In this column we talk quite a bit about the value of data and its almost magical power, but the value to the agency is often unclear or doesn’t seem initially compelling. Without clear examples of the gains that are being made via data analytics, it’s hard to pull the trigger.
The problem with early-stage innovation is the first steps always model and only marginally improve our old ways of doing things. The first car looked more like a horse-drawn wagon than a modern car. In moving to paperless processing, we embraced the PDF, which is just a digital copy of the paper. This is simply the way we evolve. We can only understand new concepts in the context of the old.
So when we think about using data in our agencies, we gravitate toward incrementally improving what we already do. This leads to marginally interesting conversations about automating all the manual reports we deal with or creating a small productivity gain for our customer service colleagues. While these efficiencies are positive, the value doesn’t always balance the change commitment required to get serious about data. So agencies often simply decide not to decide. But choosing to do nothing is putting us in a precarious position. With a 237% surge in insurance technology investment over the last 12 months (with virtually none of it going to existing agencies), we need to be aware of the risk of complacency.
The advantage to being late to the data party is we don’t need to invent the process and guess at the outcomes. There are abundant examples from other industries. So let’s take a look at some data-driven outcomes that go beyond legacy process improvement. As you read, let your mind wander to how these types of insights and outcomes could benefit your firm. This dreamy part is required because reaching this level of capability requires us to fundamentally challenge the idea that the way we work today is the right way.
Reality of Retail
If you’ve ever pulled out a loyalty card to lock in a sale price on bananas, you’ve already participated in the front end of a massive data analytics process. Whether you’re at the grocery store or Nordstrom, the things you buy increasingly determine what is in the store, how it’s presented to you and what you pay. In fact, more and more stores are linking their customer shopping data with all aspects of how they manage their company. Distribution systems decide quantity and even brands per store based on predictive models that move the inventory in and out exactly when it’s needed. These models and decisions are made automatically—no human effort or thought is required.
As retailers amass increasing amounts of customer data (how much is spent, what is bought and when) they are able to achieve almost mystical prediction rates. Through predictive modeling, one large grocery retailer in the U.K. can predict the exact day customers will return to the store and how much they will spend on that visit within £10.
While this example is staggering, in most cases this high success rate isn’t required—even a 20% foresight over the rest of the market can create an enormous advantage. The result is a retail structure that is significantly different from it was 20 years ago.
It’s impossible to talk about retail without mentioning Amazon, which barely qualifies as a retailer in the classical sense. Why? Amazon doesn’t have to sell products for more than their acquisition cost. Its analytics are able to predict when it will sell out of a product before the company even buys it. When you know you can sell 10,000 barbecue grills by 11:37 tonight, you can make some interesting decisions. Amazon can pinpoint situations where selling products for less than it paid will maximize profits through both volume and investment return on the float (the time between when Amazon sold the grills versus when it has to pay the supplier). This is the key benefit to true predictive modeling. The insight is almost always counterintuitive.
While a human can dream up creative approaches like this, an analytics system can prove them. In the old world, retail meant buying low and selling high. Today, retail is matching buyers to products at the exact moment the probability of sale is highest. Legacy retail strategies don’t stand a chance.
Ad Brokering—Every Click, Somebody Gets Paid
Don’t let Facebook fool you. You may see it as an online scrapbook for pictures of your kids, but it is actually a data aggregation and targeted advertising firm. If you are an “all-in” user, Facebook probably knows more about you than you do. Everything you say and do is merged with everything your friends say and do. These data are compiled into an internal profile of your habits that is fed into a predictive model that guesses what you will respond to with impressive accuracy.
Facebook understands the value of the raw data it holds and will likely never sell it. Instead, Facebook sells the opportunity to target you based on what it knows about you. So, as you click a friend’s post about his new car, Facebook’s ad exchange delivers an auction bid for the ad that will appear on your next screen. It provides a hashed (or somewhat obscured) version of your ad profile that the advertisers match to their own profile of you (yes, everyone has a file on you). Interested advertisers bid, and the deal is closed. The ad is placed on your next screen refresh and—cha-ching!—a receivable is posted for Facebook.
This entire process happens in the time between when you click and when the next page is displayed: about 100 milliseconds. The ultimate microtransaction performed billions of times per day. It’s so fast that it’s said European advertisers initially could not complete the auctions in time due to limitations on the speed of light between their systems in Europe and Facebook’s U.S.-based data centers. Thanks, Einstein. And guess what? Facebook makes money on that ad placement whether you click the ad or not.
How does it work? Data analytics in real time. And it doesn’t take much imagination to twist these examples into our industry. Noted predictive analytics expert Eric Siegel once defined analytics as “the power to predict who will click, buy, lie or die.” Sounds pretty relevant to insurance. Gaining even a slight advantage in identifying when an organization will change structure or leadership can help us intelligently identify at-risk accounts and, better yet, identify the at-risk accounts at our competitors. Identifying risky behavior helps us segment clients and prospects by potential profitability. Analyzing our firms’ core competencies against current opportunities can help us cull unlikely wins from the pipeline and steer our production teams to a stronger book, especially when that path is non-intuitive. We can use analytics to identify inefficient internal processes that drive expenses up and to develop models that provide direct insight and new services for clients. In short, everything we do can be improved by analyzing our data.
Like all large-scale improvements, this all starts with a few small steps. Get data-literate and get out of the hardware and process business. This game isn’t won with IT—it’s won with intelligence. Can we achieve outcomes as impactful as these examples? Absolutely, but we have to start from the beginning. Decide this is important and take the initial steps. And don’t decide to do nothing. People who already know how to do this are entering our space. My own models predict with 100% certainty that your agency will do this or wish it had.