Measuring Website Engagement September 29, 2009
Posted by admin in : Analytics, Buzz Monitoring, Social Marketing , 1 comment so farFrom conversations I have had with Marketers who have worked for an Internet property (where the term “stickiness” is used to suggest that the website is achieving high levels of engagement), some still fail to understand that measuring unique visitors and/or visits alone as a fairly unrepresentative exercise of a website’s performance.
With the continuing introduction of engagement tools for websites, many of which can be downloaded and installed for free, there is an increasing need to understand how a website engages its audience, using the most valid metrics as possible.
(Note here that we are measuring engagement with the website, as opposed to external marketing activity, which is the topic of a follow up blog on marketing measurement.)
Definition:
The concept of engagement is somewhat subjective and, as all marketing work does, needs to be rooted in the goals of the business/ website etc. However here engagement is generally considered;
“The depth and breadth of interaction visitors experience when visiting a website“
From experience working in the social media industry, measuring social engagement with individual pieces of web content, as well as the website as a whole, is important. This is due to some areas of a website needing to remain static and where engagement is not the goal e.g. A “services” page that details what a business does. As a result there are often cases where the pages that are need to remain static are removed from consideration.
So what are the best ways to measure engagement and which metrics are most appropriate?
The first is ironically unique visitors, which contrary to what might be suggested in the opening paragraph, is needed to understand total engagement that is representative of the website on average. YET this metric alone is, as previously stated, of no use and when measuring engagement simply provides the number for dividing the rest of the metrics being considered. This is because we need to assess engagement across all visitors that came to the website over any given period. As such it is not used on its own, but generalised engagement across the visitor population.
Basically:
As a simplified view of what engagement looks like we turn to session length. When session length is combined with the number of unique visitors you have a basic metric that is as close to measuring engagement as possible without needing a high level of website analytic assessment. The reason this works is that in general the aim of engaging people with web content is to keep them on the website long enough for them to want to perform a desired action, or visit a specific page. As such an understanding of how long people are staying on the website aggregates everything a visitor has done on the website as a single measure.
It must be pointed out however that this does not consider the quality of interaction. Nevertheless averaging session length over the total number of visitors to give an “average session length” is a base line understanding of how engaging a website is.
Neilson and Comscore are still working on creating a hybrid measurement for engagement. Today these industry standard audience data companies combine session length and unique visitors (as of 2007). This approach is likely to remain in force for a while, as engagement ultimately has many more variables that could be fed into it. Don’t be fooled however- this is not the whole story of engagement.
Advanced:
Web Analytics Demystified (http://blog.webanalyticsdemystified.com/weblog/2007/10/how-to-measure-visitor-engagement-redux.html) on the other hand has created a (complex) formula for measuring engagement and considers far more variables in order to aggregate the different variables that all contribute to an engagement metric.
It is excellent to see that individuals are working on ways to improve the reliability in web analytics, as all too often I have seen numbers being used that simply do not provide the backup to the story being told.
This formula is a good start, but highlights the problem with measuring engagement. There are exceptions to most of the variables used here and this will lead to a large degree of inaccuracy between websites. It must be remembered that engagement is dependent on what kind of engagement is the aim. For example there is a large difference between getting someone to submit a blog comment than visit more than 2 pages. This formula would also be too complex for the standard analytics user and would also be time consuming to solve.
A compromise needs to be reached that takes into account more variables, while still being usable.
We believe that the three most valid measures of engagement using current analytics are;
Time on Page
(this is an applied metric that looks at the depth of engagement on relevant pages)
Unique Visitors
(to be used as an average function and generalize engagement across all website visitors)
Page Views
(this measures breadth of engagement across the whole website)
Curved arrows show "Page Views" (3 are visited), while "Time on Page" assumes engagement with a page
It is always tempting to create some kind of compound number that “demonstrates engagement”, but realistically this is very challenging and would perhaps require an even more complex formula than above. To provide a relative balance the key is to set benchmarks for these metrics (which might be zero if a new metric is being employed), change the page’s functionality and then re-measure. Simply understanding a few key variables that are used in the right way to measure the right effects is a good strategy.
However as an attempt to establish how effective the pages of ‘engagement’ we have decided to bite the bullet and suggest the following;
(note: I am not a maths wizard and the following ideas suggest how the above metrics might be integrated)
“Time on Page” averaged across all pages of engagement
(Total time on website / unique visitors) = Average time on site
This will provide a number that is hopefully over 1. If not then the pages you have created to be engaging are actually less so than your static pages!
Next we establish the proportion of people who view an engagement page compared to those across whole website;
Average number of unique visitors to all engagement pages
Total unique visitors for the website
This shows us how effective the website is at driving people to engagement areas.
Combining the two through a multiplier should derive a theoretical number (not an actual numerical representation of anything, but something that should provide a virtual benchmark for how effective the areas of engaging are and to what degree they are visited.
The future?
A key addition to any engagement analytics is to look at the quality of the content engagement versus what interaction/ engagement tools are presented. For example how many actions were completed out of the total number that is presented and calls people to interact?
I might suggest that all a webpage’s engagement features are tracked via; action scripts e.g. “OnClick” command, confirmation pages, email submissions and tracking links to other pages. Each of these represents an action and would be used to assess the percentage of actions that occurred out of the total that was possible. Although this type of measurement is still not full proof, as a comment arguably requires more engagement from a visitor than it does to rate something. Despite this, measuring actual, active behaviour represents engagement far better than the “Time on Page” metric used today. It attempts to understand how the user interacts with the stimuli presented (See main image).
It is only really when analytics can start to measure this kind of data that effectively measuring engagement will be possible.
