Analytics Expert

After much debate, my team of co-presenters and I elected to cover a small number of topics, but to dive as deeply as possible into those topics during the brief time allotted to us. We tried to pick tactics that were brand new, little-known, or known-but-underutilized. I discussed the first two of these topics, Visitor Scoring and getPercentPageViewed, in previous posts. The final concept, participation, is something that may be a bit more familiar to some, but we believe that more could/should be taking advantage of this feature to understand where their truly valuable content is.

Participation—an overview

How do you gauge the value of your content? It’s easy to see what someone was doing when they added a product to a shopping cart, or when they clicked to subscribe to your RSS feed, but it may not be terribly actionable in many cases. Consider a newspaper site, where one measure of success is daily subscriptions generated. Just because users happened to be browsing the Politics section when they clicked to begin the subscription process doesn’t necessarily mean that Politics was the sole impetus leading to subscription. Previous page views across the site—Sports, Local, Entertainment, etc.—may have played an equally valuable role. And when you example these trends in aggregate, wonderful pictures of relative content value begin to emerge. Participation is how you get there.

I’ve been relying heavily on Adam Greco’s fantastic writeups for background info during this series of posts, and sure enough, he covered participation back in 2008. The key point here is that, while linear allocation in SiteCatalyst assigns partial credit to each page or other report value that led up to a conversion, participation assigns full credit to each of these values. I’ll explain below. (This can be a bit tricky to grasp, so hopefully readers will let me know if it ends up as clear as mud.)

In the Pages report, linear allocation means that if the user viewed five pages and converted—for example, placed an order for $100—on the fifth page, the standard Orders and Revenue metrics would assign one-tenth of an order and $20 of revenue, respectively, to each of the five pages leading up to the purchase.

Linear allocation can be really useful in many cases. However, as Adam pointed out, “[linear] allocation tends to reward visits with smaller number of clicks and punish visits with many clicks.  Depending upon your business model and/or site objectives, this can be a positive or a negative.” Is a page less “valuable” simply because there were more page views during the visit prior to conversion? Or, conversely, is a site section more valuable because there were fewer site sections during the visit? A $10,000 purchase divided by 20 pages assigns just $500 to each page in the visit, but if a user spent the same $10,000 after just five pages, then those pages receive credit for $20,000 instead. The order and revenue values were the same, so why does each page in the 20-page visit receive less credit? At the very least, in some situations it can introduce inconsistencies in understanding the value of the content involved in the conversion.

This is where participation comes in. It assigns full credit for the given conversion event to every value (every page name, every site section, etc.) that led up to the conversion. In the example above, each page in the 20-page visit would receive one order and $10,000 of revenue. Each page in the five-page visit would receive one order as well, and $10,000 of revenue. It levels the playing field, so that report line items, such as page names, do not display added or diluted value simply because the visit was shorter or longer.

This diagram assumes a five-page visit leading to a $100 purchase.

The really great news with participation is that you don’t need to implement anything (provided you’re already capturing your most critical success metrics already). Participation is calculated by Omniture based on visit path and conversion data that you capture using normal techniques, and is available on request in SiteCatalyst for your top conversion events; your Account Manager can help you decide which metrics should use participation and can enable this feature for you. Just to be clear, I’ll put that in other words (and in bold): contact your Omniture Account Manager to get participation enabled if you believe this feature is for you. Note that enabling participation adds new metrics to the relevant reports; it does not replace any existing metrics or data, so you don’t need to worry about the effect on dashboards or bookmarked reports from enabling this feature.

Analysis and optimization using participation

As Josh James described during the opening general session at Summit, the “old school” metrics aren’t going to cut it in the next digital decade. They don’t tell the whole story, and in some cases can be really misleading. An online marketing organization that is accountable for its decisions needs more. Take a look at the report below. If you were judging the relative value of these content types based on page views or visitors, you might make some bad choices about where to focus your efforts. When we look at “subscription participation” (i.e., the various “content type” values that participated in generating subscriptions), we see a different picture of things.

In reality, conversion is being generated by the “column,” “interactive,” and “video” content types. (”Content Type” here is a Custom Traffic variable. The example could just as easily have focused on pages, site sections, etc.) Now you know that you should examine what these content types are doing successfully to point users toward conversion. In Omniture Discover, which offers participation for all success events by default, you can even run these reports against various segments to see how content value changes based on nearly any criteria under the sun.

Page View Event Participation

Another powerful thing you can do with participation and a page view custom event is “Page View Participation per Visit,” which helps you understand the “stickiness” of various pages on your site by show, in metric form, the average number of page views per visit that follow each page on your site (For more information on setting up a page view custom event this post. The basic idea is that you will have a custom event that gets set on every page view across your entire site.)

Once participation is enabled on your Page View custom event, each page that users view receive full credit for each page view that comes after it. In the diagram below, notice that page A has a value of four, indicating that a total of four page views came as a result of viewing page A. Page B has a value of three, because three page views can ultimately be attributed to it.

Within the Pages report, you’ll want to set up a calculated metric defined as:

[Page View Event Participation] / [Visits]

This would level the playing field even further. Once you’ve got that metric, you can easily see which pages are stickiest right there in your Pages report:

Those four pages in the middle of the report are highly sticky without being the raw traffic leaders. Of course, in this particular report you may need to keep in mind what type of “content” you’re looking at; it’s possible that the “site search: results” page might be somewhat overvalued because it is naturally likely to lead to additional page views. But for certain content types, this information could be extremely helpful.

Conclusion

That should just about do it: The three topics that we covered at Summit, now in blog form. Our goal during our presentation was for audience members to come away with at least one thing that they could go home and do immediately (although, of course, if they came away with more than one thing, that’s even better). Hopefully this series of blog posts will enable even those who didn’t make it to Summit—and those who were there but wanted more information—to begin employing at least one of these tactics to do more in SiteCatalyst. Probably more importantly, I hope that the core concepts and functionality are clear so that you can apply these tactics in your own unique way to address your specific business needs.

As always, I welcome any questions, concerns, comments, etc. that you might have about any of these posts (or about anything else related to the Omniture Online Marketing Suite. Please feel free to comment on this or any other blog post, or to contact me via Twitter (@OmnitureCare) and I’ll do my best to get you the information that you need.

A few lucky SiteCatalyst users got a sneak preview showing of our Summit session, and the nearly unanimous reaction was that one of our three advanced tactics really stood out as the most exciting: the getPercentPageViewed plug-in, which (as the name suggests) allows SiteCatalyst to capture the percentage of a page (vertically) that the user has viewed. Best of all, the plug-in is completely free and available to all Omniture users. You can download it, along with basic documentation and installation instructions, from within SiteCatalyst by going to Help > Home, and then choosing Supporting Docs > Implementation > Plug-ins from the left navigation menu.

getPercentPageViewed—an overview

Omniture already offers a number of tools to help you understand what users are doing on your site. As far as page design and layout analysis are concerned, you’re probably familiar with ClickMap, which applies a graphical overlay to links and buttons to show you which page elements your visitors are clicking (and how often they’re clicking them). But this is just one gauge of page consumption. There are plenty of scenarios where it would be great to know not just what the user clicked, but also what the user saw. How much real estate do you really have at your disposal? Are users finding the page elements that you want them to find? This is exactly what the getPercentPageViewed plug-in does. It calculates the length of the page (even adjusting if users resize their browser) and then determines how far down visitors go before clicking away. This value updates as the user scrolls, and will always retain the maximum percentage viewed—even if the user later scrolls back up to the top.

To implement the plug-in, you will need the following:

1. Two Custom Traffic (s.prop) variables.
2. A correlation between the two variables.

That’s it. You can do more with additional s.prop variables and correlations, but we’ll start out with just these two items.

NOTE: The plug-in can (obviously) only determine how much of the page the user has viewed after the page has loaded, and only as the user scrolls. This means that the only reliable way to capture this data for a given page is on the next page view. Thus, we’re really capturing “percentage of previous page viewed,” but we can still marry this to the correct page name for analysis. I’ll explain how to do it below.

First Custom Traffic variable: Percentage of Page Viewed

To the developer’s enormous credit, the plug-in is extremely easy to implement. Instructions are available in the Knowledge Base, but essentially you drop the bulk of the code into your plug-ins section where you probably already have other code snippets such as getQueryParam and getValOnce. Then, within the s_doPlugins() function, you make a call to the plug-in as shown below (using s.prop1 as our Custom Traffic variable):

// capture the percentage of previous page viewed
s.prop1=s.getPercentPageViewed();

Simple. However, this doesn’t really give us anything actionable on its own. Just a bunch of numbers representing the raw percentages viewed for all pages on your site:

The really useful data comes when we also capture the previous page name (or previous site section, previous search term, etc.) into a separate Custom Traffic variable and enable a correlation between these two variables.

Second Custom Traffic variable: Previous Page Name

As described above, the second Custom Traffic variable captures the previous page name so that we can break down page name by percentage of page viewed, since that is where the magic happens. Fortunately, there is another plug-in (available in the same location as getPercentPageViewed) called getPreviousValue which makes capturing the previous page name easy. Installation of getPreviousValue is very similar to installation of getPercentPageViewed, and you would call the plug-in as shown here (using s.prop2 to capture the previous page name):

// pass the previous page name into prop2
s.prop2=s.getPreviousValue(s.pageName,'gpv_pn')

That’s it. So, finally, combining the two into a conditional, here’s how we recommend implementing these two plug-ins:

// capture previous page name; if it exists, capture percent of page viewed
s.prop2=s.getPreviousValue(s.pageName,'gpv_pn');
if (s.prop2){
s.prop1=s.getPercentPageViewed();
}

At this point, you’ll have a Custom Traffic report in SiteCatalyst with a bunch of raw percentages, and another Custom Traffic report that looks very similar to your Pages report, except that it shows the previous page name for each page view.

Correlation

Adam Greco wrote about data correlations in a previous blog post, and I highly recommend reading his advice if you’re unclear on how correlations work or on how to set them up. In this case, you will want to enable a correlation between the two Custom Traffic reports that correspond to the variables that you’re using to grab the percentage of page viewed and the previous page name. Once the correlation has been set up, you can go to the Previous Page Name report and click the green correlation icon to break down any page name by the Percentage of Page Viewed report, giving you something like this if you chose to break down the “Home” page name:

This is starting to become useful. We can see that our home page typically gets users to view around 50% of the page. If we have important links, ads, videos, etc., we probably should put them in the top half of the page. Or maybe the page is altogether too long and could be shortened.

Another useful way to look at this data involves grouping the various percentages in the report using SAINT. Looking at the raw percentages (0-100) is a good start, but bucketing can help bring to the forefront trends that may be harder to see by viewing the percentages as-is. And because you can add multiple classifications to any report, you can actually slice and dice your percentages in a variety of ways: quartiles, quintiles, deciles, etc. This gives you easily digestible reports, such as this one:

Where this gets even cooler is in the context of A/B testing. You can send various page layout names into a Custom Traffic report (using getPreviousValue, of course) and correlate with percentage of page viewed to see how the different options engage users differently. You’ll already know what they clicked, thanks to ClickMap or other reports, but the getPercentPageViewed plug-in can help you see how different variants affect page consumption.

You can do some great things with these reports.

• On the surface of it, you can let your team know exactly how much “real estate” they realistically have to work with on key pages.
• Key features that you want to highlight can be moved into the appropriate locations where you know most users will see them.
• Where appropriate, pages can be shortened to lengths more suitable to actual user behavior.
• Calendar events can help show how site/page changes affected consumption.
• Comparing to ClickMap or custom link data, you can understand the relationships between viewing a certain portion of the page and clicking on links within that portion of the page.

A few final questions (and answers)

Q: Should I use a Custom Conversion (eVar) variable as well (or instead)?

A: My feeling is generally that a Custom Traffic approach is best. Because the percentage of page viewed is going to be captured on every page, there isn’t much value to the persistence offered by eVars. The only use case that I can imagine for eVars in this scenario is the ability to understand how page consumption on the entry page affects conversion (which could be accomplished by using the plug-in to populate an eVar set to “Original Value” so that the entry page receives credit for the conversion). For the majority of users, the real value here is simply in understanding how page layouts and lengths are working, which can be accomplished using a Custom Traffic variable.

Q: What happens if users abandon my site? Do we still get to see how much of the page they viewed?

A: Because the plug-in captures the percentage of the previous page that the user viewed, if users close their web browsers, click a browser bookmark, or enter a different URL in their address bar, SiteCatalyst will not receive the percentage for that final page (as there is no “next page” on which to capture the data). However, if the user clicks an exit link, the percentage of the page that the user viewed will be captured prior to the user leaving your site.

Q: What if users resize their browser horizontally?

A: It will have no effect on the plug-in’s ability to capture the percentage vertically. If resizing the browser changes the maximum length of the page, the plug-in will adjust to it.

We have already seen SiteCatalyst users do some great analysis and optimization based on this functionality, so hopefully you’re as excited about this new plug-in as I am. Next time, we’ll cover the final topic from my Summit session: participation. (Adam already described this feature, but we have a few things to add.) Of course, if you have any questions about getPercentPageViewed, either based on this blog post or on the content we shared at Omniture Summit (or if you asked a question at Summit and we didn’t get a chance to answer it!), please let me know by leaving a comment here. You can also contact me via Twitter (@OmnitureCare).

Whew! Omniture Summit 2010 has come and gone, but hopefully you’re still deriving real value from some of the thoughts, tactics, and strategies that were discussed, both formally and informally, at the conference. As promised, I will be blogging about each of the three advanced SiteCatalyst solutions that my team and I covered in our breakout session. (It’s also worth noting that, while we discussed SiteCatalyst specifically in my session, these ideas become even more powerful when coupled with an advanced segmentation tool, such as Discover or ASI, or when used in the context of Test & Target.) Let’s get started with the first topic from our session.

Visitor Scoring—an overview

You probably have a number key activities on your site that constitute a highly engaged visitor. These are things that you want users to do, because they either constitute direct success (i.e., conversion), indicate the building of loyalty (e.g., social elements such as “Write a Review”), or suggest that the user is truly interacting with your site (e.g., internal searches, viewing photos, etc.). Visitor scoring in SiteCatalyst allows you to assign numerical “scores” to each of these key activities, and then aggregate these scores as visitors move through these different site elements.

The idea is that you can begin to see how varying visitor (and visit) scores affect conversion and customer loyalty. For example, you may notice that users with a score of 30 or higher spend twice as much money on each order when compared to users with score less than 30. You can also drill down to see which site elements are preferred by highly engaged visitors so that you can focus your efforts on those areas. Additionally, the solution allows you to see which visitor acquisition channels generate the highest levels of interaction with key site elements, and which campaigns are the stickiest in terms of leading users to these aspects of your site.

To do this, you will need three things:

1. A scoring system that assigns values to each of your top 5-10 key engagement points.
2. One “counter eVar,” which will be explained below.
3. One custom event, set to “numeric” as its type.

The Scoring System

This is the most complex aspect of the Visitor Scoring solution because you will need to lay out a relative points system for the 5-10 top site elements that you want users to engage. Some users may have access to advanced systems for scientifically assigning relative value to different site activities, but in the absence of such a system here is how we recommend doing this:

1. Identify the top several key “things” that you want users to do on your site (including, but not necessarily limited to, conversion). As described above, these should be activities that show engagement; the idea is to gauge how active a visitor’s experience is.
2. Isolate either the activity that you consider most valuable or the activity that you consider least valuable on your list.
3. Assign that value an arbitrary score, such as “5.” (This can be anything, and can be adjusted later.)
4. Assign other scores relative to that first score. For example, if we believe that “internal search” is our important, but is our least “valuable” activity of those on our list, we might give it a value of 4. Let’s say that subscribing to a newsletter is significantly more valuable; we might give it a value of 8. Reading a review may be even less valuable in our minds than performing a search; we’ll give it a value of 3.

This chart shows an example of how this system might be laid out:

This will be entirely unique to your business and, at the beginning, may involve a bit of educated estimation. Fortunately, as you work more closely with these reports, you’ll be able to see where scores are being inflated (or deflated) by user activities that have been overvalued or undervalued, and you can adjust your scoring appropriately.

The Counter eVar

A staple of almost any SiteCatalyst implementation is the “eVar” variable, which allows you to set a value (e.g., an internal search keyword) and persist it for a customizable period of time so that you can tie subsequent success metrics back to the values. (In the convenient “internal search keyword” example, an eVar allows you to view the amount of revenue, number of leads, etc. that occurred after the given keyword was searched.) These eVar variables accept text strings by default, but also have a nifty option for accepting numbers. Adam Greco defined counter eVars in outstanding detail in a previous blog post, which contains information on enabling counter eVars as well as a number of use cases.

In the case of Visitor Scoring, we’re going to take advantage of eVars’ persistence by setting the number of points for the key activities that were defined in the previous step into the eVar each time the activity occurs. For example, based on the chart above, we would set a value of “+3″ into the counter eVar immediately after the user rates a product, “+5″ every time the user e-mails an item to a friend, and so forth. Implementation-wise, using eVar1 as an example, it looks like this:

// after an internal search
s.eVar1="+4"

// after writing a review
s.eVar1="+5"

(Not that this has anything to do with Visitor Scoring, but it’s worth mentioning here—because we forgot to do it in one of our two chances to present this topic—that you CAN pass negative numbers into counter eVars to subtract from the overall visitor value.)

As the user moves through an experience on your site, interacting with the various key elements that you have defined, he/she will accrue points. If the user converts, you’ll be able to see their “score” at the time of conversion. If they don’t convert, you’ll be able to see that, too.

Pretty confusing, right? We can see that scores of 16-17 appears to be the “sweet spot,” but it’s hard to really understand what we’re looking at here. Fortunately, you can use SAINT to “group” visitor scores, and note that SAINT is flexible and allows you to rearrange/reallocate scores into different buckets depending on your changing needs and observations. After doing this, the report is much more digestible and actionable:

Now we’ve got something we can use. This makes it much easier to see how different engagement levels affect conversion. Again, how you divide up the different scores into groups is completely up to you, and you’ll probably want to adjust it as you dig into these reports. Also, note that you can do some really powerful things here with segmentation; using Discover, ASI, or Data Warehouse, you can focus in on the user experience both for users in the “Very High” group and in the “Low” group to see what they’re doing, and optimize your site around those findings.

The Numeric Event

The final piece of our Visitor Scoring system involves a numeric (a.k.a. “incrementor”) event, which you can enable using the Admin Console. (Mr. Greco documented numeric events as well in a previous post.) This might be my favorite part of the solution.

In addition to setting a counter eVar whenever a user does something that we’re scoring, this method also sets the value in an event so that you can view the score as a metric in various reports. This does involve using the s.products string, but don’t worry; we’re not going to mess with any actual product data. The implementation would look something like this (expanding on the examples given above and using event2 as our numeric event):

// after an internal search
s.eVar1="+4"
s.events="event2"
s.products=";;;;event2=4"

// after writing a review
s.eVar1="+5"
s.events="event2"
s.products=";;;;event2=5"

(Make sure to note that “event2″ exists both in s.events and in s.products in this case, and that there are exactly four semi-colons in s.products before event2 gets set.)

The great thing about this is that it allows you to see how various data dimensions affect engagement with key site elements. You’ll probably want to set up a calculated metric to divide this “score” metric by visits, because the raw score may be higher for different data dimensions simply due to varying levels of traffic; for example, when viewing this metric in the Campaigns report, a campaign that has 10,000 click-throughs will likely have a higher visitor score than a campaign that has 10 click-throughs simply because the overall traffic level is higher. When we view various marketing channels through the lens of this calculated metric, we immediately get a great report:

Social media sites, natural search, and partners are the clear winners in terms of bringing interested, engaged visitors to our site. Looks like we know where to focus our efforts—especially if we already know (from our experience with our counter eVar) how much more conversion a high-score visitor is likely to generate on our site. This gets even better when we focus on individual items within our top channels, breaking down this report by referring domain:

Not only do we know that social media brings eager visitors to our site, but we even know exactly which social media efforts/campaigns were most powerful (in this case, Fark.com, Delicious, LinkedIn, and Facebook).

Conclusion—and two bonus tips!

We didn’t mention this during the Advanced SiteCatalyst session at Summit, so be glad you read down this far. A really useful twist on this solution if you have two available eVars for use with visitor scoring is to configure one of them to expire at the end of the visit and another to a much longer expiration (such as “never”). The first eVar then gives a uniform view of individual visits, and how varying levels of interaction with key site elements affects conversion within the individual visit only. The second eVar would provide a view of “lifetime engagement” across multiple visits since the user last cleared his/her cookies. You can slice and dice both data sets to get some powerful views into how user behaviors may change over time.

Along these same lines, it’s possible that different teams or individuals may want to assign different scores to certain site activities. Don’t fight about it! Don’t let this scoring system destroy the harmony in your marketing department. Instead, similar to the tip just mentioned, if you’ve got an extra eVar, you can actually assign different scores to the same key site activities. Here’s a quick example using s.eVar1 and s.eVar2:

// this one is for the first team
s.eVar1="+5"
// this one is for the second team
s.eVar2="+30"

So there you have it. One tactic down, two to go. Next time, we’ll cover perhaps the most popular of the topics that we covered—the brand-new getPercentPageViewed plug-in. If you have any questions about Visitor Scoring, either based on this blog post or on the content we shared at Omniture Summit, please let me know by leaving a comment here. You can also contact me via Twitter (@OmnitureCare).

We love ideas. Always have. I spend most of my waking hours talking to Omniture users on Twitter, the Yahoo! Forum, blogs, etc., and many of these conversations center on ideas—how best to use our products, how to solve perplexing problems, and how to improve the Omniture Suite.

Because we love ideas—and because we especially love to use ideas—we recently launched the Idea Exchange, where any Omniture Suite user can log in and tell us exactly what ideas they have for any of our products. Then other users can “promote” the ideas that they feel would help them do their jobs most effectively, giving us a great sense of which features are most critical for our users.

(Speaking of critical ideas, enjoy this clip and just know that our building will not be demolished as we’re reviewing the ideas that you send us.)

In all seriousness, it’s a mutually beneficial situation. You have a quick and effective way of letting us know how we can improve your online business optimization efforts, and we have plenty of ideas and feedback that we can implement to make that happen. Our job then becomes to monitor your ideas, interact with you to understand the underlying business questions and your suggestions, and figure out how to get them to you. Various members of our organization spend a lot of time out there; one of the neatest things about last week’s Omniture Summit was hearing Brett Error (Vice President and General Manager, Products and Technology) speak with intimate knowledge about many of the ideas that have been submitted already. So, yeah. We’re listening intently.

You want more eVars? Submit an idea and watch others promote it. (Actually, that one has already been submitted, so you’d be best served by promoting the existing entry.)

Need rule-based SAINT classifications? Describe how you want it to work and follow the implementation process.

Want an entirely new product that does something cool? Yep, you guessed it; Idea Exchange FTW.

You can also use the Idea Exchange to get to know other users in the Omniture community. Everyone has a profile and can add information about himself or herself, upload an avatar, post contact info on several major instant messenger platforms, and more. It’s a great place to get to know how others are using Omniture tools, and it’s also a great place just to get to know others. You can even interface directly with various Omniture folks including developers, Product Managers, and more. You’ll definitely see me around. I’m one of the folks who makes sure things are running smoothly out there.

To begin using the Idea Exchange, just log in to the Omniture Suite as you normally would. Then go to Help > Idea Exchange. You will be prompted to choose a username. Once you have chosen a username, it will be paired up with your SiteCatalyst login info, and you can access the Idea Exchange thereafter either via the Help menu, or directly by logging in to ideas.omniture.com.

So have at it! We need your best ideas, and we need them now. And if you have any questions, please let me know. You can follow me on Twitter at @OmnitureCare, or just send me a message on the Idea Exchange. See you there!