Designing a chatbot for reordering your favorites

Jet.com was a shopping site where you could order groceries and essentials and have them delivered to your door. As Jet’s UX writer, I worked with our Product team to design a chatbot you could text whenever you needed another bag of Doritos.

Team Product | Jet.com

 
 
 
 

Project summary

Challenge

Make it easier for busy customers to reorder the things they need most.

Solution

Create an SMS chatbot that will automatically add items to your cart.

What I did

  • Helped design the conversation flow

  • Delivered a final script for our MVP

 
 
 

Process

 

Identifying the need

While you could buy virtually anything on Jet, our main focus was on groceries and essentials, and reorders made up a large chunk of our overall orders.

Our target customers were young, urban, and tech-savvy, and were often juggling a career and starting a family. They were busy, and one way we could help was by making it even easier to reorder their favorite brand of diapers, hand soap, and pasta.

By building a bot that you could text throughout the day with your grocery list, and have items automatically added to your cart, we’d be saving our customer both time and effort. All they’d need to do was check out.

 
 
 

Creating a flow chart

We wanted to keep things fairly simple for our MVP, and designed the conversation to cover 3 main scenarios: 

  • if we found an exact match in the customer’s order history

  • if we found only a partial match

  • if we couldn’t match the utterance with anything the customer had previously ordered

We also designed the flow to capture the customer’s email, which, along with the phone number they were texting from, would have to match what was on their account in order for them to use the bot.

 
 
 
 

Explorations

Once we had the flow down, I started exploring content. With a 160-character limit and 3 broad scenarios, each bot response would have to be short, concise, and pretty vague.

I also kept track of any questions, ideas, or opportunities for additional responses that came up while I was writing and that I could bring up during our product sync.

 
 
 
 
 

Identifying new opportunities

At our next sync with Product and Engineering, we reviewed my drafts and discussed any changes we’d need. Because this was an MVP, we didn’t want to be too prescriptive with our responses — we wanted to learn about our customers’ natural patterns and behaviors, like how they formatted lists, if they even sent multiple items at once, or if they included brand names.

One new opportunity I identified that was in scope and made it into the flow was messaging our $35 free shipping minimum. We ended up adding 2 new responses to entice customers to check out or let them know when they were close to getting free shipping.

 
 
 
 

Finalizing the conversation

After a few more rounds of revision, I delivered the final script with 12 distinct bot responses. It was important that we communicated “reorder” as opposed to just ordering, since the bot would only be able to add items that had previously been ordered.

 
 
 
 
 

Outcome

We were in the middle of building and testing when the project was de-prioritized, so I sadly never got to interact with our bot. Still, I learned a lot from the process — mostly that even the simplest bot isn’t really so simple after all.