Narvar’s mission is to enable convenient shopping experiences after a purchase is made by providing consumers with up-to-date and proactive tracking information. They recently created a bot for Messenger that helps their retail clients keep their customers updated on the status of their orders and provide real-time answers without the need for calling customer support or searching for information online.
We caught up with the team at Narvar to learn more.
In 100 words or less describe your bot ie. tell us what your bot does.
We provide proactive package tracking updates and answer related FAQs. When consumers make a purchase from a retailer using Narvar’s platform, they receive a link to a tracking page which provides up-to-date information until the order is delivered. The consumer is given an option to sign up for additional updates via SMS or Messenger. If they choose Messenger, they are sent an initial status message by Narvar’s AI-driven bot to start the conversation. Consumers can interact with the bot and ask questions about their package. Narvar’s bot answers common natural-language questions about package tracking and delivery, personalized to that customer.
What technologies/tools were used to build the bot?
Wit.ai was the main NLP engine for the conversational part of bot. We also used our own NLP engine to handle FAQs that we expected to get based on our own data & experience, which were required to be customized on a per-customer basis. Tools used included Image attachments, Button templates, Generic template, Quick replies and Sender actions.
What Messenger platform features do you leverage for the bot?
We leveraged conversational and visual features of the platform to send rich, branded notifications to customers on behalf of retailers and to respond to FAQs.
How was your experience designing and building your bot? What did you learn? Was there anything you learned that you didn’t expect?
Overall it was a great experience. One of the first things we had to learn was to slow things down rather than responding at machine speed to make the experience more natural. There were some interesting challenges in figuring out how to handle a multi-tenant, multi-page application with a single bot (there was no mechanism in place at the time to do this, so we had to solve for it on our own). One of the biggest surprises was how much more engaged users were right away in this channel. Narvar has a history of sending proactive notifications to customers for the last 4 years via SMS. We immediately saw response rates much higher than our SMS feature with some users simply being polite and saying thank you!
We have also been able to facilitate new levels of brand engagement and capture great positive feedback through the bot integration. We drive net new connections between retailers on our platform and their customers, setting up long term relationships.
Who are you (or the customers who you are building for) trying to reach and why is a bot the right way to reach them?
Narvar’s bot for Messenger helps our retail clients keep their customers updated on the status of their orders. Rather than encouraging customers to call support (which can be time-consuming for the consumer and expensive for the retailer) or visit a general FAQ page which requires the consumer to work for the information, retailers can send these proactive delivery updates and provide real-time answers to their questions to keep them engaged during this period of high anticipation in the customer journey.
How is building a bot different from a traditional mobile app?
It really boils down to conversational nature of the interface. It’s much less clinical and cut-and-dried, so you have to adjust the way you think about the user experience to flow differently, be broader and more personable.
What is your favorite feature of your bot? Why?
Our devs like how users interact with the bot – it’s very personalized to them, and they respond as if they’re not talking to a computer, but as if there’s personality behind it. It’s pretty cool seeing users interact with something you built as if it’s an actual being.
What metric did the bot fulfill for you: awareness? Engagement? Retention?
Currently the best measure of success is adoption rate as well as engagement rate. We’ve seen the speed of adoption of proactive updates via Messenger ramp up incredibly fast compared to the same type of information made available via SMS: on average, of consumers who opt-in for these additional updates, about 30% are choosing Messenger vs. 70% SMS – in less than a year. We also see an average of 7 interactions with the bot. That’s twice the number of visits they typically make to the tracking page, indicating that they are interacting with the bot and establishing a deep connection with the brand.
What are the methods of bot discovery that you are using that have been effective?
Our bot is unique in that it is currently 100% personalized to a consumer based on an order they’ve placed with one of our retailer clients, so discovery is being driven through an initial transactional email confirming their purchase and giving them the opportunity to opt-in for additional notifications via the bot.
Do you feel the bot achieves the goals you set out? Would you do anything differently next time?
Next time we’d start earlier!
What will you build next?
Our bot continues to evolve as we add & fine-tune features. We are also actively evaluating related retail use cases to expand our bot to provide additional support for consumers.
We chose Messenger primarily because of the scale and engagement of the user base, which made it the most obvious choice to reach the greatest audience quickly.
Our Building Bots for Messenger series highlights different experiences on the Messenger platform. These businesses and developers have approached building their bots in a unique and interesting way, and are seeing success on the Messenger platform. Look for more bot profiles right here on the Messenger blog.