I order something on Amazon, but the item did not arrive even though the tracking said it was “delivered”. So I looked for their customer support, and initiated their chat session.
These days, customer support chat is becoming more and more popular. Sometimes, there’s a live person responding. Other times it’s a chatbot.
From a business perspective, a chatbot could save a ton of money so that they can minimize the number of live person to respond to customer inquiries.
From a customer’s perspective, chatbot could be frustrating when you type in your questions and the chatbot doesn’t understand. Because chatbot is typically fairly limited in terms of understanding a human’s free-form questions, this happens quite often.
Amazon solved this problem by presenting a few canned answers for a customer to choose from. A customer is ONLY allowed to choose his reply from given options, instead of free-form typing.
From Amazon’s perspective, this substantially limits the error cases, almost down to none, as the customer’s reply is always predictable. This is pretty much the same as a typical automated voice system when you call a customer support phone line. But it flows much better because you can always see all the options and the chat history.
In my case, canned options worked perfectly by me choosing the following options:
“Got it!” “Yes, that’s it” “Didn’t get it” “Package never arrived”.
After choosing these 4 replies, the chatbot finally gave me a new information, saying that “OK, looks like the delivery status may have been updated too soon. The good news is it should arrive soon. Give us until the end of day Fri, Dec 06. Really appreciate your patience with this. Is there anything else I can help with?”
Then another group of reply options were presented. I don’t remember all the options, but I chose “What if I don’t get it?”
Then the chatbot replied: “If for some reason it doesn’t get there by then, definitely come back & we’ll request your money back or send another one.”
So I replied with: “OK, will do”
Chatbot: Is there anything else I can help you with?
Me: “No, that’s all”
Chatbot: “OK. We’ll post any updates and next steps in Your Orders. Thanks for choosing Amazon. I’m here to help whenever you have an issue.”
The good news is, that the item did arrive on the date the chatbot suggested. So it all worked out well.
Now a question still remained. The tracking never got updated, neither on Amazon website, nor USPS tracking site. Both remained as “delivered” which was a wrong status. I can only assume that when the chatbot responded that “the delivery status may have been updated too soon”, Amazon server actually communicated with USPS server and retrieved the real information. Or, it simply tried to buy some time without any data.
If Amazon was able to get a real information from USPS that reflects the reality, then that same information should have updated the tracking status of Amazon order page and USPS tracking site.
But because it didn’t, that made me wonder if the chatbot reply was really trustworthy or not.
These days, more than ever, various services are integrated with each other. In this case, my user experience of tracking the status of my order at Amazon consisted of me interacting with Amazon chatbot and USPS tracking website.
Even though the item actually arrived on time as the chatbot mentioned, USPS tracking and Amazon order tracking did not reflect that, which resulted in my mixed feeling. I wasn’t sure if the information that the chatbot told me was really true, until I finally received the item.
In an ideal world, I should have received a tracking update from Amazon, based on an accurate tracking update communicated from USPS to Amazon, so that I didn’t even have to contact Amazon in the first place.
These are the things that UX designer should keep an eye on, when designing a user experience. As you start looking into an end-to-end user experience, you may find that it’s actually a lot broader than what you initially thought.