I was reading a post recently outlining a somewhat incomplete attempt to measure something called “Influence” as a measure of success in Twitter. Being a champion for complicated and easily misunderstood metrics based on cognitive and behavioral psychology I was immediately drawn to the article but walked away unsatisfied … that is, until I found Twinfluence.

Twinfluence is this nifty little Twitter tool that lets you explore a Twitterer’s “influence” based on their reach (size of their network and second-level network), velocity, social capital, and centralization (see the explanation page at Twinfluence for the details behind each.) For example, here are some of the people I follow in Twitter analyzed by Twinfluence rank:

• Rank #19: Jeremiah Owyang (jowyang) from Forrester Research
• Rank #660: Bryan Eisenberg (thegrok) from Future Now, Inc.
• Rank #2,893: Marshall Sponder (webmetricsguru) from Monster.com
• Rank #3,577: Avinash Kaushik (avinashkaushik) from Google Analytics
• Rank #6,124: Anil Batra (anilbatra) from ZeroDash1
• Rank #7,195: Aaron Gray (agray) from WebTrends
• Rank #7,591: Jim Sterne (jimsterne) from Emetrics
• Rank #11,209: Omniture (omniture) from, yep, Omniture
• Rank #11,786: Dennis Mortensen (dennismortensen) from Yahoo! Web Analytics
• Rank #11,940: Nick Arnett (nick_arnett) a social media blogger

Whee, what fun! I could Twinfluence my friends and folks I follow all night and day if only client work, my family, and copious powdery snow didn’t get in the way. In case you were interested I have a rank of #5,754 based on my nearly 700 followers who are followed by over 375,000 other people and a very resilient social network.

However, after a little while I started thinking that measuring someone’s “influence” in Twitter was the wrong way to think about success in social media in general. Especially since people who have been dubbed “influential” and successful in the blogosphere have a tendency to think about their popularity in somewhat ridiculous ways … say perhaps stating publicly that they’re going to charge to re-tweet content because they want to buy expensive stuff?

Anyway, when I went down this path I immediately thought “Hey, the two things I spend the most time on in Twitter is trying to find great people to follow and trying to share interesting ideas.” To find great people I use Tweetdeck and to a lesser extent MrTweet to find folks who are having a conversation I’m interested in. To share interesting ideas I limit the majority of my updates to the sharing of links on web analytics related topics.

These combined efforts have helped me find and share ideas with hundreds of folks in Twitter interested in web analytics. So I started thinking “So perhaps the true measure of success in Twitter is being as good a listener as you are a source of information!” Being a balanced participant in your efforts, not just a “social media rock star” who spends all their time talking at people, not to them …

Of course this line of thinking let me to Dave Donaldson’s Twitter Follower-Friend Ratio (or the Twitter Ratio for short.) The Twitter Ratio is dead simple: the number of followers you have divided by the number of people you follow — the perfect Twitter key performance indicator! Dave even provides benchmarks against which we can be measured:

• A ratio of less than 1.0 indicates that you are seeking knowledge (and Twitter Friends), but not getting much Twitter Love in return.
• A ratio of around 1.0 means you are respected among your peers. Either that or you follow your Mom and she follows you.
• A ratio of 2.0 or above shows that you are a popular person and people want to hear what you have to say. You might be a thought leader in your community.
• A ratio 10 or higher indicates that you’re either a Rock Star in your field or you are an elitist and you cannot be bothered by Twitter’s mindless chatter. You like to hear yourself talk. Luckily others like to hear you talk, too. You may be an ass.

(The emphasis on that last sentence is mine … I laughed out loud when I read that!)

I think Dave’s Twitter Ratio of 10 or higher is the same thing as Perry Belcher’s “Twitter Snob” (funny YouTube video if you have 5 minutes.)  Perry comments that if your Twitter ratio is super high you may not be participating in “social media” but rather “solo media” — perfect!  Perry’s point is why are you even in social media if you don’t have time to listen to the conversation?

If I apply the Twitter Ratio to all of the fine folks I analyzed still ranked using their Twinfluence score here is what we get:

• Jeremiah Owyang earns a score of 2.95 indicating that Jeremiah “may be a popular person” and “people want to hear what [Jeremiah] has to say” plus he “may be a thought leader in [his] community.” Sounds pretty much perfect to me, but I like Jeremiah.
• Bryan Eisenberg earns a score of 1.04 indicating that Bryan is “respected among [his] peers” (or that he follows his Mom and she follows him, but with 1,951 followers we can assume the former is the best explanation)
• Marshall Sponder earns a score of 2.30 which is pretty similar to Jeremiah’s score against his 851 followers.
• Avinash Kaushik earns a score of 105.5 indicating that Avinash is “either a Rock Star in [his] field or an elitist [who] cannot be bothered by Twitter’s mindless chatter” who “likes to hear [himself] talk” but “luckily others like to hear [him] talk too.”
• Anil Batra earns a score of 1.27 putting Anil in the same category with Bryan above although with only 266 followers his reach is somewhat lower than Bryan.
• Aaron Gray earns a score of 1.49 pushing Aaron more towards Jeremiah Owyang than Bryan Eisenberg, at least on Dave’s scale.
• Jim Sterne earns a score of 17.48 which is in the same “Rock Star” range as Avinash (although an order of magnitude less rock-starry  than Google’s own analytics evangelist)
• Omniture earns a score of 1.26 indicating respect among the company’s 247 followers
• Dennis Mortensen earns a score of 13.85 showing that Dennis, like Jim and Avniash, is a true web analytics rock star!
• Nick Arnett earns a score of 0.58 which indicates that Nick is trying but alas, “not getting much Twitter love in return.”

My own score is 3.13 against 697 followers which I’m pretty happy about (especially the part about not “being an ass!”) Incidentally Perry Belcher’s Twitter Ratio is 0.98 … about as balanced as it gets!  If you have 30 seconds you can go to Dave’s site and calculate your own Twitter Ratio.

What do you think?

Is “influence” the best measure of success in social media? Or should we pay closer attention to something like the Twitter Ratio as a measure of our likelihood to actively participate in the larger conversation? It’s not hard to imagine the Twitter Ratio combined with a measure of tenure or update velocity or even something like influence to come up with a system to help us better discover which members of Twitter are providing real and substantial value to the community.

I welcome your thoughts, comments, suggestions, and perhaps more selfishly, recommendations for great and interesting people to follow and tools to help with the discovery process.

One of the major complaints about my work on measures of Visitor and Audience Engagement is that unless you have Visual Site (= Omniture on Premise), Unica, Coremetrics, SAS, or a custom data warehouse solution you’re somewhat limited in your ability to make the calculation. Now, thanks to the recent upgrade to Google Analytics and the availability of session-level segmentation everyone can use my calculation to explore engagement patterns on their site.

Yep, free measures of Visitor Engagement from Web Analytics Demystified and Google Analytics!

It was a post from Alec Cochrane about engagement that got me thinking about the application of my calculation using Google’s segmentation features, thanks Alec! Heck, had I been paying more attention to his blog I would have noticed that even Avinask Kaushik (who persists in his dogmatic assertion that “engagement cannot be measured”) refers to GA’s ability to make the calculation.

Keep in mind, what I’m describing in this post is not a full-blown measure of Visitor Engagement for a lot of reasons. Still, as I’m kicking it around it appears to be a pretty good start and per my entire approach towards measures of engagement, I’d rather have all of you banging on the idea than work in a vacuum.

So how does it work?

Step 1: Gather Your Threshold Values

The first step is to determine what thresholds you want to set for your Click-Depth, Duration, Recency, and Loyalty indices.  You can get the first two from GA’s Visitors > Overview report (shown at right) while Recency and Loyalty come from Visitors > Visitor Loyalty > Recency and Visitors > Visitor Loyalty > Loyalty respectively.

Depending on your site you may need to be creative in how you set the Loyalty and Recency thresholds, especially since GA’s reporting on these measures is not super robust. Fortunately, since the segmentation tool is pretty flexible you can play with the threshold values once you’ve set them.

Step 2: Create Your Engaged Visitors Segment

The next step is to create a segment that lets you identify “engaged” visitors on your site. I’ll first describe the basic calculation, which is essentially the same as Audience Engagement only applied to click-stream data, and then expand in a follow-up post on the idea leveraging the Interaction Index, the Brand Index, and the Feedback Index.

Start by “creating a new custom segment” and adding the visitor dimension “Page Depth” (Google Analytics’s measure of Click-Depth during the session) setting the condition to “Greater than or equal to” the Click-Depth threshold value you discovered in Step 1:

Make sure to test the segment and confirm that things are working. In the example above you can see that about 25% of the sessions to my site last May were of at least three page views. Next you’re going to add the Duration Index by adding an “and” statement and dragging in the visitor dimension for “Visit Duration” and setting the condition to “Greater than or equal to” the time on site threshold determined in step 1:

Because you’re using an “and” statement we are getting the number of sessions that were both at least three page views and at least three minutes in duration; while this is imperfect compared to the visitor by session scoring strategy we described in the longer white paper the use of “and” ensures that we’re identifying visitors who are paying Attention as measured by clicks and session duration.

The next step is to roll in the Loyalty and Recency indices using the visitor dimensions “Count of Visits” and “Days Since Last Visit”.  As I mentioned above you may need to play with the thresholds here, perhaps creating a visitor segment of goal converters (purchases, leads, etc.) and examining the return visit behavior for that segment.  Also, when you set “Days Since Last Visit” be sure to use the condition “Less than or equal to” to capture visitors who have been to the site recently:

If your site is like mine you’ll see a noticeable drop in the number of matching visits when you add “Count of Visits” or “Days Since Last Visit” because of the use of the “and” operator.  But this is good and to be expected since if everyone coming to your site was truly engaged then you wouldn’t be reading this post, you would just be rolling in money.

All you have to do now is name and save the segment and you’re in business!  I called my segment “Engaged Visitors” which is not technically correct — really what I’m tracking is “Engaged Visits” — but when you see the final application of the segment below you’ll understand why.

Step 3: Mine Google Analytics for Engaged Visitors

Once you’ve created your “Engaged Visitors” segment you can start to apply it to the various reports in Google Analytics.  I recommend comparing the engagement segment against “All Visits” to get context — and GA does something nice here in calculating the percentage of segment members (= sessions where all four engagement criteria are met) for you.  Here you can see how this comparison looks in the Visitors > Map Overlay report:

While I’m only drawing a moderately engaged audience from Australia I am feeling the love from Spain! Probably since my good friend Rene Deschamps is Spanish or perhaps since I’m talking to a web analytics consulting group in Spain about coming over for a presentation and a big Web Analytics Wednesday event this coming Spring … who knows?

Now, I am pretty delighted with how easily these segments can be applied to the various reports in Google Analytics … hell, just the fact that the segment stays applied when I navigate from report to report is nice.  And yes, there are some obvious improvements that could be made but for a first effort this is pretty nice.

The same segment can be applied to reports that are more critical to how you run you business, for example the keyword report.  When I look at three top keywords driving traffic to my site you can see a clear pattern begin to emerge (and this is without adding the Brand Index into the engagement calculation):

Here you can see an obvious difference in the level of engagement associated with external searches for my brand’s name and “Web Analytics Wednesday.”  Even searches for Judah Phillips driving traffic to my site are bringing in a highly engaged audience (Judah, since I know you’re sensitive about this, nearly 30% of the visits associated with searches for you are scoring as engaged … nice work, buddy!)

If you’re willing to keep drilling down you can learn all kinds of wonderful things.  Here is a comparison of network traffic coming from WebTrends and Omniture:

Finally, if you’re using your one user defined field to capture some type of visitor identifier (hopefully doing so in line with your privacy policy) you can actually apply the engagement segment to individuals or groups interacting with your web site and actually begin to measure true Visitor Engagement.  Here you can see my very good friends Judah and June who are highly engaged at the Web Analytics Demystified web site, shown in stark contrast to another very active visitor who does not appear to be paying me any Attention at all:

This has become a long post so I’ll stop here for now and leave you with the following summary points:

• Google Analytics, like any session-based system, is not perfectly suited for calculating a true measure of Visitor Engagement;
• That said, given the recent availability of segmentation in Google Analytics, I would encourage those of you running GA to explore the use of my Visitor Engagement calculation;
• My belief is that you will begin to see for yourselves that this measure will help you identify opportunities not easily uncovered using traditional measures like average time on site and bounce rate;
• But you don’t have to take my word for it, do you? Play with the ideas I put forth in this post and let me know what you discover.  I would absolutely love to hear what you learn using this segmentation strategy or learn about applications of the segment that I haven’t thought of yet!

Last but not least, keep in mind that I have always put forth my work on Visitor and Audience Engagement as a hypothesis, one that is still being evolved and subject to testing and application in a variety of business situations. The thing I love about our community more than anything is the willingness that most of us show to explore new ideas and have an open mind.

As always I welcome your comments and feedback.

About six months ago the management team at comScore approached me with some questions about my Visitor Engagement calculation and the Web Analytics Demystified engagement framework. Their Chief Research Officer, Josh Chasin, had taken an interest in my work and wondered how it may be extensible across multiple properties using the comScore dataset.

It was an excellent question, and today I’m happy to give readers a preview of what we believe to be an excellent answer. Today we’re announcing a measure of Visitor Engagement that, thanks to comScore, can be used to compare levels of engagement across multiple properties in a similar category.

Brand Marketing’s New Measure: Audience Engagement

Audience Engagement is a simple modification of Web Analytics Demystified’s Visitor Engagement calculation that focuses on the core site behavioral attributes, measured through the comScore panel. If you remember, the Visitor Engagement calculation is:

Σ(Ci + Di + Ri + Li + Bi + Fi + Ii)

The components of the Visitor Engagement calculation are:

• Click Depth Index: Captures the contribution of page and event views
• Duration Index: Captures the contribution of time spent on site
• Recency Index: Captures the visitor’s “visit velocity”—the rate at which visitors return to the web site over time
• Brand Index: Captures the apparent awareness of the visitor of the brand, site, or product(s)
• Feedback Index: Captures qualitative information including propensity to solicit additional information or supply direct feedback
• Interaction Index: Captures visitor interaction with content or functionality designed to increase level of Attention the visitor is paying to the brand, site, or product(s)
• Loyalty Index: Captures the level of long-term interaction the visitor has with the brand, site, or product(s)

(More information about the measure of Visitor Engagement, including the details behind the calculation and several example use cases, can be obtained by reading the white paper that Joseph Carrabis and I recently published, Measuring the Immeasurable: Visitor Engagement which is freely available on this web site.)

The Audience Engagement simplifies Visitor Engagement by applying a “zero weighting” to the Brand, Feedback, and Interaction indices. By removing these values from the core calculation we are left with Click-Depth, Duration, Recency, and Loyalty:

Σ(Ci + Di + Ri + Li)

In English:

“Audience Engagement is a function of the number of clicks a visitor generates at a site, the amount of time they spent at the site, the frequency at which they return to the site, and their loyalty to the site as a member of the category for all of the sessions to that site during the reporting period.”

We’ve selected these four indices for one very simple reason: When scored using category-level thresholds (with the exception being the Loyalty Index, see below) comScore is able to automatically generate Audience Engagement values and engagement distributions across all of the sites they track.

The result is unique view into the relationship visitors have with the thousands of web sites comScore tracks around the globe. Now, for the first time ever, marketers and advertisers are able to gain insights into the level of engagement using a much more robust measure than session duration, page views, or recency alone.

Using Audience Engagement we can say with a high level of certainty that a greater percentage of Internet users find CNN more engaging than MSNBC and Yahoo! News:

More importantly we can also say that CNN has a larger population of “highly engaged” visitors to their site (22.5% of visitors at CNN versus 15% at MSNBC and less than 10% at Yahoo! News.) We believe that assessment of the audience distribution will provide advertisers an entirely new way to evaluate sites, focusing on audience quality over more simplistic measures of quantity.

This same type of analysis applied to popular network sports sites yields similarly interesting insights:

Here we can see that ESPN, while trailing Yahoo! Sports across all traditional measures (page views, sessions, minutes spent, active days) dominates Yahoo! from an Audience Engagement perspective. A closer examination of these two sites shows that ESPN’s dominance is driven largely by the frequency at which their audience members return to the site (Recency Index of 47.2% versus Yahoo! Sports at 27.0%) — an insight that has clear value to advertisers looking to create brand awareness and drive brand impressions across a sports-minded audience.

While comScore and Web Analytics Demystified are still working on how this data will be packaged and presented, another way of visualizing the relationship between two sites or a site and the category average is using a spider chart:

This chart visually tells the same story as the table above — ESPN has a higher level of Audience Engagement (bigger footprint) that is largely driven by Loyalty and Recency.

We believe that brand advertisers, advertising planners, and marketing managers will be able to use this data to make better decisions during the ad planning and media buying process. The whole debate over the definition of engagement manifest largely from advertisers desire to find more engaged audiences juxtaposed against a lack of faith in the simple measures being proposed as proxies for engagement. Thanks to comScore, these simple measures are about to become a thing of the past, giving way to a significantly more robust measure of the level of Attention audiences are paying at advertising powered sites around the world.

Interpreting Individual Data Points

In case you don’t want to spend the time reading the 50 page white paper I wrote recently on the subject with the mathematician and cultural anthropologist Joseph Carrabis, I’ll provide a brief summary of how the data comScore is reporting can be used.

Here is a sample of sites from comScore’s automotive category:

The first line in this table says that 42.8% of the audience to KBB.com is appreciably engaged with the web site. Engagement at KBB.com is largely driven by visitors clicking deeply into the site and spending an appreciable amount of time doing so, with nearly 85% of audience members exceeding the category Click Depth threshold and over 60% exceeding the duration threshold. Finally, using the distribution data, we can also see that 63% of the audience is highly engaged versus less than 3% who are only poorly engaged.

Audience Engagement data provided by comScore can also be used in a comparative context. Looking at the most and least engaging sites in this group, the data suggests that the audience going to KBB.com is over 400% more engaged than the audience going to About.com Autos (42.8% versus 8.5%.)  This is not to say that advertising at About.com Autos is a bad idea — over 90 percent of the site’s audience appears to be moderately engaged and in some instances a moderate level of engagement may be exactly what the campaign is looking for.

A Technical Note about Audience Engagement’s Loyalty Index

In the Audience Engagement calculation, the Loyalty Index is calculated differently than in the Visitor Engagement calculation because of an advantage conferred by the comScore system. Instead of simply counting the number of times a visitor has returned to the site as we’re forced to do using a site-centric data model, comScore allow us to better approximate loyalty as more commonly used: a measure of your likelihood to prefer a single site or brand over all others in the category. This model is essentially a “share of requirements” model used traditionally in the brand advertising industry and is calculated as:

Li(AE) = Visits to Site / Visits to All Sites in the Category

So, for example, if a comScore panelist is going only to eBay in comScore’s “Auctions” category, their Loyalty Index for eBay would be 100%:

Li(AE) = 10 visits to eBay / 10 visits in the “Auctions” Category

Conversely, if another visitor goes to eBay half the time and Bidz.com half the time, their Loyalty Index for eBay would be 50%:

Li(AE) = 5 visits to eBay / 10 visits in the “Auctions” Category

The result is a distribution of Loyalty Index scores for auction sites tracked by comScore in September that looks like this:

As you can see, eBay’s Audience Engagement component indices are higher than those of their competitors, but their Loyalty Index is much higher and tells us that nearly visitors in this category strongly prefer eBay to their competitors.

One of the challenges comScore and Web Analytics Demystified face regarding the Loyalty Index is the refinement of categories. Some categories like “Auctions” are well defined and represent logical competitors in a sector; others, like “News/Information” include diverse sites like Weather.com, Discovery.com, and Court TV Online. Over time we hope to refine these categories in partnership with comScore clients to provide the most accurate view of category loyalty possible. If you’re interested in participating in this work, please contact me directly.

Next Steps for comScore and Web Analytics Demystified

This is the first time we’ve been able to apply the Web Analytics Demystified Engagement construct to a syndicated audience data base.  We’re just announcing this work today, but we can already see possibilities for the measure’s evolution. Potential next-generation enhancements could include:

• Allowing comScore clients to provide a set of branded search terms to support the inclusion of Visitor Engagement’s Brand Index (Bi)
• Allowing comScore clients to provide a set of key site interactions designed to promote visitor Attention, supporting the inclusion of Visitor Engagement’s Interaction Index (Ii)
• Incorporating third-party data sources measuring more qualitative aspects of the audience relationship with the site, supporting the inclusion of Visitor Engagement’s Feedback Index (Fi)
• Allowing comScore clients to define their own competitive set in order to drill down into a more specific engagement profile in support of the advertising sales process
• Providing comScore clients access to the details behind the Audience Engagement calculation for their site and category
• Providing comScore clients custom access to Audience Engagement data, to provide a measure of Visitor Engagement in situations where the web analytic technology deployed does not support direct measurement

These are just a handful of examples of where this data offering can go. We’re presenting this model and starting the conversation because we want to hear from you. Regardless of whether you’re a current comScore or Web Analytics Demystified client, we would love your feedback regarding the calculation, the data, and the type of insights Audience Engagement is likely to provide to your organization.

Want to Know More about Audience Engagement?

Any reader of this blog knows that I have a passion for talking about the new measures of success on the Internet. I’m tremendously excited about this announcement and happy to talk if you’re interested in how you might be able to leverage Audience Engagement data.

Also, don’t hesitate to contact us if you have concerns about how we measure Audience Engagement or, in the extreme case, don’t think engagement can be measured at all. I firmly believe that the measures of Visitor and Audience Engagement I have proposed and the work I’ve done with Mr. Carrabis and now with comScore are only the beginning of the search for more useful measures of success on the Internet. Because these measures attempt to approximate something we agree is difficult to quantify, we believe that these measures will evolve over time; nothing is set in stone.

But we also believe that Visitor and Audience Engagement are better measures than “page views” and “average time spent” and far more useful to the measurement industry as a whole than simply sticking our head’s in the sand and exclaiming “engagement is an excuse” or worse, taking a Luddite’s view and declaring that complex measures are destined to fail.

For the time being, comScore is previewing additional details on the measure of Audience Engagement with their clients selectively.  If you’d like more information about how to be added to comScore’s list, or would like to discuss the measure of Audience Engagement with me, please email me directly and we can arrange a time to chat.

I have had a ton of great feedback about the white paper Joseph Carrabis and I wrote on Web Analytics Demystified’s measure of Visitor Engagement.  Some folks have raised very good questions and I wanted to provide some of the answers to those questions here to better socialize the knowledge.

The first question came from Jonny Longden who asked:

“I was wondering if you could clear something up for me regarding your visitor engagement white paper? It is to do with the way the equation is stated:

Σ(Ci + Di + Ri + Li + Bi + Fi + Ii)

Apologies if this is me being unintelligent (I am not a mathematician), but the way I read this is that the result of the equation is the sum of the 7 different index values. However, in your examples on page 32 the VE appears to the be the average of those 7 values. Am I missing something?”

An excellent question and one that several other people asked.  When I wrote the equation I was looking for the simplest possible way to represent the relationship between the indices.  I overdid that and Jonny’s question points that out.  Technically, what I should have written is something like this:

VISITOR ( SESSION ( SUM(Ci + Di + Ri + Li + Bi + Fi + Ii) / 7 ))

Indicating that for every visitor in the set, the sum of indicies based on each visitor’s session needs to be divided by the total number of indices, which in this equation is seven.  The reason I recommend dividing by seven is that the resulting number will be between 0.00 and 1.00 every time, yielding a nice, clean number that can be translated to a percentage for communication’s sake.

My next question came from Nikolay Gradinarov who asked:

“Have you considered weighing the different indexes that are part of the Visitor Engagement calculation?”

Yes, I have considered weighting the different indices, as does nearly everyone who looks at the calculation.  The reason I don’t apply any weighting is that I don’t have any way to know what weighting to apply.  Put another way, since I don’t have another measure of engagement, I don’t have any basis for using weighting to correct components of the equation; thusly, at least to me, applying differential weighting to any of the indices seems contrived and likely to increase the complexity of explaining the calculation more than anything.

That said, those of you using the calculation are free to apply whatever weighting you like.  For example, if you didn’t have a good way to calculate the Feeback Index and wanted to exclude it from the calculation of Visitor Engagement, you would simply “zero weight” that index.  Or, if the HIPPO said that “duration is at least three times as important as anything else as a measure of engagement” then you could multiply the Duration Index value by three.

Keep in mind, relative to the last question, doing so will change the mathematics.  If you’re three weighting one index then you’ll either need to divide by 9 (i.e., six “1 weighted” indices plus a three weighted index) to get a value between 0.00 and 1.00 or understand that for some sets you’ll have a number greater than 1.00.

The final question I wanted to treat here comes from Elizabeth Robillard who asked, and I paraphrase since she sent quite the document, whether it was better to calculate the Recency Index using all of the sessions in the set or just the two most recent sessions.  Elizabeth’s point was that when she applied my Recency Calculation using all sessions and the two most recent sessions to a variety of made up situations she was confused by what the data was telling her regarding engagement.

The short answer is that Elizabeth (and any of you) can use whatever sessions you’d like when making the Recency Index calculation.  Again, if you choose to base your calculation on only the two most recent sessions then you’re functionally “zero weighting” the other sessions in the set, which is another assumption but one you’re free to make.

But the main point I would make here is that Elizabeth appears to be trying to examine a single component index and make a statement about the level of Visitor Engagement.  Tempting, I know, but not how the Visitor Engagement calculation is designed to be used.  The reason I have spent so much time evangelizing for/thinking about/understanding the mathematics behind/debating/etc. is that I believe that all of the component indices are required to understand the nuances of Visitor Engagement.

To better understand why I believe this, re-read the section on “Why a New Measure of Engagement” on pages 10 to 15.  Trying to determine the level of engagement of a visitor by looking only at the Recency Index is just like trying to make the same determination using nothing more then raw or average session duration data — I do not believe that a single measure or metric has the resolving power to determine the level of Attention that a visitor is paying to your web site (or whatever object you’re trying to measure.)

So you could have a very low recency between the two most recent sessions, yielding a Recency Index of 100% based on Elizabeth’s suggestion, but a 0% score for Click-Depth, Duration, Loyalty, Interaction, Brand, and Feedback … which would show a visitor who has been to your site twice recently but is otherwise paying no Attention.  Similarly, you could have someone who hasn’t been to the site in the last 30 days, yielding a Recency score of around 3% (1/30) but who had high scores for Click-Depth, Duration, Interaction, Brand, and Feedback and thusly appears to be paying Attention.

You can play these scenarios out all day long — trust me, I have done this again and again — but at the end of the day in my humble opinion no one index, metric, or measurement will provide you the same level of insight that these seven indices combined provides on a visitor-by-visitor basis.

Hopefully the answers to these questions are helpful to other’s of you reading the white paper.  Again, you can download the document freely from my web site and as always I welcome your feedback.

A lot of you have been following the thread in my blog about measures of engagement on the Internet. Over the past year we have certainly had a spirited discussion about the topic, and for the most part people’s interest in the subject has not apparently subsided. About six months ago I started working with Mr. Joseph Carrabis from NextStage Global on the engagement calculation and the byproduct of our work is now available as a somewhat lengthy white paper on the subject freely available to all.

You can download the white paper from the Research > Published Research section of this web site:

http://www.webanalyticsdemystified.com/link_list.asp?l=Research

The white paper includes a great deal of information about the calculation including background on it’s derivation, the calculation itself, it’s use in a business context, and the underlying mathematics.  I welcome your feedback on the paper and am more than happy to discuss the contents via phone or email.

The direct measure of a Visitor’s Engagement with a web site or set of properties is still a work in progress to be sure.  And despite some naysayers, I believe that all of us working with this or similar calculations are quite excited about the possibilities associated with moving on from more simple measures and beginning to combine metrics to create a more interesting (and potentially more valuable) view of visitor interaction on the Internet.

Let the debate begin again!