September 6th 2023
Originally published in Risk & Insurance (September 2023). Reproduced with permission of the publication.
In my early 20s, I’d often find myself standing for minutes in my nearby Publix trying to figure out which item to buy.
Did I want the $2 name brand product, or the $1.50 Publix generic product? Did I want the less tasty Gala apples at $1/lb or the tastier Honeycrisp apples at $2/lb? I would stand paralyzed by the decision.
Thankfully, at some point, stomach pain kicks in and forces you to make a decision.
But in real life, there are many people who are staring blankly, maybe even petrified and anxious, in the aisle of predictive models trying to figure out how to decide between different companies. If you can withhold judgment at the horrible intro you just read, I’ll do my best to present to you things you should consider to help move you from trepidation to triumph.
First, two caveats.
The first is that this is not comprehensive. Writing a thorough treatise on this would take too long and the editor already thinks I have wasted 150 words of my 900 word limit. Second caveat is that this article is focused on analytics firms providing underwriting, pricing, loss control, or claims-level solutions that are commonly marketed as AI.
There are a lot of considerations going into this decision, but it’s worth checking this point off before you start heading to the analytics store.
Building something in-house or even with outside support could be a better option if you’re looking for a more customized solution, have in-house technical expertise and resources, and/or need a quicker turnaround.
On the other hand, I have definitely seen companies start off with great intentions and then end up with more money spent and less of a product than if they had just outsourced the work from the get-go.
Going with an external vendor can mean avoiding diverting too much attention and resource from other parts of the business, having immediate access to specialized expertise, proven technology, and on-going support.
A Request for Proposal (RFP) process or something akin to it is absolutely crucial to ensuring a methodical approach to evaluating the potential vendors. Internal stakeholders with previous experience are the most ideal, and it’s important to also incorporate other stakeholders if not into the actual process then at least to get feedback at the outset.
Experience is key but a characteristic not to be overlooked is independence. Whomever you have leading the process, make sure that that person or group of people have the best interests of the company at heart, with no selfish motivations apparent that might impair their judgment.
It’s important that you, or the people you trust, understand the model design and how it compares against competition. I’ve seen models promoted as being robust and cutting edge, but once you get through an NDA and delve into the model details the model is shockingly dated and not well suited for anyone’s needs.
It’s also good to give it a test drive. How does the model perform on your own data? Having real-life examples can help you get comfortable with important details like the potential range of ROI and what a model “error” or incorrect prediction might mean financially and operationally.
As you go through the RFP process, some firms will connect you with smooth-talking salespeople and flashy decks persuading you that your hard-earned budget will generate an immense return.
However, after you sign the contract, you quickly realize and regret that the immediate attention you received from that salesperson is now replaced with inattentive and rigid interactions with the people actually doing the work.
As you go through the RFP process, make sure you differentiate those two groups, and talk to the actual people behind the scenes.
A subpoint to this is understanding how important of a client you are for the company. Are you a smaller or larger client for them? Will they be willing to work with you on changes in the model because you’re a valued client, or will they brush you off because you’re booked as a win on last month’s sales quota?
It’s great when companies say they want to share in your success and ensure there will be little downside to spending your budget with them, but what if they were able to codify those assertions in a contract?
Having a contract designed such that you pay little when there’s little to no results, and you pay more when you’re swimming in ROI is a true partnership.
Of course, the devil is in the details. You have to not only define “success” but be able to objectively quantify and monitor it over time.
While there are many other considerations at play and each situation is different depending on the buyer’s history, sophistication, needs, goals, etc., the above points hopefully will help reduce some anxiety when looking at the different options in the predictive modeling aisle.
Learn more about our actuarial solutions by visiting https://davies-group.com/northamerica/solutions/insurance-services/actuarial-solutions.
Originally published in PEO Insider (March 2020) Reproduced with permission of the National Association of…
Every time I attend a professional sporting event, I am always in awe…
Underwriting and claims have a lot in common. Both are easily considered afterthoughts…