Finding answers in your own customer emails

Questions that seem easy to answer, but are actually tricky:

  1. Which weighs more, a pound of gold, or a pound of feathers?
  2. Which company is the world's largest maker of tires?
  3. What is the best time and day to reach my customers via email?

Here's what most people answer:

  1. A pound is a pound - they weigh the same.
  2. Michelin.
  3. Whenever they open them.

Bad news - these obvious answers aren't the right ones.

Why we love email for customer contact

Loan servicers and finance customer service teams love email. It's easy to scale and inexpensive to send.

But as cheap as it is to send, it's even cheaper for your customers to ignore.

And if you really need to get in touch with a borrower, it's worth investing the time to figure out these three pieces of information:

  • Which day of the week they're most likely to read and respond
  • What time of day they're most likely to read and respond
  • What message they're mostly likely to read and respond to

Learning from your own customer information

You're the expert on your customers, and you've got the data to prove it.

But until now, it's been tricky to act on that data. Sure, you've got thousands (and thousands and thousands, probably) of emails, but how can you extract the information you need?

We talked to two of our AI engineers about how we're extracting exactly this information for a customer right now:

We're taking in all of their accounts and predicting what is the best time time of day and what is the best day of week for them to send an email, so that when they send that email, they have the maximum engagement and the and they get the most out of it. - Akshat Vaidya

But we're not stopping there.

We're modeling user-level features and email-level features like how they interact with certain emails. We have the timestamps, we have the content of the emails, and the content of the replies. And we can analyze this text data to figure out what is the best type of email to send to a person. We're also modeling features based on payment details, transaction history. And the idea is to use their past interactions, as well as current interactions to keep making the best prediction that we can make for them. - Sheyril Argawal

Why now?

If you could have been extracting all this information all along, you would have, right?

But as Akshat explains, the sheer volume of data you have to analyze has been impossible to manage.

Except now we have AI. And specifically, we have our AI Intent Engine that we've trained on emails and conversations like the ones you have with your customers every day.

We have deep vertical intelligence into the consumer finance industry. We understand and we have been ingesting and analyzing these conversations a lot more. We already have built our own intent engine which understands these calls. And now we are taking the next step into the digital realm where we are saying that all of our learnings in the call space we are able to apply to the digital outreach space as well.

Check out the full conversation here:

Oh, and the real answers?

  1. A pound of feathers weighs more. It's true! Gold is measured in Troy weight, so a pound of gold is 12 ounces, not 16.
  2. LEGO holds the Guinness World Record for producing the most tires.
  3. And when's the best time to email your customers? It depends...but we'll help you figure it out!