Better ways to measure content engagement than time metrics
I spent five years responsible for web analytics for a major ad-monetised content site, so I’m not immune to the unique challenges of measuring a “content consumption” website. Unlike an eCommerce site (where there is a more clear “conversion event”) content sites have to struggle with how to measure nebulous concepts like “engagement.” It can be tempting to just fall back on measures like “time on site”, but these metrics have significant drawbacks. This post outlines those, as well as proposing alternatives to better measure your content site.
So … what’s wrong with relying on time metrics?
1. Most business users don’t understand what they really mean
The majority of business users, and perhaps even newer analysts, may not understand the nuance of time calculations in the typical web analytics tool.
In short, time is calculated from subtracting two time stamps. For example:
Time on Page A = (Time Stamp of Page B) – (Time Stamp of Page A)
So time on page is calculated by subtracting what time you saw the next page from what time you saw the page in question. Time on site works similarly:
Time on Site = (Time Stamp of last call) – (Time Stamp of first call)
A call is often a page view, but could be any kind of call – an event, ecommerce transaction, etc.
Can you spot the issue here? What if a user doesn’t see a Page B, or only sends one call to your web analytics tool? In short: those users do not count in time calculations.
So why does that skew your data?
Let’s take a page, or website, with a 90% bounce rate. Time metrics are only based on 10% of traffic. Aka, time metrics are based on traffic that has already self-selected as “more interested”, by virtue of the fact that they didn’t bounce!
2. They are too heavily influenced by implementation and technical factors unrelated to user behaviour
The way your web analytics solution is implemented can have a significant impact on time metrics.
Consider these two implementations and sets of behaviour:
- I arrive on a website and click to expand a menu. This click is not tracked as event. I then leave.
- I arrive on a website and click to expand a menu. This click is tracked as an event. I then leave.
In the first example, I only sent one call to analytics. I therefore count as a “bounce”, and my time on the website does not count in “Time on Site”. In the second example, I have two calls to analytics, one for the page view and one for the event. I no longer count as a bounce, and my time on the website counts as “Time on Site.” My behaviour is the same, but the website’s time metrics are different.
You have to truly understand your implementation, and the impact of changes made to it, before you can use time metrics.
However, it’s not even just your site’s implementation that can affect time metrics. Tabbed browsing – default behaviour for most browsers these days – can skew time, since a user who keeps a tab open will keep “ticking” until the session times out in 30 mins.
Even the time of day your customers choose to browse can also impact time on site, as many web analytics tools end visits automatically at midnight. This isn’t a problem for all demographics, but perhaps the TechCrunches and the Mashables of the world see a bigger impact due to “night owls”!
3. They are misleading
It’s easy to erroneously determine ‘good’ and ‘bad’ based on time on site. However, I may spend a lot of time on a website because I’m really interested in the content, but I can also spend a lot of time on a website because the navigation is terrible and I can’t find what I need. There is nothing about a time metric that tells you if the time spent was successful, yet companies too often consider “more time” to indicate a successful visit. Consider a support site: a short time spent on site, where the user immediately got the help they needed and left, is an incredibly successful visit, but this wouldn’t be reflected by relying on time measures.
So what should you use instead?
Rather than relying on “passive” measures to understand engagement with your website, consider how you can measure engagement via “active” measures: aka, measuring the user’s actions instead of time passing.
Some examples of “active” measures on a content site:
- Content page views per visit. A lot of my concerns about regarding time measures also apply to “page views per visit” as a measure. (Did I consume lots of page views because I’m interested, or because I couldn’t find what I was looking for?) For a better “page views per visit” measure of engagement, track content page views, and calculate consumption of those per visit. This would therefore exclude navigational and more “administrative” pages and reflect actual content consumption. You can also track what percentage of your traffic actually sees a true content page, vs. just navigational pages.
- Ad revenue per visit. While this is less a measure of “engagement”, businesses do like to get paid, so this is definitely an important measure for most content sites! It can often be difficult to measure via your analytics tool, since you need to not only take in to account the page views, but what kind of ad the user saw, whether the space was sold or not and what the CPM was. However, it’s okay to use informed estimates. For example:Click-through rate to other articles. A lot of websites will include links to “related articles” or “you also might be interested in….” Track clicks to these links and measure click rate. This will tell you that users not only read an article, but were interested enough to click to read another.
- I saw 2 financial articles during my visit. We sell financial article pages at an average $10CPM and have an estimated 80% sell through rate. My visit is therefore worth 2/1000*$10*80% = 1.6 cents. This can be a much more helpful measure than “page views per visit” since not all page views are created equal. Having insight in to content consumed and its value can help drive decisions like what to promote or share.
- Number of shares or share rate. If sharing is considered important to your business, clearly highlight this call to action, and measure whether users share content, and what they share. Sharing is a much stronger indicator of engagement than simply viewing. (You won’t be able to track all shares, for example, copy-and-pasting URLs won’t be tracked, but tracking shares will still give you valuable information about content sharing trends.)
- Download rate. For example, downloading PDFs.
- Poll participation rate or other engaging activities.
- Video Play rate. Even better, track completion rate and drop-off points.
- Sign up and/or Follow on social.
- Account creation and sign in.
If you’re already doing a lot of the above, consider taking it a step further and calculating visit scores. For example, you may decide that each view of a content article is 1 point, a share is 5 points, a video start is 2 points and a video complete is 3 points. This allows you to calculate a total visit score, and analyse your traffic by “high” vs “low” scoring visitors. What sources bring high scoring visitors to the site? What content topics do they view more? This is more helpful than “1:32min time on site”!
By using these active measures of user behaviour, you will get better insight than through passive measures like time, which will enable better content optimisation and monetisation.
Is there anything else you would add to the list? What key measures do you use to understand content consumption and behaviour?