How we built our knowledge base by analyzing our Intercom conversations
In the world of SaaS, you need to practice what you preach. As an AI SaaS provider, we preach the value of our product day-in, day-out. But can we find value in it ourselves?
We recently started building our help center, for a variety of reasons including:
1. Everyone else is doing it, so… it has to be the cool thing to do, right?
2. We’re great at supporting our customers (so they tell us), but really they expect to be able to use our solution without having to ask us for help all the time.
3. The more users find answers to their questions, the more they feel we understand them. Even if we are only explaining an unexpected error message. This in turns drives them into using our solution more.
4. Writing help content forces us to face the fact that, honestly, sometimes we just didn’t do the right thing. The harder it is to explain in a Help article, the more screwed up our UX approach was.
After trying out a few solutions, we opted for Intercom’s Article, which is a simple but really quick solution to build a customer-fronting Knowledge Base. We already use Intercom for our own support and sales engagements, so it made sense to keep using this platform for our Help Centre.
But whatever KB solution you pick, you still need to build your articles. And this is where the real fun begins.
Our first approach was to document every single feature of our great and brilliant SaaS solution. And as we have a lot of features, we managed to generate a lot of articles. 550+ articles about 3,627 AMAZING features and 17 cups of coffee later we had a comprehensive database of every single corner of our solution. Instant success, right? And we know our product and features well, so it didn’t take us too long to write these articles. And it didn’t take too long for our customers to read them either. Because they didn’t.
Our limited brains slowly came to the realisation that a Help Centre isn’t there to help you do a sales demo of your solution, it’s here to help customers, by answering their queries. So the question arose: what are they actually asking us? Most of my colleagues remember their last 40-50 intercom support conversations (personally I remember 8 or 9…). So we could have started making a list of what we remembered. Hoping we remember it correctly.
But we are not that dumb (honestly, once the caffeine kicks in we can actually pass as clever). Actually one of the many reasons our customers use our AI solution is because it is supposed to help them do exactly this: understand what their customers are contacting them about. I know this for sure because it’s written on our web page. And also because they tell us. We just didn’t think of this use case as being beneficial to ourselves, simply because a few years ago, we were so small I could actually remember all my support conversations.
So we went ahead. We plugged our own AI solution (Cx MOMENTS) into our Intercom instance and started analysing those customer support conversations. And it worked!
Within minutes, we discovered that customers were asking us about stuff we thought was already there or that we thought was the most obvious thing in the world. And sometimes it was stuff that had nothing to do with the app itself, but how to get a web demo from us, the difference between paid plans and free trial, error messages, data updates, and so on.
Connecting Intercom to Cx MOMENTS took just a few clicks (Pheww! this is what we promise on our website. So I can keep writing this article.)
Importing pre-defined topics into our Cx MOMENTS dashboard was actually good fun. Our Library feature enables users to decide, in a few clicks, which of our 250+ pre-defined/automated AI topics they want to bring into their dashboard. As we did this, we realised that some of the most frequent queries were about error messages. We didn’t invest much in generating eloquent and engaging error messages, and it definitely showed in our customer’s queries. Our customer-facing team’s first move was to ask our dev team for a list and description of these error messages, that seemed to be randomly appearing. We didn’t the
They got a short answer. Very short.
So they asked the question differently. They used Cx MOMENTS to list all the terms associated with “errors”, which gave them a good idea of what these specific messages were. Then the marketing team re-formulated their question to be much more actionable, referring to these specific messages, and supported by evidence from our customer conversations. The answer from Dev came back and was a much better one. We can now document these messages in our Help Centre… or get them corrected if they were not right.
This is the beauty of using customer feedback to write FAQs. They provide “facts” about what people are asking and volume trends that help prioritise them. The discussion between Cx teams and Product Development is then quite different. It becomes specific, actionable and evidence-based, and much, much easier to progress to a positive conclusion.
Another benefit of analysing this customer feedback is that it helps you formulate your FAQ in a way that your customers can actually find them. For example, customers don’t always ask to “reset their passwords”, sometimes they say “my account is blocked”, or “I entered my login details too many times”. By writing your FAQ Articles with the same words and formulation that you find in your customer’s questions, you increase the chances that they will find the right FAQ when searching your Help Center.
And this is what we did and will continue to do in Cx MOMENTS:
– Find the top 20-25 topics our customers are contacting us about
– List the sentences they use to formulate their questions
– Start writing Articles to answer these questions
– Discover and trend the next 25 topics and do the same.