Who is tracking your site? And what are they doing with the data? If you are reading this you probably answered “Webtrends, so I can get useful metrics on my site that help my business”. But are you sure that this is the only one? I am not talking about spyware, but tracking placed on your site for (not necessarily your) business purposes.

My job as a TAM (Technical Account Manager) is helping customers tag sites, configure profiles, and create reports. For those unfamiliar with the Webtrends TAM service this is sort of a combined consultant, admin, and analytics expert that you contract for a period of time rather than per project. But what does this have to do with who has your site data?

I got a request from a company who wanted to get more information on their social marketing efforts, starting with a series of Blogs. We were developing a strategy to track a long list of blogs done by their employees. Some were hosted on their company site but many were hosted on various other sites and platforms including personal domains.

What surprised me was the variety of tracking already implemented on the sites. Much of which I doubt the site owners knew about. I am not taking about spyware but actual tracking tags embedded in the page code that send data about their site somewhere else. Tracking included:

• Tracking from the CMS software
• Google tags, some of which seemed to be dropped in with other sections of code
• Hosted blog site tracking (e.g. blogengine.com)
• Site Host tags (You’d think that server hosts would do server side, not page based tracking)
• Tracking implemented by, with data going to, a consulting company who built the site
• Ad tracking tags

In some cases there were already 2 to 3 tracking tags on the site – before the Webtrends tag was added! Are you sure these types of tags aren’t on your site?

Is a site developer/hosting service/IT staff/banner ad/CMS software/etc. tracking your site for “quality control”? Where is this data going? Is your marketing data going to someone outside your company? Are they open about what they are doing with the data? Is it even your data even though it is coming from your site?

Ways to tell if your site is being tracked:

• Use a tool such as HttpWatch or Fiddler. These track requests coming from page loads. Watch for Post and GET requests to domains other than yours. This means data is going somewhere besides your domain.
• It is often easy to copy and paste a convenient function without realizing you are including a tracking code. When implementing outside code watch for calls using 1×1 transparent GIF files. This is a common method of sending tracking methods using an image not visible on your page.
• Check if your host and CMS have built in tracking and if so what information is being tracked and where does it go.
• If you have outside content on your site (e.g. a frame with ads) find out what they are tracking. Is it just ad response or is it including data on your site.

Even for your main site tracking – where does the data go? Can you see it or does it go into an unknown pool that you can pull some reports from. Do you have access to all the data being taken on your site? Who else has access to your data?

Some people worry about Big Brother watching them. While most of the companies that put tags on the sites were probably meaning well (“Let’s see how people are using our CMS product!”). They seemed more like a little brother. Not a benevolent “Watching you for your own protection” big brother. But the kid brother hiding behind the couch when his older sister brings home her new boyfriend – just to see if he can hear anything of interest. Occasionally, check behind the couch.

Learn How the Webtrends and Teradata Integration Solution Paints the Whole Picture

There is little doubt of how highly critical the importance of integrating online and enterprise data is. It is necessary to analyze both types of data in order to get the whole picture in a customer’s buying habits and satisfaction, and to improve conversion rates. Knowing when a customer connects with sales staff, knowing which direct mailings and catalogs the customer received, and knowing if the customer is chatting in social networks about the company or its products – all of these are tiny pieces to the whole picture. Unifying this information helps a company fine-tune their marketing strategies, reduce their marketing costs, and helps gives the customer a more personalized shopping experience and an incentive to purchase again.

This is a marvelous idea in theory, but very challenging to achieve in practice. Unless, that is, a company is using Webtrends’ enriched online data and Teradata’s Integrated Web Intelligence (IWI) solution, a partnership that is part of Webtrends’ Open Exchange program.

On December 2, 2009, in a live, interactive webinar, guest speaker John Lovett, a Senior Analyst for Forrester Research, Inc., will discuss why bringing online and enterprise data together is crucial. Next, Webtrends and Teradata will demonstrate their collaborated efforts solution of integrating online and enterprise data, and then will facilitate a question and answer session at the end.

Sign up today for the Integrating Online and Enterprise Data webinar and learn more about Webtrends’ enriched online data and Teradata’s Integrated Web Intelligence (IWI) solution for painting the whole picture to your customer’s purchasing experience.

Integrating Online and Enterprise Data Webinar
December 2, 2009
10am PDT / 1pm EDT

Hi again everyone!  I know I’ve been slow getting back to this thread, and for that I apologize.  It’s been a busy time here at Webtrends, and I was caught up in the whirlwind.  I haven’t forgotten my promise, though, to continue this thread, so here’s my next installment.  I’m skipping over the “How to read the Report” entries to post this one first about the numbers on the scenario analysis report.

What you’ve all pointed out in comments is true:  the numbers in this report don’t add up in the ways we might expect.  There’s a reason for this, though – actually, several reasons – that I’ll walk us through in this post (settle in; this is a thick read).   See, in a perfect scenario, all our visits would enter the process at the first step, convert through the steps in order, and complete the scenario without ever going anywhere else.  Most scenarios, though, have room for improvement – information is missing for the visitor, or a step is optional, or … well, you get the picture.  In such cases, it’d be useful for us to know how users meander in and out of a scenario, so we can identify steps we could improve upon, right?  That’s what this report is designed to do – and at the risk of sounding a bit self-serving, I have to say it does it quite well.  And that’s precisely why the numbers can seem so confusing – because this report follows the user’s meanderings, focusing on their activity rather than on totals.

Let me explain with an example (with many thanks to Xavier Le Hericy, who built the example!).  Let’s say I have three people visit my site and interact in a scenario.  Here are the paths each of them take:

Visit A:
(1) Views Page 1.
(2) Enters Step 1 (Product Page View) in the scenario.
(3) Views Page 2.
(4) Returns to Step 1.
(5) Views Page 3.
(6) Enters Step 3 (Started Checkout), skipping Step 2 entirely.
(7) Goes back to Step 2 (Cart Add).

Visit B:
(1) Views Page 1.
(2) Enters Step 1 (Product Page View) in the scenario.
(3) Views Page 2.
(4) Views Page 3.

Visit C:
(1) Views Page 1.
(2) Enters Step 1 (Product Page View) in the scenario.
(3) Goes directly to Step 2(Cart Add).
(4) Goes directly to Step 3 (Started Checkout).

None of them complete the checkout process for one reason or another.

Fairly straightforward visits, which lead to the following results in the scenario analysis report.  First, with step transitions:

Okay, let’s walk through our examples so we understand what we’re seeing here.  The text above from our engineering pals helps, but let’s translate it clearly into the examples we have above.

The Product Page View we see at the top left comes from Visit A. The two visits we see on the top right come from Visit A as well, though – in that visit, the user viewed a product page, then jumped to starting checkout.  So, if we try to add up numbers, we’d be literally tripling the number of visits – this entire line refers to a single visit, but show multiple paths through the scenario.  That’s one reason the numbers don’t add up.

Okay, moving on to Step 2.  Visit C converted from Step 1 to Step 2, so that’s the single visit we see next to the green arrow leading to Step 2.  But we also have Visit A represented once more – when the user moved from Step 3 to Step 2.  See how this is affecting the numbers?

Moving to Step 3, we see again that Visit C converted directly from one step to another – hence the 1 visit by the blue down arrow. Both Visit A and Visit C saw Step 3, which is why we have two visits at this step.  Visit A also is reflected on the left, since the user viewed a product page, then came to Step 3.  Finally, Visit A is also on the right, since it’s the visit that went back to Step 2 from this step.

So, this report shows us the flow of visits through the scenario steps, but there’s no way we could add up these numbers to get an appropriate number of visits to/from the scenario itself.  That’s just not what the report was designed to do!

Let’s check out the other view – scenario entry and exit pages:

Again, thanks to the Xavier for the explanations above – now let me walk you through the visits one more time.

Visits A, B, and C all enter the scenario from viewing page 1; therefore, they’re reflected both in the three visits on the left and the three visits to the product page view.  We don’t see the second visit to this step that took place; that’s reflected in the Step Transitions view instead.  On the right, we see that Visit B has moved on to another page and then never came back to the scenario – whoever they were, they played around elsewhere on the site, then headed out.

Visit C converted from Step 1 to Step 2, so that’s our one visit we see coming down to this step.  But we have two visits at this step, and none coming in on this step, which looks confusing until we take a look back at Visit A.  Ah – this visit did not convert from Step 1 to Step 2 (it jumped from 1 to 3 and back to 2), nor did it come into the scenario at Step 2 (it saw Step 1 first).  It was simply out of order – but it was still a visit to Step 2.  Hence, the two visits on this step – and we see another reason why these numbers won’t necessarily add up.

Visit A just keeps complicating things – this is the step from which this visit exited the scenario and either idled out of the visit or left the website entirely.  That’s the visit to the right of Step 2.

Visit C is now the only visit we have left, and it’s the visit that converted from Step 2 to 3 that we see.  However, again we have two visits on this step, and again it’s good ol’ Visit A that’s being reflected.  That was view #6 in Visit A’s progress on our website.  The visit we see exiting at this step is Visit C, since it didn’t move on to Step 4.

So, once again, these numbers won’t add up because of a couple of meandering folk on our website, weaving their way in and out of this scenario.  Multiply that by 10,000 visits or more, and you can start to see why these numbers seem out of whack on even infrequently-visited scenarios.

Let me say that I agree wholeheartedly that this lack of the clean-cut ability to add up numbers and see exactly how many visits were part of this scenario makes this report confusing.  However, that’s also what makes this report so valuable – we can see exactly what’s happening at any step in the scenario, and we can tell when visits meander through and out of the scenario.  That’s the actionable information this report is designed to provide – the kind of information that helps me optimize each step along the scenario, ensuring that my visitors always have the information they need to keep moving through the scenario and complete it.

Just to make you feel better, though, there actually are a couple of places in which the numbers add up.  Check out the image below:

Isn’t Xavier great? These are his notes, which were honestly a great relief to me.  I really wanted to be able to tell you that the numbers do add up, and here’s proof – some of the numbers really do!  Just not always the ones people ask you about.

So, I hope this post is helpful.  I’ve still got a couple more of these posts in the works, so I’d love to hear more about what you’d like to know about this report!

One of the challenges Webtrends staff and customers share is navigating through reams of information, both current and legacy. With many hundreds of book pages and help topics, sometimes searching for a simple concept or a feature explanation can turn into a crazy scavenger hunt through old printed manuals, saved PDFs, online customer center, and online help.

Which information has the highest value, and what terms, groupings, and search methods make the most sense for different kinds of users? What’s the best way to ensure information is accessible, relevant, and easy to explore? Even Webtrends insiders and industry experts don’t always agree on the answers to these questions. In the end, the real test of documentation usability is whether our customers can find the information they need to generate and understand analytics measurements.

Our Information Development team, formerly Documentation, has been more aggressively collecting information about how the information we publish gets used, understood, reused, transmitted….even crumpled up and discarded.  The investigation methods (which Julia talked about in a prior post) we’ve been using include card sorting (to help create a mental model of the terms and categories customers use to organize information), improved documentation tagging for better usage stats, and direct customer input.

At our Engage event in April we ran through some card sorting alongside our paper prototyping group. A few of you readers may have had the chance to attend the group or  fill out our survey at event. If you haven’t, please take a few minutes to tell us about how you use Webtrends information.

If you are interested in participating in usability testing at our location, online or just have a question - you can contact us at documentation@webtrends.com or comment here. We’re always eager to hear from you.

In the previous two articles*, I introduced the newly redesigned Webtrends custom tracking-code creation tool called Tag Builder and then provided some background into the user centered design methodology, Paper Prototyping, used to construct the flow of the application. In this post, I will highlight the most significant design improvements that defined the new look of Tag Builder and influence products to come.

1. Palette

The most noticeable change with the new Tag Builder is the monochromatic color palette. It is sparse, simple, and gets right to the point. As we push the design of Webtrends applications forward, we are intent on establishing a professional, concise, and highly engineered look and feel. Think German automobile. Think professional photography equipment. Webtrends products are professional products. The palette of charcoal, magnesium, and white dominate the design while hints of litho blue and stop sign red reveal themselves on hover states and error messaging. Color is still there, just reserved for when it is effective.

Color palettes (before and after)

In an addition to Tag Builder, you may have also noticed that the brand identity for Webtrends received a facelift. The webtrends.com website reflects this and is being rolled out across all of our marketing materials. The new wordmark is modern. The new palette, is complimentary to Tag Builder. Both are heavily monochromatic but the modern blue and warm grays have a much stronger presence with the outbound marketing while the product reserves its use of color for important indicators and highlights. This sophistication in coordination is like the outfits of a well dressed Hollywood couple attending the Oscars. Tag Builder was released before the branding update was revealed and so you will see future refinements to product design to reflect alignment.

2. Grid

The grid has received a lot of attention in the web design community in the past couple of years (thanks to pioneers like Khoi Vinh at the New York Times). The grid has been used for decades to organize typographical information in print so that blocks of type define the geometry and patterns of the design. This allows for the elimination of ornamental design clutter and reduces design to its essential elements. The new Tag Builder uses a 960 based grid and this allowed us to simplify the design dramatically. It may seem a bit open at first with excess whitespace, but as soon as you interact with Tag Builder, you’ll notice that the whitespace gives way to hover help text that no longer requires a click just to get the basic concept of each fieldset.

Tag Builder and the grid

3. Language

Because the previous Tag Builder required this additional click to access any help, the field labels ended up being sentence like in places and overly descriptive. This created a wordy design that left the user with a lack of confidence and uncertainty as it required more comprehension even for basic fields. The new labels are short and conversational in tone. This is easier to comprehend and it leads to a more confident user. For instance, a certain label read, “Single first-party cookie (use one first-party cookie across the primary and each subdomain: Cookie domain attribute.” In the new interface it now reads, “The site domain you want to track,” followed by the entry field. When the user hovers over the field, hover help appears and provides the user with an opportunity to confirm their assumptions as well as providing a link if the user wishes to explore the topic in depth. In addition to improving the readability of the labels and hover help, we also improved the error messaging for improperly filled out fields. The language short, concise, and straight to the point.

4. Indicator Dots

One of the most troublesome design challenges we faced with the new Tag Builder was how to clearly communicate completed fields when only one tab was visible at a time. We solved this challenge with a unique solution that we named, “indicator dots.” We noticed in the prototyping tests that users would click through the tabs a few times just as a driver in a car trying to make a left hand turn onto a busy street swings their head back and forth. With the indicator dots, now they at least knew which tabs that had completed some information in. This challenge grew as we realized that there was no clear way to message the user if they had improperly filled out a field on one tab when clicking the “Build Tag” button. So, we also leveraged the indicator dots to turn red when there is a field error on a tab. In the end, the indicator dots communicate to the user three things; how many choices are there on a tab, how many were filled out, and which fields contain errors when submitting the form.

Indicator dots

5. Confirmation

A related challenge to the indicator dots was the lack of any confirmation before the user clicked the “Build Tag” submit button. The previous version of Tag Builder directly triggered the “download” function of the browser when this was clicked. This worked ok in Internet Explorer but in Firefox and Safari, the download function didn’t allow the user to name the file. We realized we needed a confirmation page that condensed the information related to that particular tag setup on one screen paired with the option to name the zip package.

Confirmation page

Summary

We made many additional improvements with Tag Builder and I hope if you are a user, you have noticed an improved workflow. In the next article, I’ll walk through the changes we made regarding the web standards architecture.

I’d love to hear your feedback (thoughts, comments, questions, or critiques) on this new design direction.

*When I set out to write this five part series, I didn’t intend for it to be as drawn out. A new baby and product design improvements beyond Tag Builder are keeping me quite busy.

Not too long ago, I asked a former colleague to give me his view on the implementation of a measurement tool.  While I had partially expected a response related to the technical implementation, his response aligned with an important truth–implementations are only as important as their inherent backbone–a measurement strategy.

With that I would like to introduce Martin Cook, a Performance Measurement Consultant for EMC Conchango. Based out of London, England Martin has a deep background in measuring online behavior and helping organizations understand the benefit in producing sound measurement strategies as a basis for all digital investments.  Implementations and the software used are merely tools – while some are better than others none of them will do the job without proper application.

Next week at Engage we will be discussing implementations in my workshop, Planning for Success:  Implementation Strategies.  While the content will be focused around how WebTrends can be implemented on your website, I’d like to think of this idea of measurement strategy and ROI during our workshop discussions.

Why are measurement strategies so important?

Measurement strategies enable us to measure performance against business goals and justify the business case.

In this day and age how many clients still do not know how their online offering is performing? Especially in the current economic climate, it’s vitally important to understand your company’s performance against its goals and what is working well and more importantly what can work harder! Without this knowledge it can and has proved catastrophic to many businesses throughout the world. If your organisation is even the slightest bit slack at establishing an effective measurement strategy, your competition will be more agile at reacting in these times and capitalising on opportunities. We need to look beyond the credit crunch and economic downturn to focus on the opportunities that will arise, upon the upturn.

(more…)

Hi Everyone,

Last week I had the pleasure of listening to Eric Peterson speak not once, but twice. The first time was during a Coremetrics webinar on campaign attribution and the second later that evening at the local Web Analytics Wednesday where Eric delivered a longer presentation that included the same attribution material. And while I have a great deal of respect (and even friendship) for Eric and an equal amount of respect for Coremetrics, I feel a need to challenge the content.

For awhile I’ve been speaking about the emergence of the third generation of web analytics, as I call it. For those that haven’t heard me present this before, the first generation was characterized by IT departments measuring web site activity via software installations of log file analysis tools. The second generation was dominated by marketing departments utilizing hosted solutions and page tagging. The primary value these two generations of solutions provided were aggregate reports, along with rudimentary ad-hoc analysis capabilities (rudimentary, that is, compared to modern business intelligence systems).

Whereby the first two generations were characterized by reports, the third is certainly about the data – the open access to un-aggregated visitor detail data and the endless forms of true analysis that can be performed with it. Knowing that Coremetrics is one of a few major vendors to store un-aggregated data in an industry-standard database (along with WebTrends) I was expecting a thoughtful discourse on statistical modeling. Alas, what we were told was to utilize not one, but three flawed attribution models (last, first and equal), in hopes that three wrongs would make it right I suppose.

Since our high school statistics classes we have been taught the difference between correlation and causality. Statistics show that as ice cream sales increase, so do drowning deaths. Therefore, ice cream causes drowning, right? Of course not – it is the onset of warmer temperatures that indirectly leads to both. As trite as this example may seem, it is no different than the fallacy that a campaign’s inclusion in a visitor’s click-path prior to conversion means that it had a causal affect on the conversion, or that it belongs in our campaign portfolio. The same campaign may have been clicked on by many more non-converting visitors … at substantial expense.

True, if a visitor clicked on a campaign prior to conversion, it’s certainly more likely to have had a causal impact. That’s especially true for the last campaign. But if we’re going to finally break away from the flawed last-click attribution model, why not do it correctly? We have the data – let’s use a statistical model.

Now for the less-than mathematically savvy user of web analytics, no, this doesn’t mean your solution will be more complicated. Quite the contrary. Before credit card companies implemented mathematical models to detect fraud, we consumers would first learn of fraud only after we received our statement. And then after weeks arguing with our vendor we might have gotten the charges removed. Today we get a phone call within hours of the questionable transaction and a new card sent overnight to us, no questions asked. Math made our lives easier.

So will it be for campaign attribution. Imagine a campaign report that tells you, in a statistically valid way, which campaigns and campaign attributes actually had a positive contribution to conversion and to your campaign budget, versus those that didn’t. Then imagine that same report telling you how to improve results. I propose the following report:

Don’t sweat the details – I just punched some example data into a spreadsheet. Instead, focus on the bigger picture of having a report that shows you how your campaigns truly performed and recommends to you an adjusted mix based on the current set of campaigns. Then imagine the data for auction-based networks being automatically passed to an automated campaign optimization system. Now that would be progress towards true optimization of campaign budgets while also making the marketer’s job much easier.

Note that at the moment WebTrends doesn’t provide the above report either (but we do have the requisite data in a readily accessible format). My point is that it’s time to embrace the third generation of this industry and start truly leveraging the data in mathematically and scientifically valid ways.

- Barry

P.S. Please send me your thoughts on the dream campaign report.

Hi Everyone,

Last week I had the pleasure of listening to Eric Peterson speak not once, but twice. The first time was during a Coremetrics webinar on campaign attribution and the second later that evening at the local Web Analytics Wednesday where Eric delivered a longer presentation that included the same attribution material. And while I have a great deal of respect (and even friendship) for Eric and an equal amount of respect for Coremetrics, I feel a need to challenge the content.

For awhile I’ve been speaking about the emergence of the third generation of web analytics, as I call it. For those that haven’t heard me present this before, the first generation was characterized by IT departments measuring web site activity via software installations of log file analysis tools. The second generation was dominated by marketing departments utilizing hosted solutions and page tagging. The primary value these two generations of solutions provided were aggregate reports, along with rudimentary ad-hoc analysis capabilities (rudimentary, that is, compared to modern business intelligence systems).

Whereby the first two generations were characterized by reports, the third is certainly about the data – the open access to un-aggregated visitor detail data and the endless forms of true analysis that can be performed with it. Knowing that Coremetrics is one of a few major vendors to store un-aggregated data in an industry-standard database (along with WebTrends) I was expecting a thoughtful discourse on statistical modeling. Alas, what we were told was to utilize not one, but three flawed attribution models (last, first and equal), in hopes that three wrongs would make it right I suppose.

Since our high school statistics classes we have been taught the difference between correlation and causality. Statistics show that as ice cream sales increase, so do drowning deaths. Therefore, ice cream causes drowning, right? Of course not – it is the onset of warmer temperatures that indirectly leads to both. As trite as this example may seem, it is no different than the fallacy that a campaign’s inclusion in a visitor’s click-path prior to conversion means that it had a causal affect on the conversion, or that it belongs in our campaign portfolio. The same campaign may have been clicked on by many more non-converting visitors … at substantial expense.

True, if a visitor clicked on a campaign prior to conversion, it’s certainly more likely to have had a causal impact. That’s especially true for the last campaign. But if we’re going to finally break away from the flawed last-click attribution model, why not do it correctly? We have the data – let’s use a statistical model.

Now for the less-than mathematically savvy user of web analytics, no, this doesn’t mean your solution will be more complicated. Quite the contrary. Before credit card companies implemented mathematical models to detect fraud, we consumers would first learn of fraud only after we received our statement. And then after weeks arguing with our vendor we might have gotten the charges removed. Today we get a phone call within hours of the questionable transaction and a new card sent overnight to us, no questions asked. Math made our lives easier.

So will it be for campaign attribution. Imagine a campaign report that tells you, in a statistically valid way, which campaigns and campaign attributes actually had a positive contribution to conversion and to your campaign budget, versus those that didn’t. Then imagine that same report telling you how to improve results. I propose the following report:

Don’t sweat the details – I just punched some example data into a spreadsheet. Instead, focus on the bigger picture of having a report that shows you how your campaigns truly performed and recommends to you an adjusted mix based on the current set of campaigns. Then imagine the data for auction-based networks being automatically passed to an automated campaign optimization system. Now that would be progress towards true optimization of campaign budgets while also making the marketer’s job much easier.

Note that at the moment WebTrends doesn’t provide the above report either (but we do have the requisite data in a readily accessible format). My point is that it’s time to embrace the third generation of this industry and start truly leveraging the data in mathematically and scientifically valid ways.

- Barry

P.S. Please send me your thoughts on the dream campaign report.

Hi Everyone,

Last week I had the pleasure of listening to Eric Peterson speak not once, but twice. The first time was during a Coremetrics webinar on campaign attribution and the second later that evening at the local Web Analytics Wednesday where Eric delivered a longer presentation that included the same attribution material. And while I have a great deal of respect (and even friendship) for Eric and an equal amount of respect for Coremetrics, I feel a need to challenge the content.

For awhile I’ve been speaking about the emergence of the third generation of web analytics, as I call it. For those that haven’t heard me present this before, the first generation was characterized by IT departments measuring web site activity via software installations of log file analysis tools. The second generation was dominated by marketing departments utilizing hosted solutions and page tagging. The primary value these two generations of solutions provided were aggregate reports, along with rudimentary ad-hoc analysis capabilities (rudimentary, that is, compared to modern business intelligence systems).

Whereby the first two generations were characterized by reports, the third is certainly about the data – the open access to un-aggregated visitor detail data and the endless forms of true analysis that can be performed with it. Knowing that Coremetrics is one of a few major vendors to store un-aggregated data in an industry-standard database (along with WebTrends) I was expecting a thoughtful discourse on statistical modeling. Alas, what we were told was to utilize not one, but three flawed attribution models (last, first and equal), in hopes that three wrongs would make it right I suppose.

Since our high school statistics classes we have been taught the difference between correlation and causality. Statistics show that as ice cream sales increase, so do drowning deaths. Therefore, ice cream causes drowning, right? Of course not – it is the onset of warmer temperatures that indirectly leads to both. As trite as this example may seem, it is no different than the fallacy that a campaign’s inclusion in a visitor’s click-path prior to conversion means that it had a causal affect on the conversion, or that it belongs in our campaign portfolio. The same campaign may have been clicked on by many more non-converting visitors … at substantial expense.

True, if a visitor clicked on a campaign prior to conversion, it’s certainly more likely to have had a causal impact. That’s especially true for the last campaign. But if we’re going to finally break away from the flawed last-click attribution model, why not do it correctly? We have the data – let’s use a statistical model.

Now for the less-than mathematically savvy user of web analytics, no, this doesn’t mean your solution will be more complicated. Quite the contrary. Before credit card companies implemented mathematical models to detect fraud, we consumers would first learn of fraud only after we received our statement. And then after weeks arguing with our vendor we might have gotten the charges removed. Today we get a phone call within hours of the questionable transaction and a new card sent overnight to us, no questions asked. Math made our lives easier.

So will it be for campaign attribution. Imagine a campaign report that tells you, in a statistically valid way, which campaigns and campaign attributes actually had a positive contribution to conversion and to your campaign budget, versus those that didn’t. Then imagine that same report telling you how to improve results. I propose the following report:

Don’t sweat the details – I just punched some example data into a spreadsheet. Instead, focus on the bigger picture of having a report that shows you how your campaigns truly performed and recommends to you an adjusted mix based on the current set of campaigns. Then imagine the data for auction-based networks being automatically passed to an automated campaign optimization system. Now that would be progress towards true optimization of campaign budgets while also making the marketer’s job much easier.

Note that at the moment WebTrends doesn’t provide the above report either (but we do have the requisite data in a readily accessible format). My point is that it’s time to embrace the third generation of this industry and start truly leveraging the data in mathematically and scientifically valid ways.

- Barry

P.S. Please send me your thoughts on the dream campaign report.

Years ago, when my wife and I were new parents, we decided to go out on a date one evening and leave the care of our son to a babysitter for the first time.  We were nervous parents and our heads were filled with visions of everything that could go wrong, nearly convincing ourselves to abandon the plans and stay home.

We had to go out and decided we had to get over our fears.  We started by writing down all of the emergency information we could think of on a single page for the babysitter.  This included phone numbers for our cell phones, doctors, the hospital, poison control, grandparents, friends in the area, and anything else we could think of.  We listed our son’s favorite foods, when he should be put down for bed, and what his favorite bedtime stories and lullabies were.  We prepared, packaged, and labeled food and placed it in the refrigerator, made up a bottle of milk, laid out his pajamas, as well as many other preparations.  Again, we were nervous.

When the babysitter arrived, we walked her through the house, showed her all the preparations, and went over and over the emergency information a few times.  This all took less than ten minutes.  She seemed to understand it all and her let us know that she had done this before and everything would be OK.

(more…)