How to launch your chatbot – part 2

Customer Experience 2 October 2019

In our first article, we outlined which questions you should be asking before launching your chatbot. Now we’ll show you how to actually launch the bot, including challenges to overcome pre- and post-launch.

After identifying business challenges, distribution channels, chatbot personality, conversational flow, and UX, you’re ready to launch your chatbot! Or, actually, you’re ready for the pre-launch…

Step 1: Prepare for the launch phase internally

Good things come to those who wait. Once your chatbot is ready, you should plan a series of tests to make sure that it works well. Consider implementing a beta release followed by a round of testing before the official launch. Several groups of people, including some who have not worked on the project, should test the bot, giving you a chance to spot and fix any discovered problems.

During this phase, it is essential to get all internal stakeholders involved, particularly to ensure that testing goes well in addition to communications and marketing related to the chatbot. Creating a bot is a capstone project that can affect a brand’s reputation. At the very least, all teams should be aware that the chatbot is being developed and invited to test it during the various phases (beta, go live, etc.).

Step 2: Plan a budget and a detailed communication strategy

If you think your work is over once the chatbot is live, think again. You’ll need to set aside a budget to implement a communications strategy and promote the new service. No one will know the chatbot is there unless you tell them. Though digital natives are used to this type of communication with brands, the rest of the population is still adjusting to this new medium and it may take time for them to get the hang of it.

For the bot to take off, it needs to be accessible and visible via calls-to-action displayed on your site. Social media is another good way to raise awareness (such as Facebook’s Messenger Ads). Calls-to-action can also be integrated into your online interactions with clients (emails, landing pages, etc.) as well as your offline interactions, particularly in the context of customer touchpoints. Teams on the ground must thus be aware of the chatbot’s existence, and perhaps even prepared to answer certain questions (how it works, potential benefits etc.). 

Let’s take the example of the Messenger chatbot for French department store Galeries Lafayette. This tool is not accessible from the store’s website or digital communications; it is only available if customers choose to use it while active on Facebook. What’s more, its capabilities remain limited (personal shopper, inspiration, stores, etc.) which makes us wonder what purpose the chatbot is meant to be serving…

Step 3: Measure, measure, measure!

It can’t be repeated enough: measuring performance is the crux of any marketing project… And chatbots are no exception. 

First, identify your KPIs so that you can measure and analyze your chatbot’s performance. This will help you make data-driven decisions regarding its evolution, such as whether or not to integrate new features or delete certain conversational paths that are not often used. There are many KPIs that could be useful (number of users, conversation length, average number of messages per conversation, drop-out rates,  conversion rates, etc.). To avoid drowning in a sea of data, choose only the most relevant KPIs that will help you reach the goals of yourfor the chatbot.

Once you’ve decided on your KPIs, define your performance monitoring plan which should include both a specialized measurement tool (often built-in to the chatbot creation tool or its hosting platform, e.g. Messenger) but also a process for collecting and recording user feedback. Customer service and human resources should be prepared to track the chatbot’s performance both quantitatively and quantitatively, including analysis of user feedback. 

Step 4: Continuously improve the bot for optimal user experience

The job is not complete even when your chatbot is up and running. The post-launch optimization phase is also very important as it serves to enrich the user experience. User interactions with the bot should be analyzed to understand usage, and to gauge if conversational flows are working or not. Does the bot go far enough? Does it successfully solve user’s problems? Or, alternatively, is it too complicated, failing to meet user needs? Making adjustments as you go can make all the difference.

Don’t forget that the chatbot is an extension of the brand and aims to offer privileged interactions with the client or prospect. As part of customer service, the chatbot must answer questions to the best of its ability and be a part of the brand experience. Imagine how you’d feel if a salesperson or brand representative didn’t understand your question or comment… Well, expectations are the same for the bot!

For example, in 2016 Sephora launched its Beauty Bot to provide advice to customers shopping during the holiday season. Using the chatbot, which was active between November 17th and December 24th, users could interact with the brand to find gifts depending on several criteria including gender, age, and preferences. The initiative was rolled out simultaneously in 5 European countries, including France and Italy. The test seemed to be a success, given that Sephora now has a “full-time” chatbot called Ora. Placing online orders, displaying new products, and making appointments… Ora can do it all! Sephora’s bot is a good example of the test and learn model, to create a chatbot that would be useful to its users in the long term.

Example of the Sephora chatbot (2019).

There is no doubt that chatbots must stay “alive” to stay relevant and evolve over time. Should they ping users after 24 hours of inactivity, for example? Or should they recommend products based on past purchases? We recommend implemententing regular product development sprints in order to improve and promote your chatbot following its launch. You could do this seasonally or more often, depending on your brand and its business rhythm – retail brands, for example, might want to have sprints leading up to Christmas or end-of-season sales.

With these pointers, you should be ready to launch your chatbot! Still have a few questions? Don’t hesitate to  get in touch with our team of experts.

Would you like another cup of tea?