Visit scoring and engagement value are two metrics that can be used to assess the validity of a website visit. Understanding both the benefits and the limitations of these two metrics is important.
The Power of Engagement Value
Not all visits to a website are of equal value; marketers know this. Often they use conversion metrics to assess the value of visitors to the site delivered by a specific campaign, but this is a very all or nothing metric – it allocates no value to a visit during which a visitor progressed significantly through the sales cycle but did not convert.
Sitecore DMS (and other marketing-aware CMS products) provide marketers with an alternative – score the value of a visit based on the content viewed and the actions taken. Visits that include high “engagement” activities such as requesting a whitepaper or signing up for a newsletter might be assigned higher scores than a visit that consists primarily of perusing job openings (assuming that sales trumps recruitment for the organization).
With an engagement model in place, marketers can refine campaign expenditures based on the value of the delivered traffic. Problem solved, right?
The problem with engagement models? Building them
Well, sort of. The big challenge with engagement models is building the model in the first place. How many “engagement points” should I give to any given action? Does viewing content count or must the visitor perform an action? The options commonly proposed for creating this model are “top down.” They require the marketer to invent a model and then apply it to the content. Specific approaches recommended include:
- The investment model:
Score visits that require action or investment by the visitor more highly than those that require passive viewing. The underlying concept is that individuals who “invest” are more valuable than those that do not.
- The real world model: Talk to your sales team and probe them to find out which activities, questions or content requests indicate a prospect’s value. Use this to guide your engagement model values.
Both of these seem reasonable approaches, but in my work with clients, I've found that neither is particularly useful. Here's why:
- The investment model is flawed because it assumes that information gathering is less important than action. For organizations that have a long sales cycle or whose clients move through a considered purchase process, this simply isn't true. Engagement models built on this approach effectively assign value to visitors late in the sales process, ignoring the value of the early sales funnel.
- The real world model is flawed in two ways:
- Most visits to your website happen long before your sales team gets involved in the deal. The content these early stage visitors need is not what your sales team will tell you is of value.
- A minimal sample size – even your best sales person simply does not see enough prospects in enough geographies to give you reliable information.
An engagement model based on this approach is doomed to visits in the early stages of the sales cycle and is more likely to reflect the idiosyncratic beliefs of a handful of sales people than any underlying truth.
The truth is in there – the Big Data alternative
There is an alternative approach. It’s simple in concept, but challenging in execution:
Let the data tell you the value of any given piece of content or action on your site.
Products like the Sitecore DMS can be configured to retain a profile of visitors over time – each page visited, action taken, search term used, etc. It can also record the eventual outcome associated with that visitor:
- Purchased a product
- Became a lead
- Applied for a job
- Never returned to site, etc.
Machine learning techniques can tell us explicitly how much value any given piece of content or action taken has in predicting the eventual outcomes we desire. This is very much a Big Data problem. Tracking every aspect of every visit generates huge volumes of data, though most of it won't apply to the problem at hand, you cannot know until you try.
Academics have been working on this kind of problem since the advent of the commercial internet – and they have had success in creating one off models for predicting everything from the final price of a product auctioned on EBay to the lifetime value of a customer based on actions and clickstream.
Making it real – a plea to marketing technology vendors
We could build individual models for each of our clients, but this is both expensive and time consuming. So the challenge we’re issuing to marketing technology vendors is to develop these models and embed them in their infrastructure solutions. In an ideal world a product like Sitecore DMS would:
- Build visitor profiles that include content viewed and actions taken
- Allow marketers to assign values to goals or outcomes
- Automatically run baysian networks or other machine learning tools to determine the actual value of content on the site or actions taken by a visitor
- Apply these engagement values to every page on the site automatically
- Provide marketers with an easy way to view and adjust (if required) the value applied to any given piece of content or action
We don’t know of any product that provides this today; this is a critical gap if concepts like engagement value are to become reliable tools of the marketing toolkit.
A stand-in alternative
In a subsequent post, I will discuss how we here at nonlinear digital use Google Analytics data to generate a model for the value of any given piece of content. The Google Analytics approach has significant limitations – including an inability to look cross-visit when determining value – but it offers a real, data-based way for creating a model of content value.