The Pareto Principle, also known as 80/20 rule, is one of the most universal principles that exist. You can find in many aspects of your personal, social and working life. In its simplest formulation, the Pareto principle states that, for many events, roughly 80% of the effects come from 20% of the causes. This principle is named after Vilfredo Pareto who observed that income distribution typically followed this proportion (e.g. he noted that, in his time, 80% of the land in Italy was owned by about 20% of the population).
Chatbots are not an exception to the repeated occurrence of this kind of power-law distribution. Indeed, we can find plenty of applications of this 80/20 rule in the design and maintenance of (chat)bots. I’d like to highlight a couple of them so that you keep them in mind when creating your next chatbot (or your next bot development platform). Let’s start with the most generic and obvious one, applicable to any kind of business:
80% of chatbot revenue come from 20% of clients
So make sure you detect early on these “good” clients and evolve your business to cater to them as best as possible. But there are also other instantiations of the Pareto Principle more specific to the chatbots field. For instance:
80% of user utterances will match 20% of your intents.
Keeping this one in mind is important to fight one usual “disease” of our clients: trying to make a bot able to properly respond to any possible user request. First, because this is impossible, you’ll never be able to cover every single possible interaction scenario (unless, of course, you restrict the paths a user can take). But, secondly, because even if you could, you’d be wasting your time. Remember the well-known saying: “The first 90% of the code accounts for the first 90% of the development time. The remaining 10% of the code accounts for the other 90% of the development time.” – Tom Cargill
Instead of aiming for full coverage wasting your time on defining intents and training sets that a very teeny tiny percentage of users will ever match, focus on improving the quality of those that are frequently used. Note also, that, in fact, the quality of your bot may decrease if you add too many similar intents that may end up having overlapping training sets, confusing your NLU provider.
Recursive Pareto: 64% (80% of 80%) of your input utterances will be reduced to 4% (20% of 20%) of your defined intents.
This requires a good chatbot dashboard to help you identify these common intents and monitor the quality of their matches (e.g. to make sure you add/update their training sentences for optimal performance). Note that it could happen that a popular intent is one that in fact you don’t have yet!. Make sure you use your chatbot as a discovery tool to understand what your customers are looking for!. You can always add an automatic redirection to a live chat environment to catch those users before they leave and ask them for additional inputs.
The Pareto principle should also guide the features and platforms you add to the chatbot platform itself since
80% of your chatbots will require only 20% of your platforms
No need to keep adding more and more connections and platforms to your chatbot framework. Get to know your clients and you’ll quickly understand whether they typically use Slack or Facebook or WhatsApp …. to interact with their customers. Better than providing some support for every communication channel, make sure you provide great support for those your clients actually use. For instance, if you target more business/enterprise bots, focus on more professional communication tools (like Slack or Microsoft Teams) and forget about Whatsapp / Facebook.
Let Pareto guide you and you’ll be able to focus your effort on what will bring more results! (and let us know in the comments whether you use other applications of this principle in your life as a chatbot expert).