Predicting User Questions – tcworld India 2016
I’m attending tcworld India 2016 in Bangalore. Mayur Bhandarkar gave a presentation entitled “Predicting User Questions to Build an Information Repository”. These are my notes from the session. All credit goes to Mayur, and any mistakes are my own.
I’m interested in this session especially because it advocates the use of FAQ, which is a document type often criticised in the technical writing community. One of the items in the presentation overview was “What the FAQ”. Funny and clever!
Mayur started by saying that presenting information in a structured way is a failure. Rather, queries are the path to success.
What are the advantages that FAQs offer? FAQs give the reader the impression that the document is going to answer a problem that the reader has. The content makes you feel that you’re having an interaction with the system. And an FAQ provides a complete solution for a particular problem. Another advantage is that FAQs seem more informal.
Look at LinkedIn for example: they provide documentation in the form of FAQs.
How can you predict users’ questions? Mayur gave an example of predicting questions for users of a dish washer, and then for a user interface.
How to go about it:
- List the features in the user interface.
- Decide the types of questions: why, what, and how.
- Build the question repository.
- Why should I use the x feature? (concept)
- What actions can be performed using the x feature? (reference)
- How do I use the x feature? (task)
Mayur related the above questions to the DITA types of concept, reference and task. Using the above examples, you can generate questions for each feature.
The next step is to organise the information so that it is useful to the user. Mayur showed us a system that prompted the user for information, including the language in which they want the information, and the type of information they want, and then showed a list of relevant questions. The user can select a question to see the answer. Mayur also discussed online sites that organise FAQs by most popular, most recently added, or modified features.
Users don’t like reading long procedures. When creating the FAQs, if a procedure ends up being very long, make the call whether to change the procedure into a video, simulation, or other easily consumable media. You could split the topic into smaller topics, but Mayur says that the users prefer media. An example is installation guides.
A trend that the team is following is the use of long-tailed keywords for SEO (search engine optimisation). A “keyword” can actually represent the whole task that the user is trying to perform. Using the FAQ format, an FAQ cis specifically related to such a keyword.
Someone from the audience asked about the problem of maintenance of videos. Mayur confirmed that these are maintenance heavy, and that the team produces videos based primarily on user demand, as represented by the support sites.
Another audience question was about basing your documentation on the UI, which is something technical writers don’t normally recommend. Mayur replied that this approach doesn’t work for troubleshooting or command-line applications, but only for simple UI-based documentation. It’s a solution for showing how to use the product.
An interesting point that Mayur mentioned: his team found that their PDF documents were being crawled more thoroughly by search engines than their online HML docs with the same content.
Thanks Mayur, this was an interesting perspective on FAQs.