Skip to main content

How insurers can unlock data-enabled business growth

(Image credit: / Pressmaster)

There are many industries where data can be used to drive growth, but there’s nowhere where this is more fundamental to survival than the UK insurance market.

Successful insurers know that pricing a policy is a delicate balance of aiming low enough to be competitive, but high enough to cover the predicted claims cost and make a sustainable profit. Doing this in a price comparison fueled market where every home and every person has a different risk profile means the arms race for data is a fight no-one can afford to lose.

Simply having data is not enough, the differentiator is having the teams and tools which can spin that data into gold.

Indeed, 2020 showed us that speed and adaptability is vital. Deloitte’s recent survey found that 79 percent of insurance leaders believed the pandemic exposed the shortcomings in their company’s capabilities.

The best in the market know that the scale and granularity of insight required to win are now beyond traditional techniques. AI and machine learning are the new digital frontier. Recently we’ve seen them driving business growth in three different ways:

1. Using a real time data pipeline to predict competitor pricing

Price is often a powerful factor in attracting customers, and the UK insurance market makes this particularly true when buying via price comparison websites.

Insurers of all sizes have to work hard to make sure their price lands a coveted spot at the top of the comparison site.

Machine learning techniques including random forest and gradient boost models are now being used very successfully to make predictions such as the best price a customer will find in the market, how far apart the top 5 insurers will be, and the total number of insurers who’ll be willing to quote competitively. The price and product can then be tailored in real time to best match what the customer is looking for.

The most powerful examples of this process in action are in cases where a ‘standard’ pricing model would place an insurer in price positions 5 to 10. In these cases, the models quickly recognize the gap and reoptimize accordingly. All this occurs in under 700 milliseconds.

The data pipeline is the often overlooked element of a set up like this. Whilst it is the ever popular AI or machine learning which makes the final predictions, it is the data pipeline which ingests data quickly enough to know that a competitor has changed their pricing and gives the models a chance to adapt in real time. Enabling proactive features which optimize the price can add 7.5 percent increased commission through additional sales.

2. Forecasting risks

The biggest expense insurers face is paying out claims. For home insurers, one category represents about a third of their claims cost – this is known as ‘escape of water’. In simple terms this is leaks or household water damage.

Insurers tend to look at all risks through three lenses – people, property and place. For a truly data-led approach this can mean looking at over 5 billion internal data points in all.

Typically, the choice is between using machine learning when data is plentiful and perhaps a Generalized Linear Model (GLM) when there may be less data available. However, if the company has the right architecture, then it may be able to employ a ‘multi model’ approach where several models are blended side by side. Advanced statistical tests will indicate whether this results in more predictive power. Combining different models also gives the company an additional layer of parameters that can be tuned as the models are compared to actual claims that come in.

This data-led approach sounds impressive, but on the ground, what it is doing is allowing insurers to intelligently assess customers in far more depth than any other process would allow, leading to prices which more fairly represent what each home really does cost to insure.

To measure the benefit of modern techniques, insurers replay quotes from previous years but price them using their latest technology in order to see how well they would have predicted claims. For escape of water alone, applying these sophisticated techniques should see up to 4.5 percent reduction in claims costs.

3. Using data to target the right customers

Marketing is an area where insurers traditionally spend a vast sum of money either advertising or paying affiliate fees to price comparison sites for new customers, the cost of which has to be covered by that customer’s purchase price.

Paying for digital keywords is an expensive game, so doing this profitably means you need to know precisely which customers are valuable enough to justify the effort, and how to reach them.

Sharing data and machine learning capability across departments is critical to moving the whole business forward. This enables you to train marketing systems using predictive data feeds from the pricing department, improving targeting accuracy and so advertising budget only needs to be spent going after the most valuable customers who are likeliest to convert.

The predictions used in this process go beyond the simple profitability of the policy. The company uses models to predict how long each customer typically stays with an insurer before switching and how often they typically phone their insurer, something which can influence costs.

The effectiveness of these techniques plays on a growing trend known as hyper personalization – realizing the uniqueness in each customer and tailoring your service to match them at an individual level, significantly improving ROI on digital marketing spend

Accurate predictions create a critical advantage

What these examples show is that whilst data is an essential element for growth, data for data’s sake is not enough.

Increasingly industries are turning to more sophisticated techniques to ensure they get both granularity and responsiveness from their data. This is where more traditional processes could simply never compete.

Tech companies often cite their data pipeline as being as essential as the data or the machine intelligence. After all, if you can’t push clean, accurate data into your process quickly enough, you shouldn’t expect much value to come out the other end.

Whilst insurance seems like a hot bed of competition right now, it is easy to see parallels in other industries. Banking has long been faced with the same challenges around which customers to win and making quick decisions about who to lend money to.

As the technology and techniques evolve quickly they still have one very simple goal - those who make the most accurate predictions in the least amount of time have an almighty advantage.

Dan Huddart, Chief Technology Officer, Avantia (opens in new tab)

Dan Huddart is the Chief Technology Officer for Avantia with responsibility for its platform and technology strategy. Avantia is a home insurer that trades across price comparison and direct channels through its consumer brand, ‘HomeProtect’, using its own state of the art machine learning platform. Prior to Avantia, Dan worked in a number of technology and data roles in insurance, telecoms and retail.