A little caution never hurt anyone. So double-check these things before your chatbot launch.

We published a short handbook on preparation for the first chatbot deployment not that long ago. The mentioned steps should ensure that you deliver value to the end-users while maintaining superior quality. However, as the project speed accelerates towards the launch, certain aspects might be omitted or neglected at the very end. It is fair to say that nobody wants their work to go belly-up, right? We are not here to cool your enthusiasm, but let us take five minutes to make sure that you have covered and tested the following areas:

Handle manual and automatic tests

Testers are at a premium throughout the project and particularly close to the launch. There are various good practices around their work, but today we will spill some tea about our approach. First, we share the dialogue flow and the list of cases with the tester (Miro is our best friend here). They check if the bot appears correctly on all channels, using different devices. The documentation should be detailed enough for the tester to check the chatbot’s behavior in specific scenarios. Let us say that one of the features is to fill out the form. The bot should inform about process completion, redirect the user to a pre-filled form and minimize itself to show the fields. Sometimes it is worth providing clickable UX wireframes so that the tester can easily compare the desired and actual state. We believe that testing does not end the second we deploy on production. That is why we advocate preparing automatic tests. We record test scenarios at the beginning of dialogue configuration to trigger them regularly. Post-go-live, it happens automatically every week, plus after any changes to the flow. Whatever your practices are, make sure to handle one final round of tests before the launch. If possible, consider automation — to us, it is worth its weight in gold.

Check if your audience can see each component of your assistant (and the channel behind it)

We bet that at this point, your UI experts and testers know the design of the window inside out. If you are one of them, take a deep breath before reading the upcoming sentence. Let us check how it looks once again. Don’t hate us just yet — we have been there ourselves, we mean well! Do you know that subpage where the chatbot covers some crucial information should stay minimized? How about the cookie bar that appears from nowhere and covers half of the window? Oh, we almost forgot about a survey prompt that overlaps the send button. These could be the protagonists of a horror story, but we aim for a happy ending here. While going through pages or screens of your channel, emphasize mobile. According to Statista, mobile devices account for almost 55% of website traffic. If you launch your solution on various channels, bear in mind the limitations you may face. WhatsApp, for instance, offers less sophisticated UI options. Be ready to check if the simplified components fill the role of those on the website.

Prepare an analytics panel to quantify your chatbot’s performance

Peter Drucker once said: ‘If you cannot measure it, you cannot improve it.’ This quote resonates with our beliefs — we put a strong emphasis on analyzing our activations to identify areas for improvement. If you’re also a supporter of an efficient analytics panel, you may want to verify what kind of data is collected. Be sure that you capture vital metrics such as containment and coverage –some systems require additional integrations or context variables to enable them. If you are using an internal BI environment, handle the integration with chatbot analytics with reasonable notice. You will likely need to integrate with the engine’s API by providing key and authenticators. Give yourself some time to compare both sources and see if you fetch all selected information without discrepancies. Finally, you should carefully plan what data to distribute internally and externally. Not to sound bossy, but it is high time you drafted these reports. This way, you can catch nuances that might be impossible to capture post-launch.

Verify the basic framework: welcomes, goodbyes, small talks, and emojis

Unless you are a meteorological company, speaking about the weather might not be the core domain of your chatbot. But don’t cross it off that fast! Users may tend to challenge your assistant’s intelligence, especially at the beginning. You can experience swear words or popular ice-breaker topics occurring now and then. It is not too much of a strain to add a couple of general responses or surprise the interlocutor with a relevant joke. Make sure that your assistant has their opinion regarding Santa Claus's existence, handles both compliments and criticism, not to mention saying “welcome” or “goodbye”. Handling responses to emojis may also be a virtue of your assistant. It all adds up to its personality. Reacting to other languages is another asset. The benefits far outweigh the costs — the chatbot that socializes is more likely to gain long-term trust.

Maintain a healthy habit of saving versions and backups

We all know how instrumental versioning is. That said, during the hustle in the last week of a project, versions are often off the radar. Here’s your daily reminder to save your workspace before adding any last-minute changes. It will not only save you from potential mistakes but will also give you flexibility when it comes to MVP launch. The team can easily resign from a feature while having the freedom to add it at any future time. When everyone is ready for this very go-live moment, save (and ideally, download!) one more version, just in case. If there’s some unfortunate error, returning to your safe spot is a matter of seconds. After the launch, it is worthwhile to save a backup version ever and again, ensuring that your work is not forfeited.

Equip your LiveChat consultants with knowledge and reliable technology

If you enable human handoff within your solution, you should look out for certain aspects. First, make sure that the consultant team is onboarded and knows the ins and outs of the LiveChat software. If possible, conduct a training session with core functionalities of the console, including best practices regarding the handover. Double-check the agents’:

• roles,
• permissions,
• and department assignment.

During your final tests, you might have changed the standard business hours for your convenience. Now it is time to adjust them to standard 9–5 operations. You would not want your end-user to wait in the queue with no one available on the other side. If you intend to deploy your chatbot on new channels or extend the handoff in the future, a separate LiveChat instance for tests may prove invaluable. You are free to test novelties thoroughly while not disrupting consultants’ daily work.

The pre-launch road is full of ambushes, that’s for sure. Hopefully, we managed to shed some light on them. If you need any additional help, feel free to contact us at marketing@action.bot.

P.S: Thinking about implementing a chatbot for your channels, but still unsure what can it improve for your company? Check out our Chatbot Examples page and see what chatbots can do for you!