it's even much longer than say a year as they review their rates outside of just small adjustment adjustments based on underwriting input just every sort of couple of years but with the gbms and the not needing to take
00:14:35
Speaker
two to three months to recreate the price and just build upon what you already have. I think that would be a good place for the industry to move towards. Yeah, i think that's I think that's right, getting more of the rapidity. And um I think bringing together lots of skills, because I do understand the kind of concern with areas like large losses that are quite hard to understand statistically.
00:15:01
Speaker
But on the other side, areas like in-car sort of third-party damage, Cars are kind of cars. Then there's we we know a lot. We've got a lot of data and information about those that we can understand and model. And you're absolutely right. Being rapid, getting getting to market, getting out with good analysis and getting it deployed in a much more rapid way does make a lot of sense.
00:15:27
Speaker
Is anything you'd like to add? um So outside of the areas we talked about. I think one thing that I do notice sometimes is how far having a good sort of common sense works within the actual modelling space. Yeah. I think it's underestimated a lot of just having a little bit of a subject approach.
00:15:48
Speaker
knowledge of being, seeing something that doesn't feel quite right, I think. And it so as we move towards the data science side of things and GBMs and further and further into more niche data and more enrichment, we can sometimes lose that bit of, does this and look right?
00:16:10
Speaker
Yeah, I definitely think that's fair. I e i think domain knowledge is a little bit underrated, actually. and So I'm not saying everyone needs to be sort of a car enthusiast or a builder or whatever to to do these things. But knowing about what you're ensuring does make a lot of difference. um Even on the side of like data reconciliation of seeing or just like the validation of models.
00:16:40
Speaker
seeing if something completely goes against what you would expect can help a lot and possibly find that you find some data errors within there that you may not have picked up on had you not had that to faint knowledge.
00:16:53
Speaker
Brilliant.
Contact Information
00:16:54
Speaker
All right, Max, so it's really kind of you to come in to be on the show today. It's been really good catching up. It's nice to see you as well. And I'm glad glad things are going well. So if um if people want to get in contact or anything, how should they drop you a line?
00:17:06
Speaker
ah They can find me on, obviously I'm on LinkedIn.com. Great. Thanks so much for for for coming on the show today. And we'll, learn yeah, thanks a lot. Have a good rest of your day. Thank you. Cheers. Thanks. Bye.