Customize Your Customer Service

Tags: books customer-service programming

Have you considered purchasing over seven hundred customer service numbers? Twilio rents toll numbers for a single dollar a month, and toll-free numbers only cost a single dollar more. Twilio’s great, but they aren’t even the lowest cost provider. At those prices, can you afford to not own seven hundred customer service numbers?

That’s what the developers at did to help customer service agents quickly identify why a customer was calling and what they could do to help:

The company even created separate phone numbers (more than seven hundred in total) that dynamically appeared based on the page and how the customer got to it, so that when customers called a certain number, it would be obvious where they encountered problems.
Analytics at Work, chapter 8, page 142

A casual inspection of the site shows that this program was successful, because it’s still in use:

You can check it out yourself by watching the number in the upper left corner of the screen change as you first visit the site, search for a hotel, and click on a hotel for additional details.

How can you use data to customize customer service?

Anonymously collecting and aggregating it is a good first step. If you aren’t tracking usage of your sites and applications, you’re ignoring valuable data. There is a major US based website that was localized for countries across the globe, and their Japanese site had conversion rates significantly lower than any other location. Were there cultural differences at play? Analytics reports for their multi-page sales form showed that 90% of their customers bounced between two specific steps. Analysts found that the bounce inducing page was never translated from English to Japanese.

You may know about the DRY principle. Some programmers live and die by this one single acronym: don’t repeat yourself. Don’t make your customers repeat themselves, either. Ask for confirmation that the number on their account is the number their calling from, ending in 1234, rather than making them punch in the whole thing.

Allow users to opt-in to deeper data collection. If something goes wrong, don’t just show them an error page and apologize.Allow them to file a support request from the error page and automatically associate it with their account and bundle a log of the activities they were trying to accomplish with the ticket.

Just don’t be creepy. You probably remember Target taking all of this a little too far when it announced a teen daughter’s pregnancy before she did.