Nick Lamparelli interviews Shannon Shallcross, Head of Client Services at Pinpoint Predictive about predictive modeling and its important function in the Florida Insurance marketplace

FULL TRANSCRIPT

1
00:00:00,000 –> 00:00:04,030
…Hi We’re back. We’re at the 2023 Florida

2
00:00:04,030 –> 00:00:09,020
Chamber annual insurance summit in…Disney World.

3
00:00:09,020 –> 00:00:12,565
Orlando, Florida. And who am I speaking with?

4
00:00:12,565 –> 00:00:16,545
I’m Shannon Shallcross I’m the head of client services from Pinpoint Predictive.

5
00:00:16,545 –> 00:00:22,255
And what does Pinpoint Predictive do? So Pinpoint Predictive is a predictive modeling company

6
00:00:22,255 –> 00:00:28,050
We do risk predictions for the insurance industry, and it’s not cat predictions it’s not cat

7
00:00:28,050 –> 00:00:33,930
modeling but what we’re really looking at is the individual behaviors of a person.

8
00:00:33,930 –> 00:00:39,920
Understanding that someone’s behavioral characteristics…has a bearing on if they’re gonna file

9
00:00:39,920 –> 00:00:45,165
a claim…if they’re gonna be involved in negation if they’re gonna cancel early.

10
00:00:45,165 –> 00:00:51,945
And it’s kind of a premise that is similar to the fact that credit is used in insurance so with

11
00:00:51,945 –> 00:00:56,570
credit…the the understanding there is you’re a good

12
00:00:56,570 –> 00:01:01,205
insurance risk if you have a high insurance based credit score.

13
00:01:01,205 –> 00:01:05,805
Similar premise for us that we’re looking at the behaviors of an individual but our process does

14
00:01:05,805 –> 00:01:12,105
not use credit at all Mhmm And our predictions are available at the earliest possible point So

15
00:01:12,105 –> 00:01:18,155
really before someone even fills out an application, You just have their name and address Don’t know anything else about them.

16
00:01:18,155 –> 00:01:21,205
You can pull a prediction to understand their risk propensity.

17
00:01:21,205 –> 00:01:26,090
Okay So don’t use credit score Correct Okay So…that’s unique.

18
00:01:26,090 –> 00:01:32,020
How is that possible? Magic. Imagine we’re we’re in Disney We’re in Disney world It’s very

19
00:01:32,020 –> 00:01:38,935
fitting…No Just kidding so The way that that’s possible is we’re using a different

20
00:01:38,935 –> 00:01:43,825
data source and it’s a data source that usually is not used by the insurance industry.

21
00:01:43,825 –> 00:01:47,775
It’s behavioral like omics data. And when you have enough

22
00:01:47,775 –> 00:01:51,795
of this type of data and I’m talking about trillions of data points Okay.

23
00:01:51,795 –> 00:01:56,310
And we have this on two fifty million US adults.

24
00:01:56,310 –> 00:02:00,700
there are a lot of things that you can do when you would apply deep learning machine learning in

25
00:02:00,700 –> 00:02:05,890
the modeling process And that’s what we do Yeah So that’s amazing So what would be I’m assuming

26
00:02:05,890 –> 00:02:12,572
social media, what what would be…No So no social media And the reason for that

27
00:02:12,572 –> 00:02:17,802
is when you do social media scraping, there is a potential that you’re adding bias into the

28
00:02:17,802 –> 00:02:22,942
modeling process and we have put in some pretty tight controls to make sure that we’re not doing

29
00:02:22,942 –> 00:02:28,162
that especially in light of some of the you know the NAIC had just really to bulletin on

30
00:02:28,162 –> 00:02:33,060
insurance companies’ use of artificial intelligence, we wanna control for bias.

31
00:02:33,060 –> 00:02:38,510
So we’re using data that is behavioral like an my data So it’s essentially what people are doing

32
00:02:38,510 –> 00:02:42,395
on screens what media they’re consuming, what are they buying.

33
00:02:42,395 –> 00:02:48,205
again they’re they’re data points that literally have nothing to do with what we’re predicting.

34
00:02:48,205 –> 00:02:52,865
But when you’re able to develop this behavioral fingerprint of a person Yeah In the model

35
00:02:52,865 –> 00:02:55,905
training process we can find out Okay.

36
00:02:55,905 –> 00:03:01,510
So this behavioral fingerprint looks a lot like someone who ends up becoming a…or it looks a

37
00:03:01,510 –> 00:03:06,770
lot like someone who is gonna file three claims in the first policy period.

38
00:03:06,770 –> 00:03:11,760
But on the opposite side of that coin and this is the part that is really important for carriers

39
00:03:11,760 –> 00:03:17,402
You can also identify who do we know is gonna be profitable at that really early point.

40
00:03:17,402 –> 00:03:21,732
So that’s very helpful to know because then you can understand when exactly you can quote and

41
00:03:21,732 –> 00:03:26,825
bind as quickly as possible Yes so it’s it’s that fast.

42
00:03:26,825 –> 00:03:30,985
So we’re talking potentially seconds to Subs tracking.

43
00:03:30,985 –> 00:03:34,745
Sub seconds to turn information around. Yes Yes.

44
00:03:34,745 –> 00:03:38,670
So we’re essentially…the integration is an API integration.

45
00:03:38,670 –> 00:03:43,930
So wherever a prediction is pulled, name and address, you get a prediction.

46
00:03:43,930 –> 00:03:49,120
Amazing. so why Florida? Why Why Florida Well why are you here?

47
00:03:49,120 –> 00:03:54,370
My question. Because of the magic…Well I I think I already know why Florida because you’re

48
00:03:54,370 –> 00:03:58,810
predicting claims but Yeah Let’s let’s specifically talk about like why Florida?

49
00:03:58,810 –> 00:04:04,095
Yes So a couple things about the Florida market are really important to our clients.

50
00:04:04,095 –> 00:04:08,350
number one, the citizens’ depopulation…has been

51
00:04:08,350 –> 00:04:12,860
something that’s been a really hot topic It’s a great opportunity for carriers here.

52
00:04:12,860 –> 00:04:18,180
But at the same time, the onus is on the carrier to really understand the nature of risk as

53
00:04:18,180 –> 00:04:23,627
they are evaluating…any new book of business to acquire It doesn’t even have to be related to

54
00:04:23,627 –> 00:04:30,357
that particular thing…but we are providing a prospective view of risk It’s not looking back

55
00:04:30,357 –> 00:04:33,397
It’s looking forward, and it works really well.

56
00:04:33,397 –> 00:04:38,217
The other one that we’ve had a lot of success with in Florida is a litigate prediction.

57
00:04:38,217 –> 00:04:41,390
Yeah Because litigation, I I know that reforms

58
00:04:41,390 –> 00:04:45,450
passed…last December, and those have been really great.

59
00:04:45,450 –> 00:04:51,110
But this has just been a constant issue for years and years and years in insurance where, there

60
00:04:51,110 –> 00:04:56,472
are loopholes…in the auto industry and the homeowners industry…And as soon as those

61
00:04:56,472 –> 00:05:00,682
loopholes are found, they turn into enormous problems for carriers.

62
00:05:00,682 –> 00:05:07,242
So it’s really helpful to be able to understand what’s the likelihood that this policy

63
00:05:07,242 –> 00:05:12,435
holder is gonna respond to a doorkeeper…that’s actually something that you can predict Yeah

64
00:05:12,435 –> 00:05:19,020
Yeah And and Steven Weinstein…who’s premier Permuting…reinsurance expert He was interviewed

65
00:05:19,020 –> 00:05:24,640
by AM Best recently and he was going he was talking about how this has been a problem forever

66
00:05:24,640 –> 00:05:29,720
Yes…In Florida’s like first it was, you know mold is gold.

67
00:05:29,720 –> 00:05:35,680
Right And then sync sync whole problems Like, so I was there for that.

68
00:05:35,680 –> 00:05:40,660
And so like that’s really interesting because I think in when I think when most most folks think

69
00:05:40,660 –> 00:05:47,650
of citizens takeout, they think of cat modeling…and trying to optimize the portfolio…

70
00:05:47,650 –> 00:05:53,605
Yes But this is like an additional layer on top around like okay…the problems that we’ve had

71
00:05:53,605 –> 00:06:00,405
in the last few years are not really cat…oriented They’re litigation oriented And you and you

72
00:06:00,405 –> 00:06:06,070
and that’s not gonna probably go away with the reforms They’ll just find some other…entrance

73
00:06:06,070 –> 00:06:11,400
Exactly…It’s about that people part of the risk And especially in Florida that people part of

74
00:06:11,400 –> 00:06:17,750
the risk, can turn into very high dollar claims if you aren’t able to get your answer out of it

75
00:06:17,750 –> 00:06:24,072
Yeah…so personal lines commercial lines, So we’re doing business in both personal lines and

76
00:06:24,072 –> 00:06:28,412
commercial lines personal lines primarily auto and homeowners insurance.

77
00:06:28,412 –> 00:06:35,192
For commercial lines…we do predictions for companies that have less than twenty employees

78
00:06:35,192 –> 00:06:41,792
because we found that using this process there’s a very strong signal with the business owner.

79
00:06:41,792 –> 00:06:44,992
So the behaviors of a business owner actually do play

80
00:06:44,992 –> 00:06:48,835
out in their loss experience and their loss trend.

81
00:06:48,835 –> 00:06:55,382
So commercial lines…we’re operating there as well workers comp Okay So I’ll tag…I’ll tag

82
00:06:55,382 –> 00:07:00,952
Pinpoint and yourself on this How how should people get in touch with Pinpoint or for yourself?

83
00:07:00,952 –> 00:07:06,652
They they can reach out to me happy to talk with them We have an awesome sales team too We have

84
00:07:06,652 –> 00:07:12,522
information on our website as well so really happy to have any discussions We work with carriers

85
00:07:12,522 –> 00:07:18,650
every single day on developing…strategies like this that help them to optimize all of their process.

86
00:07:18,650 –> 00:07:25,650
Okay So we we are in Disney World. So…who is your favorite Disney character…Oh

87
00:07:25,650 –> 00:07:32,650
boy This is not something that I We can include Pixar…You know, this

88
00:07:32,650 –> 00:07:37,810
you’re gonna laugh but I have to say the little mermaid, and this is a throwback to when that

89
00:07:37,810 –> 00:07:42,540
movie first came out, my younger sister it was her favorite movie.

90
00:07:42,540 –> 00:07:46,460
And my fun fact is I know every single line

91
00:07:46,460 –> 00:07:50,600
to that movie, every single song that’s in that movie.

92
00:07:50,600 –> 00:07:57,552
Because she probably watched it at least two hundred times when I was growing up Yeah So old school…here.

93
00:07:57,552 –> 00:08:03,122
So Pinpoint Predictive is using new technology, but it’s being sold by old school.

94
00:08:03,122 –> 00:08:07,003
Yep. Yeah K Shannon thank you Thank you.