Further | Blog

Part I: How Further is Using Machine Learning & AI to Solve The Digital Lead Problem

Written by James Johnson | Jun 19, 2023 9:22:57 PM

Digital Customer acquisition for Senior Living Communities is hard.

There are many challenges that make connecting with digital prospects difficult, but in recent years one problem has become a particularly acute:

For every 100 leads sent to a senior living community, that team is doing a great job if they generate five move-ins.

This means there is a lot of wasted effort by sales teams and low return on investment for ad spend by marketing departments. This problem has been getting worse every year: leads go up, conversion goes down, sales teams experience more failure and as a result, occupancy doesn’t improve fast enough.

This is the most universal issue we see at Further with senior living communities we work with.

A simple solution to this problem would be generating a small number of higher-quality leads, but this is a difficult thing to do.

How can you identify which leads are likely to move in, when it takes more than 3 months on average for that to happen? How can you identify which leads are a good fit for your community before a salesperson speaks with them?

AI Can Help Solve the Digital Lead Problem

There has been a lot of buzz recently about ChatGPT and advances with AI and there are many exciting implications for the senior living industry. One of the opportunities Further is most excited about is leveraging Machine Learning to improve digital lead conversion.

Background

Further has been working on the lead quality and segmentation problem since our start in senior living in 2016. This came out of a necessity, because when Further was first installed on communities with high traffic volume, the number of leads we would generate would quickly overwhelm the community salesperson.

We work with our agency partners on the SEO targeting and SEM campaigns, but there are limits to how much control you have over who comes to the website.

For some of our customers where this problem was most extreme, we started asking prospects questions in our flows, and avoided creating leads based on certain answers. These questions and configurations got more complicated and customized over time, but we learned certain questions were great indicators of a prospect’s likelihood to move in! 

To give a tangible example: we identified that if a prospect scheduled a tour after viewing pricing instead of telling us they couldn’t afford the community, the difference in conversion from Lead to Move-In was enormous:

 

Now you might read this and think this is obvious, but at the time we were under the (wrong) impression that prospects wouldn’t be truthful when interacting with a digital experience. It turns out they will. 

Where Does Machine Learning Fit In?

Companies that have successfully used Machine Learning to drive business value typically have a large data set to work with. Further has developed our own unique data collection, based on our normal business operations. In order to continually improve performance for our customers, we capture as much data as we are able to about the digital customer acquisition funnel.

We capture all of the marketing data, behavioral data, website engagement data, conversational data and move-in data. Our primary job is to drive more move-ins for the communities we work with, so measuring end-to-end performance is essential to be able to make improvements. 

Following our discovery of the large delta between prospects who take different paths, we asked ourselves several questions:

  • What if we looked at everything prospects do?
  • Could we use Machine Learning algorithms to train on our dataset of millions of conversations?
  • Could we predict how likely a prospect is to move in, solely based on their digital journey?

Over a period of 6 months we analyzed, dissected and modeled out the behavior of 80 million visitors, 300M events, 7 million conversations, 1.5 million prospects and over 40 thousand move-ins. The results (to us at least!) are mind-blowing:

  • We can predict how likely a prospect is to move into a specific senior living community with 80% accuracy through the digital journey alone
  • 75% of move-ins come from just 18% of prospects
  • The bottom 55% of leads represent only 3% of move-ins

You can view the results of real data funneled through our prediction algorithm below. The Score is how Further rated the prospect and the other columns are the breakdown for how each of those groupings of scores actually performed:

Taking a look at the extremes of this analysis shows how effective this kind of approach is at predicting who is likely to move-in and who is not:

  • 39% of leads represent just 1% of all move-ins. These leads convert 0.06%
  • 26% of move-ins come from 2.5% of leads. These leads convert at 22.13%

How To Use This Data

Being able to predict who is going to move into your community creates new opportunities to improve performance that didn’t exist before. Here are the four actionable steps that senior living operators can take using this new information:

Marketing: Marketing teams can do more experimentation and create faster feedback loops by using this predictive move-in score and driving it higher over time.

Sales: Sales teams should not be investing as much time in the 62% of leads that represent less than 7% of all move-ins. 

Reporting & KPIs: Lead quality is just as important a metric as lead volume as it relates to occupancy.

Training: Create Sales KPIs and processes around the top 18% of leads to improve sales teams experience and performance.

 

Community Implications

In addition to the corporate implications, it is also important to educate community sales teams on how predictive move-in analytics can help them hit their goals. Further communicates this data to the community via our dashboard and email notifications.

 

If we can help sales teams win more, waste less effort, and drive higher quality leads, all these items will result in better performance and higher occupancy for senior living organizations. 

If you are interested in learning how we developed this training algorithm keep reading here for the details.

 

Keep Reading: Part II: How Further Created Our Move In Prediction Algorithm Using Machine Learning