Lead generation alone isn’t enough in the insurance business if your company wants to beat the competitors. You also have to know how to convert leads to sales in the most effective manner.
However, this isn’t as simple as it might sound. No magic formula exists to help your company increase its conversion rate in insurance. It’s not like you offer more incentives to new customers or make sure your email marketing stands out, and that’s it. While these tactics may bring you value to some extent, there should be consistent work on improving the company’s lead quality and increasing agent efficiency.
In this case study, we describe the latest insurtech project we have worked on at Intelliarts, which aimed at boosting the company’s sales performance and improving lead generation with a machine learning solution.
Using our customer’s experience, we prove that a data science and machine learning (ML) approach gives a chance to insurance businesses to make their decisions more data-driven. Thus, they can achieve a higher lead-to-sales conversation rate in the future.
Our customer (under NDA) is one of the fastest-growing health tech insurance companies. This insurtech company specializes in selling Medicare-eligible insurance coverage by working as an intermediate between US insurance giants and their end-users.
Before entering the market, the customer discovered that today’s insurance sector is subjected to an increasing diversity of insurance plans, which usually creates a confusing enrollment experience for buyers. The company has decided to bridge this “information gap” by making contacts with prospects, usually with call calling and other successful sales strategies, and helping them choose the best coverage. Only when the lead is hot enough to make a deal, this insurtech business transfers them to the insurance agent of one of their partners (a large insurance company). The agent takes the initiative and makes final moves to close a sale.
In some sense, our customer’s business model is reminiscent of insurance brokerages when a third-party company helps consumers check different insurance plan options and select the best fit for them. Still, there is more to it since our customer doesn’t leave the leads as it is after the enrollment but tries to follow up and keep in touch with them to be sure their partnering company has provided the most quality service to the end-users.
In brief, the customer’s workflow can be described as follows:
The insurtech company operates within an online lead auction that has a base of insurance lead phone numbers
A bot places bids to get a lead, and the partner (large insurance company) that has placed the highest bid gets the phone number of the lead
The customer shares the lead’s contact with this partnering company
The insurance agent calls the lead
Now let’s go over to the company’s pain point as well as lead generation for insurance with machine learning.
The company is a big supporter of disruptive innovation and always seeks ways to improve its business operations. It has contacted us as the provider of technology consulting and software engineering services with the request to improve its sales effectiveness by generating leads with ML After a series of the first meetings with the customer and after we dived deeper into their business model, we came to an agreement that the best way for them to convert leads into sales will be through:
Improving lead quality
Increasing insurance agent efficiency
The vast majority of insurance companies can use the same formula. Let’s discuss briefly the reasons.
Why improve your insurance lead quality?
When we’re speaking about insurance lead generation, quality always trumps quantity. In theory, the more leads you have, the better, since more leads will convert into sales. This is far from reality, though. Your company can generate as much as 10,000 insurance leads in a week but convert, let’s say, only 2500 of them because the rest didn’t intend to make a purchase nor had the power to buy.
In comparison, your insurance company can convert 8000 out of 10,000 generated leads if it’s focused on quality leads. In this case, the business wins by spending less time nurturing the lead and increasing its return on investment.
If the insurance company goes with non-quality leads, it potentially has higher expenses. As Jacob Rodriguez, the director of sales, auto, and home insurance at ActiveProspect, puts it:
“In the insurance world, too many bad customers making too many claims can really destroy your loss ratios.”
At the same time, most insurance businesses do not experience problems with getting new leads, but struggle to transform them into customers. Again, this refers to poor and inconsistent lead quality. An overwhelming majority (87%) of insurers recognize it as the critical issue they’re dealing with, according to the poll conducted during the insurance lead gen webinar.
What about insurance agents’ efficiency?
Having a base of quality leads isn’t enough for an insurtech company. You should also have talented insurance agents in your team, so they could convert those leads into sales. Skill shortage is now holding back innovation and growth in the insurance sector, as evident in the study below. Still, the success of your insurance company is directly linked to your insurance agent’s efficiency. So, insurers should invest in increasing their agents’ productivity consistently.
An insurance agent is a sort of mediator between the company and the lead. A good agent advocates for the interests of the insurer, aiming to close a deal. Nonetheless, the customers’ interests go first, and a skillful insurance agent does their best to leave the person satisfied. Communication with insurance agents is the number one factor that affects the Net Promoter Score (NPS) after the first contact between the customer and the insurance company. 42% of leads/customers say it’s important to them, as compared to, for example, 22% of those who mention price.
Both improving insurance lead quality and insurance agents’ efficiency could be achieved by taking a data science and machine learning approach, which we have taken in this project to help the customer boost their sales effectiveness. Now let’s discuss how the solution was built.
Read also: Customer Churn Prediction in Insurance
As said, our ML solution for insurance lead generation was composed of two parts. The first one was related to increasing the lead quality. Our major goal was to determine the effectiveness of leads (only those used in telemarketing). We wanted to score and filter them better and, hence, improve the lead quality.
Analysis of zip codes and SCF codes
The Intelliarts team has started with analyzing data at a zip code level. We have received the demographic data (age, marital status, mortgage rate, population density, and so on) and financial data (Adjusted Gross Income (AGI), Gross Total Income, etc.). And our data science experts made an effort to find a pattern between these demographic data and sales data (conversation rates and the total number of insurance policies sold) on a zip code level. They aimed to predict the average profit in the regions to choose those where sales would be the most profitable.
The preliminary research helped us understand, though, that we didn’t have enough sales data for a quality data analysis at the zip code level. So, we moved from a zip code level to using a destination sectional center facility (SCF). This refers to a geographical area defined by three-digit zip code prefixes.
The following data analysis provided us with several insights into the lead quality:
1. We have discovered that the conversion rate was a bit bigger in sub-urban areas
2. There was no linear dependency between AGI and conversion rate
Our next step was to detect SCF codes with low conversion. So, our data engineers labeled all SCF with the conversion of less than 20% as bad (0) and the rest as good (1) and built a principal component analysis (PCA) visualization. The latter stands for an unsupervised ML technique that is used to reduce the dimensionality of data and helps to visualize data.
From the two graphs above, we can see that good and bad SCF aren’t separated visually. This implies the need for more detailed analysis and the use of more complex ML models.
To go further with the idea of lead generation with ML we tried to build a classification model for good and bad SCF. The ML model should have been highly explainable to product owners. It’s important for the management to be able to see the patterns on their own and understand whether the logic of the ML model coincides with the business one. So, we used the decision tree approach, along with clustering analysis techniques, parameter tuning, and feature engineering.
To explain it further, the decision tree is a classic algorithm that is built in such a way that it divides data into two groups based on some characteristics in every iteration until it meets the stop criteria. As a result, the decision tree provides the dendrogram of decision-making. This is a tree-like diagram visualizing the relationship between the data points in the system, according to which we could understand why the SCF was bad.
Besides, clustering analysis helps with clusters, which we could then compare to statistical tests and interpret how one cluster (good) differed from another (bad). In the end, we achieved the result of predicting correct good SCF (91%). We also excluded 41% of low conversion SCF (bad), which led to improving business metrics and optimized agent calls.
Analysis agent efficiency
Next to the ML-based insurance lead generation solution, we aimed at increasing the efficiency of insurance agents. We hypothesized that those agents that sell well (quickly, much, and the most successfully) deserve the most promising leads (those who potentially are the most profitable). And this was possible to do since the base of the top-performing agents was updated regularly. Specifically, the insurance agents were divided into three categories (low, medium, and high), depending on the results of their work.
The idea was to build an algorithm that could detect agent efficiency and connect the top-performing agents with the most promising leads. Our data scientists collected a lot of data about agent calls and aggregated this data on different levels, for example, by partners, lead source, state, and opportunity for sales. All this data affected the probability of conversion and the income received from that conversion. Hence, we built a statistical model that should consider all those parameters and return an expected agent conversion per week.
This expected conversion was then contrasted with the actual conversion rate and statistical significance. So, the insurer could decide whether the agent has to be promoted to a higher tier, dropped into a lower one, or stay within the same one. Respectively, when the insurance agent is moving to the higher tier, they get more quality leads and could earn more. This should improve their motivation to work better and increase their efficiency afterward.
Speaking about the results we achieved, the ML-powered solution on quality lead improvement helped the customer establish a better lead selection strategy. Now the insurtech company doesn’t have to waste time on SCF with possibly low conversion. It can focus instead on the most promising SCF regions in terms of profitability. Accordingly, the conversion rate should increase, and the customer’s profitability should be maximized. The already achieved result is up to a 5% increase in lead quality, which shows the efficiency of the built machine learning solution for insurance industry.
Related reading: Learn about lead scoring which can help you improve lead quality.
The solution we built has also inspired the company to try to create a lead-level model in the future. This ML model should work in a way that the team can check the information about the lead with its help.
Then, the team decides whether it’s worth placing a higher bid in the auction to get the lead’s phone number. If the probability of selling the lead later is low, then the company won’t buy it.
If we consider insurance agent effectiveness, this solution increased the overall conversion rate as well as the conversion rate in all tiers. The explanation here is simple. Insurance agents work harder to be able to move to a higher tier and earn more. The health insurtech company has already achieved a 3% increase in agent efficiency.
Increasing your sales conversion rate in the insurance sector means that a company earns new customers while also saving precious time and money resources. Still, wasting your time contacting those leads that have little to no chance of converting into sales isn’t reasonable. Instead, the company can focus on targeting quality leads to increase its conversion rate. This is exactly what we tried to achieve together with the health insurtech company when building a machine learning solution to improve their lead quality.
Another goal of the solution we built and deployed was to increase the insurance agents’ efficiency. By assigning the most skillful agents to the most promising leads and, hence, allowing the agents to earn more, the ML model works well to improve the motivation of insurance agents and increase their efficiency. It then improved the conversion rate.
In case your insurance company is looking for talented data engineers, the Intelliarts team is ready to assist. With deep industry knowledge and expertise, we can help you increase your conversion rate, improve lead quality, optimize prices, or solve any other business problem relevant to the insurtech field with our ML application experience.