Introducing a better way to track page views

Hi, I’m Richard Loa, product owner of Analytics & Search at the CBC. My team focuses on collecting data that gauges how our digital product suite — cbc.ca and mobile apps for News, Sports, Radio, TV, Music for iOS and Android phones/tablets — is used by Canadians.

Photo By: Melissa Martinez; Photo of a book loop in a bookstore.

Have you ever wondered if you are tracking the right events on your website?

My team and I pondered this very question with respect to page views, the action of loading a web page, and came up with a solution that I’m excited to share with you.

Motivation for tracking the read verb

Imagine you’re at a bar with a friend, drinking caesars and enjoying the (too short) patio season. Your friend brings up an interesting news item they heard on the radio. The next morning, sober and curious, you want to learn more about the story.

You go to Google and search. Google returns a list of links and you click on the first item. After 5–10 seconds you realize it’s not what you were seeking. You hit the back button. You go back and forth between the list of content and story pages until you find what you’re looking for. You might repeat this three, four or five times before landing on the right page, and, in doing so, trigger three to five page views (i.e., false positives) for the websites you visited but didn’t actually contain the information you were looking for.

For those working behind the scenes on those websites, these false positives are captured on analytics platform, and, depending on what those sites are hoping to achieve, this may present a problem.

Clicks may help sell more ads but they don’t provide a good understanding of whether or not the requested content is relevant. Since optimizing for utility as a public service to Canadians is our team’s top priority, false positives throw off our understanding of how people are responding to the content we present on our websites and apps.

We used to say: “Woohoo! We are great! Look at all the people visiting our website! People are reading a ton of our content!” when in fact the page views were rejections of the content — false positives.

These false positives led me to wonder if there was a tool out there to detect if a person has read an article. Even if there was such a tool, would it account for:

• When the user switches tabs in a web browser?
• The occupied space of the article in the viewable screen?
• The amount of time the user had the article in view?
• The word count of the article?
• Average reading speed?
• How far the user scrolled down?

In short, the answer was none that I could find — so, I built it. It’s called Read JS. It’s open source — free to use, and completely flexible in how you want to define what “read” means.

Here is how CBC has chosen to define whether an article has been “read”*:

1. The article text is in the active web browser tab or window
2. The article text is occupying at least 15 per cent of the web browser’s visual space for at least X seconds (where X depends on the article word count)
3. At least the top 50 per cent of the article text has been displayed

*Note: The definition above is not always going to be this way — we are experimenting as we go!

Tying It All Together

To help ground the conversation about false positives, here’s a random seven-day period to consider:

Image of conversion funnel report. The first step of the funnel is loading a sports story and the second step of the funnel is reading the story.

From the data, we know any article tagged as “sports” from this time period was loaded by 392,809 devices. They were read approximately 20 per cent of the time and took an average of one minute and 42 seconds to read. Cool, eh?

This is just one of the ways we’re measuring if Canadians find the CBC’s content useful. If you have comments or constructive criticism, I’d love to hear it!

If this kind of work interests you, take a look at how we changed the way we work and an example of an A/B test that we ran. We’re also on the lookout for other people who can help us discover new insights from the data and analytics we work with. If you have a passion for this (or even just technology and public service), apply here. We look forward to hearing more about your ideas on what can be done with tools like Read JS!