Nine key learnings after building the first Swiss retail chatbot

Last April, Facebook introduced the Messenger platform, making it possible for developers to connect with more than 900 million people around the world. Since then, there’s been a lot of buzz about chatbots and we’ve seen the ecosystem launch into rapid growth.

At Valora, we strive to be at the forefront of the latest trends in customer engagement, which is why we developed and launched the first retail chatbot in Switzerland. Swiss customers can chat with the ok.– chatbot to receive the best ok.– deals and redeem them at more than 800 stores located in Switzerland.

Introduced several weeks ago, the goal of this experimentation was to learn how chatbots work and determine if they could become an effective digital channel for retailers. Now that we have accumulated sufficient data, we are ready to share our first key learnings with the community.

Build experiences that matter.

People don’t care about chatbots, they care about the benefits you offer them. When building a new product, it’s easy to become so enthralled by the technology that you lose focus on everything else. But for any product to be successful, you must also focus on the customer needs.

We have seen and tested many bots and one of the biggest mistakes we’ve witnessed is bot builders trying to replicate existing experiences, like one you might have with an app. Chatbots, however, are a different kind of platform that warrant a different context for interaction.

Our advice: Start from scratch and think about the best way you can leverage the Messenger platform with a customer-first, benefits-driven approach.

Make your design flexible.

We’re still in the early days for the Messenger Platform and many customer experiences simply aren’t possible, due to the lack of specific features. That said, the Messenger Platform is evolving rapidly — we are already in the 1.3 release since the first launch in April. Promisingly, each release has arrived with a set of new features and experiences that help build better experiences for the customers.

Our advice: What’s not possible today might be tomorrow. Keep your design open and flexible so you can adapt and improve your chatbot design as soon as Facebook brings new features to the platform.

Building a bot is surprisingly easy.

Building and maintaining an app can become really complicated and expensive, really fast. As a result, we were surprised by how easy it was to build and deploy a chatbot.

You can build a chatbot from scratch in a matter of days. If you’re using a platform like Chatfuel, you can build it even faster and can launch in a matter of hours.

At Valora, we built and deployed a chatbot using several tools that are available to developers: Api.ai; Dashbot; Heroku; SendGrid, Microsoft Luis, Amazon Web Services, Papertrail.

This complex communication system is managed by a central component that we call the Bot Brain. It controls and manipulates data, invokes the appropriate services, and handles outgoing responses.

Our advice: Before writing any line of code, use a platform like Chatfuel to build your MVP and test it with some users.

User acquisition is easy, while retention is a challenge.

Once you’ve built a chatbot, the most difficult part is finding users. The good news is that we’ve found it relatively easy to get people using our chatbot.

Why? There’s less friction with chatbots than there is with applications — people don’t need to download anything, they can start a conversation with your chatbot right away without having to register or create an account.

Spending less than $500 in advertising, we reached our first hundred users in less than a day, and reached our goal of 1,000 users three times faster than expected. And the users we’ve acquired are engaged — we’ve received more than 8,000 messages on the platform.

Like any other digital channel, the challenge for chatbots will be user retention. You have to find ways to continue engaging with your customers so they don’t forget about the chatbot and use it on a regular basis.

We believe chatbot retention is going to be even more challenging than on an app due to Facebook’s strict guidelines around re-engagement. If you don’t abide by these guidelines, the alternative Facebook offers is sponsored messages, but of course, those come at a cost.

Our advice: Think about how you will foster continuous engagement with users when you design the chatbot experience.

It’s difficult to measure which acquisition channels are the most successful.

While it’s easy to get users, it’s difficult to measure which acquisition channels are the most effective. When you run Facebook ads, distribute flyers, or publish Facebook posts, people are redirected to Facebook Messenger. While you can measure the number of clicks, you don’t know which clicks convert to a chatbot user because people still have to press “Get Started”.

With that, it’s nearly impossible to track which channels are the most successful, making it challenging to decide which one would be the most effective to leverage. Recently, however, Facebook seems to have solved this issue with the support of referral parameters in the latest Messenger Platform release.

Our advice: While it wasn’t possible to use this feature when we ran our first user recruiting campaigns, you should leverage these new features to make your user acquisition more effective.

Conversing with your customers is still a long way out.

In last few years, NLP has made tremendous progress and is now good enough to understand most user intents. However, we still have a long way to go when it comes to having open conversations with customers via chatbot.

Because there are still gaps in NLP technologies, we’ve noticed that it’s more effective to guide users with actions so they don’t have to type excessive text to interact with the chatbot. Using menu buttons and quick replies are an effective way to drive the conversation forward and help the users get what they need, fast.

Our advice: Constrain your use cases so that the bot you build will be able to answer most of the questions from customers within that specific case.

People don’t mind receiving messages from businesses, as long as they’re useful.

Maybe the most exciting aspect of chatbots is the opportunity for businesses to reach out to customers directly within the Messenger app. Of course, customers are likely concerned about spam, not wanting their messenger inbox to become a place full of junk. The good news is, Facebook has strict guidelines when it comes to sending messages to customers within the Messenger app.

The best chatbots, however, actually offer value to customers. When this is the intention, people don’t mind being contacted by brands through that channel. Our chatbot, for example, sends notifications to customers who saved a deal, reminding them when a deal is about to expire and that they should think about visiting a store to redeem it.

This notification was appreciated because our customers found it useful. After this notification was sent out, we noticed a spike in redemption, and took note that only a handful of people decided to unsubscribe after receiving it.

Our advice: Only send notifications your customers will find valuable.

Online to offline conversion rates look promising.

When we came up with the idea for this chatbot, we didn’t know how effective chatbots would be at sending users to our stores to redeem the deals.

In our case, the ok.– chatbot has been incredibly effective at transferring people across channels, from online to offline. So far, we’ve seen a 10x increase in deal redemptions at stores compared to other channels, and more than a third of our users visited a store after using the chatbot.

Our advice: Again, focus on customer benefits that are enticing enough to pull your customers offline.

We’re still in the early days.

There has been a lot of hype around chatbots — some industry voices are even saying chatbots will kill applications. With that in mind, it’s important to understand where we are on the hype cycle curve.

The truth is, Facebook has still a lot of work to do to streamline the platform and make it really a good tool for developers to deploy great customers experiences. Product managers also have to do their part and learn how they can successfully build impactful chatbot experiences for users.

Our advice: Keep abreast of what’s happening with this trend but don’t be quick to jump to any finite conclusions.

Final Thoughts

Chatbots seem to offer an interesting channel for the retail industry. It’s still too early to know what the detrimental use cases might be, or whether or not customers will continue using chatbots once they realize brands have inundated their messenger inbox.

We believe that if product managers and Facebook focus on offering customers experiences that actually matter to them, chatbots have a bright future.