Websites today are jam-packed with content, actions, and tactics to get you to engage and transact. Is it working? Well, the only way to truly know is to track it, and there is a lot to track! Websites can easily average over 10,000 pages with multiple actions on each page. Tracking this amount of data requires complex implementations, massive data storage, and sophisticated data visualization to make sense of it all; and in the end, it may be impossible. Even with the help of web analytics software such as Omniture and Google Analytics, the amount of data is still dense and becomes overwhelming, sometimes counterintuitive, and reporting may seem unproductive.
So what is the answer? Engagement Scoring! Don’t look for answers in the data -- instead allow the data to provide you with the answer. Aggregating and calibrating large sums of data to build a score is nothing new. In fact, statistical scoring models have been around for decades, especially in the financial field. Banks use Credit Scoring to quickly assess the magnitude of risk by calibrating thousands of attributes (e.g. FICO score). So why not apply the same methodology to the wealth of website data?
The mechanics of a score are straightforward:
1. Define your website’s objectives, i.e. the actions or conversions which signal the successful use of your site.
2. Determine which site actions drive more or contribute less to your objective. These may be well established or you may need to build a separate model to isolate highest contributing drivers to your objective.
3. Create a score by giving more weight to most important actions, moderate weight to actions which contribute only slightly, and ignore anything that is not directly tied to your core objective but may be pertinent to your website.
4. Select your methodology for weighting. Weighting for a score can be defined in two ways:
a. Intuitive weighting: assumptions based on valued activities, using historical reference and business acumen; typically a result on insufficient data
b. Empirical weighting: values derived through statistical modeling, regressing each online activity against one/several business KPIs; requires adequate sample size and data for development
5. Establish benchmarks to gauge ongoing success or signal issues
6. Define cadence. How frequent and granular do you want your score to be? Data can be aggregated each time a visitor comes to your site or, more generally, data can be summarized on a daily, weekly or monthly basis.
The intent is to calibrate weights for only the key site actions, reduce manual review of large amounts of data, simplify reporting, and score your website regularly to monitor performance. Depending on your objective, a weight can be provided for just about any site action (e.g. repeat visit, content download, time spent on site, video view, registration, and purchase). A score consistently higher than your benchmark would indicate you are exceeding your site objective through the actions you deem most important. A spike or a drop in your score (larger than 2 standard deviations) should signal a detailed review; likewise an erratic score should draw attention to the calibration of the score since it could signal tracking, data collection, or paid media issues.
I believe there are many benefits to Engagement Scoring, especially for websites focused on brand objectives:
Of course, not all campaigns and websites warrant the need for a score, but for the ones that do -- happy scoring!