onsdag 23. november 2016

Web analytics in service of accessibillity

A graph from google analytics.
Why is it so hard to track assistive technologies in web analytics?

Web analytics is a well known tool for better user experience. If we couldn't see how our users use our site, we would be quite lost in knowing how it works.

For a long time I have been playing with the idea of filtering users who use assistive technology to see how their behaviour is compared to those who don't. Maybe we even can detect barriers by doing this. But assistive technologies don't make this easy because they don't identify themselves to the server. I find it strange that web analytics can tell me see how big screens the users have or which versions of Windows they are running, but I can't know if they actually read or hear the content.

It's hard to believe that I am the first person to think about this, so I started googling web analytics and accessibility. What I found was lots of scepticism about identifying people disabilities. Many of the arguments are well funded, but many think they know my intentions before I have even opened my mouth. This article wraps it up quite well.

These are the arguments I find:

My disability is not your business

People don't want "dyslexia" written on their forehead. I can understand that. I don't want everyone to know every detail about me either. People also fear this information can be used for commercial purposes. I still see ads for bicycle glasses everywhere after I bought this several months ago, so I get their point. People want to protect their privacy.

But to be honest, I don't care about you. Not when I analyse web traffic. I care about you as a group, not as individuals. I want to find out how the group of users who use screen readers or high contrast mode handle our pages. If I discover that their conversion rate is lower than average, maybe there is an accessibility barrier somewhere.
We are living in the age of Edward Snowden. Even if I swear that my intentions are pure, others may misuse the same information. But how new is this really? If you are blind, I suspect that Google's and Facebook's algorithms already have figured this out.

I don't want to be treated differently

Some web sites have separate "accessible pages" in addition to the fancy pages for "normal" people. Nobody wants to be sent here. They want to play on main pages like the cool kids.

I don't like to separate the users either. I don't want twice as many pages to maintain. I'm a fan of universal design. One web to rule them all!

The numbers will be wrong anyway

When assistive technologies don't identify themselves to the servers, it gets difficult to track them in web analytics. It's possible with some creativity, but it takes different measures for different tools and some will possibly go under the radar. Besides, detecting a screen reader does not necessarily mean that there is a blind person on the client. I have tested plenty of sites with VoiceOver, but I'm definitely not blind. I have also seen my uncle holding a magnifier in front of the screen. How can we detect that? We will never get the numbers right.

But web analytics is not about exact numbers. It's about trends. We want to see if a change is for better or worse, how long the users stay on a page, how many pages they see per visit etc. And now I want to see if these data are different for users of assistive technologies. All I need is a representative group. It doesn't need to be complete, as long as the deviations in the data are noticeable. The real challenge is to be aware of what we are tracking.

I don't want to be counted

People are not always happy to be a part of some home made statistics of the users' abilities. When I advocate accessibility, I'm almost always asked how many blind users we have. I must admit that this was the first thing I was thinking when I got the idea of using web analytics for accessibility, but those numbers would not convince anyone anyway. The percentage of blind people is too low to make a noticeable difference for most sites. First when we add the numbers from all kinds of disabilities, it will make a difference. As mentioned before: web analytics is not about counting, but analysing users' behaviours.

I doubt many people find anything suspicious in my intentions. If they do, there is a problem with web analytics in general. The challenge is that the same identification also can be used for other, more dubious purposes. It's hard to make people count if they don't want to be counted.