Which users actions are worth logging, to gain better perspective on their behavior?

Question

I am currently implementing a "User action log", that will keep interesting actions of users within our system.

Any advice on what to log? Of course, the answer is "everything you feasibly can", but I want to focus on the important actions, that will help us gain true insights.

Log every button click? Every hover on a button? Same for hrefs? What else?

Note:

1. The goal is better UX - we want to study our users, understand how they behave, and build a better system for them.
2. I'm aware of some 3rd party solutions that track everything users do, starting from Google Analytics, through apps like Clicktale. I'm not saying we won't use those - but still, a custom log with some select actions might provide extra value over these "log everything" type of systems.
3. The focus is websites, built for consumption over standard computers (mobile/tablet clients are less relevant for us a.t.m).

Answer 1

No no no! The answer is not "Everything you feasibly can"!

You really need to turn the situation around. There is no need to log user actions just to be sure. You need to know what you want to know more about. What is the purpose of this data collection?

1. Start out by defining some goals. What do you want to know?

2. Then you need to figure out what to look for to answer those questions.

3. Finally you need to pick some key indicators you can log to feed your metrics/measures.

   (Take a look at the GQM method. This is a very handy method when you need to figure out what to measure/log.)

BTW, the log information does not (necessarily) provide you with valuable information alone. I don't know how many times I've accidentally hit the F1 button when I really aimed for the ESC button. What does these data tell you. Say you have 95% of the users hitting the help-button once or twice during a session. Did they intend to get help? Did they accidentally hit the help-button? Is the help button helpful to these users or is the form/page itself confusing?

What you need to do is to observe users in action. This is the best way to improve the UX of your product.

Answer 2

If you're asking yourself this question, then to be honest, I'd suggest not making this kind of decision up front at all, but instead incorporating an adaptive logging solution which you can control from the server side what is logged (and when).

Since you don't know what it is you want to log, then you need to log everything so that you can find out what it is that you want to carry on logging and more importantly find out what it is that you don't want to carry on logging.

Otherwise you run the risk of not logging something that you didn't realise might be of interest. It's all very easy to have an idea of what might be of interest and log that, but really that's just going to tell you something you kind of thought you knew already, where as the point is that you don't know what you don't know!

I don't mean to be awkward, but if you can't answer the question given your own domain knowledge, it strikes me that you certainly can't expect someone who knows nothing about your system to answer it for you.

Having said that, it's also worth saying that in the same way that whitespace is an important part of content and gaps are an important part of speech, you may consider it useful to get an idea of periods of inactivity between actions. Obviously you can't assign a single user's inactivity to a particular problem - I'm talking about when you have significant enough data that you can form a trend or pattern.

Answer 3

Start by deciding on what you want to measure and what you can do with these metrics. What sort of improvements and design decisions can be made with this info? E.g. Are you trying to find out how frequently users access the "Help" link

Think about the additional data around clicks that will provide context e.g. sequence, frequency, interval...

Before collecting data, I find it helpful to think about how I would report the findings and the purpose and audience for this report.