About a month ago I asked about analytics widgets. Today I stumbled across one from WebTrends.

Cool.

Hey Omniture, where's mine?

Posted by Craig Danuloff at 9:43 PM | Permalink | Comments (1)

With the recent Google purchase of Feedburner, I presume we'll soon see RSS subscriber stats included in the Google Analytics reports.

This is great for anyone who relies on Google Analytics and blogs or uses other RSS feed distributions as part of their online marketing - RSS is just another channel and should be included in traffic reports alongside of search and email. I didn't see this talked about when the acquisition was being discussed, but this is a big win for Google Analytics.

I've never understood why the analytics vendors didn't try to compete with Feedburner themselves - or buy them - a huge new channel and tracking opportunity emerges and NONE of them even bother to try and capture it. Assuming Google eventually moves the entire Feedburner UI into Google Analytics they'll all regret it even more.

Posted by Craig Danuloff at 8:51 PM | Permalink

By the way - wasn't something like this the original widget?

And where exactly are the cool web analytics widgets from the major vendors that let me share interesting stats with my site/blog readers? Not just 'page view counters' but cool configurable stuff with selectable metrics or metric-slide-shows and graphics etc.

Omniture has their cool Dashboard Player - how about widgetizing that? Come on guys, have a lunch-hour developer contest or something.

You'd think they'd want the branding visibility even if they don't want to just give us some cool toys.

Posted by Craig Danuloff at 1:01 AM | Permalink | Comments (1)

Comscore says that in April 2007 there were 177,000,000 unique viewers for widgets - which by their definition are "data files that can be embedded into a site’s HTML code and are typically displayed in a small viewing pane on the site." And they seem to have skipped YouTube, so that would have probably increased the number by at least a couple...

Most widgets have controls or interactivity options for users, so who knows how many clicks took place inside these widgets? Nobody. And that's a problem.

Widgets currently exist outside the purview of web analytics - you can't track the clicks inside widgets on your site, and if you distribute your own content in widgets you can't incorporate those 'remote' views or interactions either.

So web analytics are ignoring the Widget Economy. As the numbers above show, this is a material oversight. Analytics vendors need to get themselves together and create a standard widget reporting mechanism, and present it to the widget developers of the world, pronto. Widget consumers (bloggers in large part, but other sites too) then need to demand that the widgets they support embrace this standard, enabling them to track the clicks that take place within these wonderful little windows.

By any expectation we're at the very start of Widget World. Time for the analytics folks to join the party.

Posted by Craig Danuloff at 10:41 PM | Permalink | Comments (2)

Now that the Sopranos ending is out of the way, there's just one more secret left - The Future of Web Analytics.

I'll be revealing the answer to that one this Weds evening at the Web Analytics Wednesday event in Philadelphia. I'm pleased to have been asked to speak at this event, and hope this reveal doesn't spark the same level of controversy as that other one did.

In case you aren't familiar, Web Analytics Wednesday is the world's only distributed networking event for web analytics professionals. Open to everyone, practitioners and vendors alike, Web Analytics Wednesday is a free event allowing you to meet folks with similar work interests.

If you're in the area, please join us at 6pm (RSVP requested). The event is being held at Teikoku Restaurant @ 5492 West Chester Pike - Newtown Square, PA 19073 - Phone: 610.644.8270

Posted by Craig Danuloff at 10:58 PM | Permalink | Comments (3)

Web Analytics is struggling for the soul of online marketing. At least it should be. There is far too much time and money spent badly, far too many users subjected to horrible web experiences, and far too many misplaced priorities – all the very things that good analytics and analysis can help prevent.

And although they might not see it in these grand terms, many of the Analytics Elders have been talking lately about some fundamental issues that we as practitioners need to get right before what has traditionally been the tail can finally start wagging this dog.

First we have the relative role of tools and talent. Avinash now famously proposed the 90/10 rule for allocating analytics budgets. I’m fairly certain it wasn’t meant to be the literal dictate some like to argue about but was intended to be a theoretical ideal and probably a little bit of an exaggeration intended to make the point that 10/90 split found in most companies was really wrong. The vital point here is that it doesn’t matter that you buy (or install) web analytics it matters that you do web analysis.

Eric Peterson suggested the debate shift from money to time, and more specifically that 20% of time be spent on hiring and 70% of it on the actually doing of the work. More importantly, he makes the case for a known and repeatable analytic process so that results aren’t based on raw talent and individual superstars, both of which are in short supply. I’m sure he’s right but it will take a lot of effort and discipline to define these processes and given the huge range of efforts that analytics can measure I think this one’s going to be a long time coming before most companies (or the analytics world in general) have so many processes well defined that it doesn’t take an exceptionally skilled analyst to deliver a constant stream of valuable and actionable insights.

The second issue is that of exactly what analytics should deliver: reports or insights. Avinash proffered the 20/80 rule (again I believe as a guideline) for allocating time between reporting and analysis. While ‘less reporting more analysis’ has become the analysts’ theme song, the idea that any reporting that does get delivered should be ‘actionable’ has also become a frequent refrain. Gary Angel effectively took the counterpoint in this debate when he argued that it isn’t practical for reports and KPIs to be actionable because the KPI itself cannot contain enough context. With some great examples and follow up posts (#1 and #2) he clearly demonstrates both the value and process of analysis – while setting a simple goal for reports which justifies their existence even if they aren’t actionable: “Reporting is designed to provide information back to key decision-makers within an organization about the web channel.”

Eric wrote a reply to Gary’s post that is worth reading, and Jim Novo did a fantastic job of putting both into some larger context. My own sense is that KPIs and reports are vital for understanding, which is in and of itself important and the prerequisite to ‘actionable insights’ for both the marketer/executive and the analyst.

Finally we have a step which I’d never heard discussed in the context of analytics but clearly is part of the analytics process. Bryan Eisenberg pointed out that the step after actionable is action. Analytics only works if you have reports, and analysis, and someone ‘talented’ enough (Bryan’s word) to take action.

Talk about raising the stakes. Now to be successful with web analytics we need to choose and properly implement the right software, gain complete and accurate data capture, select the right KPIs, define and prepare the right reports, have the time and skill to perform analysis and define insights (with the inevitable trackbacks for more tagging, data, and reports), possess the ability to recommend actions based on those insights, AND find a client (internal or external) who has the awareness, authority, and resources to go make significant changes to campaigns, web pages, offers, or whatever else needs changing.

Of course this last piece (the client) isn’t really the last step – they’ll need to be pre-qualified as having the right attitude and proclivity to action way up front, informed and updated throughout, and probably play at least a semi-active role in the acceptance of the analysis and recommendations. In other words they aren’t just going to do what you tell them.

I don’t know that there’s any aspect of online marketing which is moving faster in its development than web analytics is right now (both in technology and in practice) but I also can’t think of any areas that has further to go or in which more is at stake. My instinct is that as a group we’re not going to make it – and that broadly speaking the metric-less marketers will keep driving their boats into rocks without even knowing it for quite some time to come.

Having said that, this is an astounding opportunity for those who can make it work. I would suggest that the sky is the limit for both salary and influence for those who can really demonstrate command and the ability to execute to the potential of this space over the next few years.

And it’s worth pointing out what should be obvious - there are a very impressive group people writing and working to bring maturity and ultimate success to this market space. I came across all of the above referenced articles in just one session catching up in my RSS reader. Those I’ve linked here and another dozen or so really smart folks are certainly going to make it a good and interesting fight.

Posted by Craig Danuloff at 9:26 PM | Permalink | Comments (3)

Our friend Avinash Kaushik's book is now available, and the Commerce360 copies arrived today. I've only skimmed it thus far, but it looks great and I really can't wait to dig in this weekend. Anyone who reads Occam's Razor already knows the man can think and write and has a tremendous perspective on web analytics - so go get a copy!

It's also worth pointing out again that Avinash is donating 100% of his royalties to charity in the form of The Smile Train and Médecins Sans Frontières (Doctors Without Borders) - so don't just buy one, order a case or so for the office (and use the link at left to order and the Amazon affiliate fees will be donated too).

To help promote the book, Avinash has asked for photos of his readers, so here is one from some of the Commerce360 team:

Posted by Craig Danuloff at 10:30 PM | Permalink

While at the emetrics conference earlier this week I had the chance to sit down with Jeffrey Veen, just a few minutes after he unveiled Google Analytics Version 2 to the world. I recorded our conversation and am pleased to present our first PodCast™ below.

Jeffrey is UI Design Manager at Google, formerly of Adaptive Path, and author of The Art & Science of Web Design. Jeffrey is very well known and respected for his design work, and his impact is very clear in the fantastic UI advances in the new GA version.

In our 12 min talk, we covered:

• The design and usability goals and process they used to improve them
• The presentation of 'related data' on report screens
• New segmentation abilities
• Report distribution features
• New date selection and data-range features

Before you listen, I must admit that it sounds very much like I called him Jeffrey Been (which I may have although I knew his name long before that day - if so offer my apologies) but I assure you his name is Jeffrey Veen.

To listen, either click the Play button or download the MP3.

I've had only a few minutes with the new version myself since then, and hope to share some more thoughts soon. There is extensive coverage from other bloggers, including a pretty detailed look from Avinash Kaushik. UPDATE: Full round-up of coverage is here.

Posted by Craig Danuloff at 3:58 PM | Permalink | Comments (1)

If website analytics are such powerful tools which reveal truth and/or insight into online marketing, why is everyone so reluctant and even resistant to using them as a central part of their daily routine? Why does using these tools feel like such a chore?

As something of an analytics nut, I’m in some ways amazed at how active online marketers seem to view and treat a dip into their analytics package with the zeal usually reserved for a trip to the dentist or doing your taxes. One would think that marketers couldn’t help but check these ‘scorecards’ nearly constantly to see how they’re doing, catch mistakes early, figure out how to improve, etc. But it seems that analytics are treated (in terms of actual use not conceptual approval) like a necessary evil rather than an exciting or empowering experience.

I’ve come to the conclusion that it’s because the marketers are process centered and the analytics tools are data centered. The marketers want a nice comfortable chair to sit in and the analytics software keeps handing them some sticks, a box of nails, a hammer and a roll of fabric. That may be ok if you have a carpenter (or an analyst) around, but it gets tiring when it happens every time you want to sit down.

There are too many levels of ‘translation’ necessary to get from the process I’m thinking about relative to my SEM campaign, website content update, or client segmentation issue to how the information in these reports tells me anything near a complete picture.

I want to know how the new SpiderMan-3 keywords are doing relative to our spidy-pajama bottoms and lunch boxes. As a product manager or marketer, or even a search marketer, this means what I’d like to see is a quick snapshot summary of the spend, traffic, and results for the related keywords, how these visitors used my site (consumed or interacted with pages and content), who they were (how many were first time visitors vs frequent buyers) and how purchasers really behaved (what else they bought in the same order, over time what they came back and bought later, how their basket size compared to other current product groups or promotions), and maybe a few more things. That’s a lot of stuff, but I spent some money, dedicated some time, and I want to know what happened.

Does a non-analytics-expert have any chance of getting the answers to those questions using today’s analytics software? Could even a moderately experienced analyst get the majority of these answers in less than 30 minutes?

PPC-tracking, path analysis, segmentation controls, lifetime value information, fallouts and revenue analysis are distinct in terms of how analytics software tracks data and present reports. Many of these data elements cannot be joined or cross-referenced in any way within the current interfaces. Analysts know and (begrudgingly) accept this. Non-analysts are surprised and appropriately befuddled by those facts.

PS: I’m heading out to emetrics.org in SF next week, where I look forward to righting all these wrongs with my vendor friends, and drinking to the challenges we must endure with the analysts.

Posted by Craig Danuloff at 9:18 PM | Permalink | Comments (2)

Moving quickly beyond the fact that it is a a major panel-based audience measurement company releasing a study that says (surprise) cookies aren't accurate, the study is most likely completely accurate and there are several interesting facts contained therein:

• 31 percent of U.S. Internet users cleared their first-party cookies during the month.
• Within this user segment, the study found an average of 4.7 different cookies for the site.
• Among the 7-percent of computers with at least 4 cookie resets, comScore counted an average of 12.5 distinct first-party cookies per computer, accounting for 35 percent of all cookies observed in the analysis.
• Web site server logs that count unique cookies to measure unique visitors are likely to be exaggerating the size of the site’s audience by a factor as high as 2.5, or an overstatement of 150 percent.
• comScore’s analysis of third-party cookies revealed an average of 2.6 distinct cookies per computer in December, indicating a similar rate of overstatement as the first-party cookies.

Further proof that web analytics are true but not accurate (a most excellent quote I heard from Jim Sterne).

Definitely a problem if you're a brand marketer paying for coverage across unique humans. Much less of an issue if you're an online retailer trying to understand visitor behavior. Sure the numbers won't be right (did you really think they were?) but the trends should be true and accurate. Don't forget it's not just uniques that is affected - things like ave-number-visits-before-purchase and similarly 'cookie-based' cumulative measurements are also equally impacted.

Two thoughts:

• What if web analytics software provided an option to compensate for the likely overcounting of uniques due to cookie deletion. A simple place we could tell it to decrease the unique visitor count by 33% (or whatever) because that's what we believe our overstatement rate to be. Wouldn't it be nice to just quickly see reports with this clean-up IF we believed that made them more accurate?

• Any chance we could get all those cookie deletion utilities to empty them out and then write a new cookie telling the rest of us the date of the last flush?

Related Discussion: AVC AttentionMax

Posted by Craig Danuloff at 8:25 PM | Permalink | Comments (1)

Eric Peterson author of Web Analytics Demystified (and other forms of fame) is asking all Web Analytics professionals to complete this survey.

To encourage you, Eric is providing the followignto thank you for the few minutes it should take you to complete this survey:

1. A complimentary copy of the results of this survey (available in mid-May 2007)
2. A twenty-five percent discount off the purchase price of Eric T. Peterson's Web Analytics Demystified and Big Book of Key Performance Indicators (normally $49.99, you'll save $12.50 and pay only $37.49 which is like getting the Big Book of Key Performance Indicators for FREE!)

I just completed the survey in just a few minutes. Go ahead, do it.

Posted by Craig Danuloff at 11:20 PM | Permalink | Comments (1)

We held our first formal internal Intro to Web Analytics/Omniture SiteCatalyst training session today, and while I did a poor job of time planning (we only got through about 1/2 the planned 90 min presentation in a full 2 hours) the conversation and interest in continuing to push analytics to the center of our work was great.

There were 15-17 people in for most of the session, including everyone from members of the sales team to account managers to the search team to strategists to new-to-Omniture analysts to a person from our finance dept. (We've grown a lot lately, and I've not posted enough about that - will catch up there soon.)

Preparing for this was interesting, in that it caused me to try and define the role and potential for the kind of web analytics and analysis we strive to provide our clients in a way that only some blank PowerPoint slides and a deadline can.

At the core I think web analytics should assist in four areas - optimizing marketing resources, satisfying visitors, improving the site, and increasing revenues. Many of those are inter-related but each drives a hierarchy of questions which the analytics can help answer.

We talked a bit about having the right expectations with web analytics in relation to a project, how it was a long road from tagging to reporting to analysis to insights and finally to actions. And beyond preparing for that time-frame, how to think about web analytics software as a tool that enables a larger process where questions result in answers which beget more questions and only after a whole bunch of those cycles can you start to form opinions and then get to the work of starting to figure out what to do about it. Then of course worrying about how to convey the ideas and share with others (particularly time-pressed managers) all that data in a useful context to someone who didn't go through that process.

This is only a bit of what was touched on, and all that was before we even got into SiteCatalyst! You can see why we ran over time! But I thought it was important to set the tone before we looked at the software that this isn't a 'quick answer lookup system'. It's easy to get that impression but I think it leads to either disappointment or massive underutilization.

We did finally get a bit into SiteCatalyst, touring the core aspects of the interface and trying to establish the pattern by which it is used, at least in relation to generating basic reports. To summarize and help newbies consider all of their options I came up with this list:

1. Develop the question
2. Find right report
3. Set correct time-frame
4. Display relevant metrics
5. Configure The Graph
6. Sort appropriately
7. Sub-relate (optional)
8. Compare or Trend
9. Wish for more data/relations
10. Print
11. Email
12. Bookmark
13. Schedule
14. Set Alerts
15. Add to Dashboard
16. Share

What did I miss on this list? Comments appreciated.

Tomorrow we finish what we didn't today, getting through lots of sample screens and reports covering all the program basics. Next week we'll take a quick tour of the many advanced features.

Want a free seat at the class? Come work with us!

Posted by Craig Danuloff at 9:28 PM | Permalink | Comments (2)

Just before the Omniture Summit (OMTR) began, the company announced a new professional certification for SiteCatalyst, and that initial tests to gain this certification would be offered during the conference.

Commerce360 Web Analyst Andrew Orsini was among the first to take and pass the test, earning his Omniture Certified Professional status for SiteCatalyst. Andrew had already been certified as an Omniture Certified Professional for Implementation.

Andrew was the first full-time web analyst to join Commerce360, and while he's since been joined by several others he's been the lead in our implementation work and worked on analysis for many of our clients. Congratulations to Andrew on both of these certifications.

“In today’s competitive environment, online marketers are looking for a way to distinguish themselves with current or potential employers,” said Steve Wellen, vice president of client services at Omniture. “The Omniture certification will assure businesses that the professionals they are hiring or staff they have in place have the highest levels of expertise. We encourage individuals that regularly use or plan to use Omniture solutions to gain these respected credentials.”

I too sat down to take the test while at the Summit, but some problems with the WiFi signal in the hotel resulted in the cancellation of that session. This disappointed certain people who thought it would be great fun to tease me were I to fail that test (I had perhaps given the indication that there was no chance of that...). But I'm glad to report that I was able to retake the test later, and while it was sufficiently difficult that I'll admit to being a bit nervous when hitting the final 'submit' button, I did also pass and gain my OCP : SiteCatalyst certification.

Posted by Craig Danuloff at 12:14 AM | Permalink | Comments (1)

Website analytics is sold as if it's a solution, but it's really a problem. Buying analytics software just means you have to install it, learn how to use it, figure out what reports you want, and the figure out what they mean. And while a good report or dashboard may answer a few questions, a really good one creates a lot more.

It's necessary. It's beneficial. It's useful. But it isn't easy.

Last week at the Omniture conference was incredibly motivating in terms of the potential of website analytics.

But it was also sobering in terms of how much there is to do to get even a fraction of the potential implemented and accurately reported on. Session after session covered big interesting topics filled with possibility that could barely have the surface of their surface scratched in 60 or 90 minutes. And even with a dozen or so sessions there were probably another 100 or so topics not even covered. This industry/technology and our collective expertise is at the 1 yard line with 99 yards to go.

And another interesting note; while all the best minds will tell you it's not about the reports but rather about the analysis, insights, and actions that follow - there wasn't to my knowledge a single session or organized discussion on anything near analysis or the 'process' of gaining insight from of all this tracking. Everything was implied - do this, you'll be able to track that, then "you'll know".

But we all know that the buzz from a new report only lasts a few minutes. Then you stare at it and wonder what you can do with that information, or think of 5-10 follow up questions and realize you can't think of an easy way to get at least a couple of those answers, or have some ideas and need to do some serious work to implement them to test them out. The report may have moved you forward but it didn't get you to any destination.

To be fair there just wasn't time. If you don't yet have the tagging right to even collect the data let alone format the basic reports, insights aren't your most pressing problem. And oh man is tagging and data collection and import still a problem. In fact, what led me to this cheery post is a great post on that subject from my new friend Gary Angel over at SemAngel.

Anyone want to buy an 'Analytics Isn't The Answer' t-shirt? Or have ideas for the clever punchline that should be printed on the back?

Posted by Craig Danuloff at 10:50 PM | Permalink | Comments (3)

As usual, Jeffrey Eisenberg got me thinking. He was commenting on an interesting post about the cost (and intellectual ingredients) of developing a single landing page, and offered some ideas about indirect costs associated with that page (analysis, testing, etc.) which should also be considered when measuring ROI. He ends challenging agencies to defend their own ROI.

Why isn't there an easy way to collect the fixed and variable costs associated with a web page and then look at the ROI of that web page? Or a campaign?

In Omniture SiteCatalyst it's possible to use SAINT to assign the landing page associated with each keyword in paid search, and then see the expense, revenue, and ROAS. But assigning landing pages to tens of thousands of keywords is A LOT of work. The software knows which landing page each paid click arrived at - couldn't this be automatically populated? (hm, rule-based classifications, there's an idea...)

The problem is harder for other traffic sources. Even if you know the variable costs associated with clicks coming from other sources it can be very hard to get that data into the system. And assigning values in SAINT classifications against the tracking codes doesn't help because you can't do calculations against that data. I'd imagine Omniture Genesis connections properly pass this data from email and some other sources, but it's impractical to assume that you have that level of integration for all traffic sources.

A default classification of variable cost for each tracking code and fixed cost for each campaign name - both of which could be mathematically operated on as calculated metrics - would address this issue.

The idea of assigning a fixed cost to a campaign and having it automagically allocated across all of the activity for that campaign would be very interesting. What if we could see the ROI for a campaign based its direct variable costs, these variable costs plus direct fixed costs, these fixed and variable costs plus an allocation of staffing costs, etc? The numbers may be scary but the honest evaluation of these programs would be quite interesting.

The idea of a page having a fixed cost associated with it is interesting too, although in many cases assigning the combined fixed costs at the campaign level would be sufficient. However it's implemented there's little doubt there is a 'real world gap' between the available reporting in most current analytic software and the way marketers really consider and evaluate these campaigns. Implementing more complete cost accounting would go a long way to close this gap.

Posted by Craig Danuloff at 10:14 PM | Permalink

The Omniture Summit in Salt Lake City has ended, although the snow on the east coast meant it took many of us a lot longer than expected to get home. The conference was great, and I hope to post a series of comments and thoughts about the content and ideas from the event this coming week.

For now however, I just want to thank everyone at Omniture responsible for putting on what was undoubtedly the best organized and executed conference I've ever attended. Both as a sponsor / exhibitor and an attendee they made everything a pleasure. I spoke with many others during the week and I know my sentiments were widely held.

Here's a brief list of things that made the event stand out:

• The facilities at Grand America were excellent. It's a four-star hotel that is elegant without feeling stuffy, spacious without being sprawling, and even the night manager and security team were good guys (don't ask.)

• Nothing was rushed or short-changed. Buses to the external events ran on infinite loops so you didn't have to feel rushed coming or going, when it became clear that folks wanted to keep going late into the night Omniture planned an 'after-party' to follow a huge night out on the town and kept the entertainment and drinks going well into the early hours, there were an abundance of terminals to check email, laptops and support people in the 'hot zone' help rooms, and more than enough friendly staff to assist with registrations or logistical issues anytime throughout the three-four days of the show. There are many more examples but all of them contributed to a relaxed atmosphere and a real demonstration that they cared more about their guests than saving a few bucks on logistics.

Continue reading "Omniture Summit 2007 : Wrap Up #1" »

Posted by Craig Danuloff at 1:43 AM | Permalink

The quote that best summarizes the goal and opportunity of blogging to me is one which I believe I read at GapingVoid sometime last year:

"The most interesting conversation wins."

It says that if you can simply think of and then articulate the best series of cogent thoughts in your chosen area of interest, that the world will beat a path to your (blog)door. Your ideas will spread, you will gain access to (and in fact become a member of) the 'A' list in that particular circle, and you can then leverage this audience in amazing ways.

There are no costs, no barriers, no limits. You against the world in a surprisingly fair fight.

Today we have another case study.

Avinash Kaushik, who has been mentioned here many times before, is leaving his position at Intuit to become an independent consultant and will become some type of ambassador for Google Analytics as his first gig.

In just over nine months on the strength of his ideas, personality, and dedication alone Avinash has risen to the highest levels of respect and visibility in his community, been able to pursue his passion in terms of a career move, and been tapped by the world's most powerful company (as least as far as we're all concerned) to add value and credibility to their product. If that's not a demonstration of the power of blogging, I don't know what is.

(It's also a fact that Avinash has earned a place as lead singer in the "THE ANALYSTS" a little known supergroup featuring Eric Peterson on electric guitar, Jim Sterne on Bass, Bryan Eisenberg on drums, and Jason Burby on keyboards - famous for their amazing late night jam sessions when the other campers have gone to bed at emetrics.)

Congratulations Avinash. We all look forward to your ongoing blogging and contributions to this marketplace.

Posted by Craig Danuloff at 9:25 PM | Permalink

I just found out that two of the very best web analytics minds have new books coming out shortly: Avinash Kaushik and Jason Burby. This is great news for anyone with an interest in analytics or online marketing.

Readers of this or any other web analytics blog know that Avinash has proven himself an extremely clever and insightful fellow, leading online thought in analytics with his trinity theory, 90/10 rule, tips for selecting analytics software and vendors, and many other fine posts.

His book, Web Analytics:An Hour A Day is set for release in early May, and can be pre-ordered at Amazon. Based on the great posts he's produced for his blog over the past year, and what I've heard in seeing Avinash present several times (once live and multiple times through the magic of web video/audio) there is no doubt this will be mandatory reading for anyone who wants to be successful online.

It's also worth noting that Avinash is donating all of his royalties and affiliate earnings from the book to some charities which are important to him. Read more about it here, and use one of our links (which use his Amazon affiliate code) or those on his page to place your order so he gets the extra commission from Amazon for his donation. I know we'll be buying these by the dozen for our staff and clients, and encourage you to do the same.

While checking out the Avinash book, I noticed that Amazon was offering to bundle it with another upcoming book from Jason Burby and Shane Atchison. Actionable Web Analytics: Using Data to Make Smart Business Decisions is also bound to be fantasic. Jason and Shane have been way out in front of the deep use of analytics for many years, building ZAAZ into a market leader and (I think) the largest web analytics agency in the US.

We're all lucky guys like this are taking the time to share their knowledge and experience.

Posted by Craig Danuloff at 12:11 PM | Permalink | Comments (3)

If someone sees only one page of your web site before they decide to leave, did they bounce or did you bounce them?

I don’t think it’s simply a semantic distinction. Talking to marketers about bounce-rates the general impression is usually that ‘these things happen’. And while often the conversation includes a discussion of reducing the bounce rate, there often seems to be an unspoken assumption is the visitors were unqualified or fickle or otherwise to blame.

We’re guilty (at Commerce360) of perpetuating this mindset in the way we often describe the issue to executives new to the term: “It’s as if people came to visit your business, took one step inside the door, then turned around and left.”

I think we’d all be better off if we turned the language around. They didn’t bounce. You bounced them.

They came to your site – on purpose. They clicked a link because they thought you could answer their question, solve their problem, or otherwise help them attain whatever it was they were seeking. In most cases, you created the impression that your site could help, with the ad you ran, the keyword you purchased, or the content that inspired someone to link to your site. Somehow your marketing or your site attracted this person.

But when they arrived, the site didn’t fulfill their expectations. In effect, the site told them to leave, because the information or solution they sought probably wasn’t there, or wouldn’t be complete or accurate, or just that it might be available but would require some work to get at it. Or maybe your site created an impression that just made them not want to do business with you – I ‘get bounced’ from small online stores all the time not because the business is small but because the site makes them ‘seem’ too rinky-dink to be credible.

There are hundreds of possible reasons why people leave so quickly, and obviously some of them are more about the source or the visitor than the site. But the vast majority of bounced visitors are failed opportunities, and when marketers accept these numbers as ‘normal’ they’re doing both the business and the visitors a great disservice.

The people you bounce from your site – especially when stacked up by the hundreds of thousands - should become the motivation and justification for spending the time and money necessary to help your website live up to its (expressed or implied) promises.

Posted by Craig Danuloff at 11:18 PM | Permalink

With this rush of analytically minded visitors we're seeing today based on the link from Avinash, I thought I'd take the opportunity to solicit some information. So you may have thought you were here to read a blog, but I'd like to you to participate in a little survey / project.

Like everyone else who spends time with Analytics software, I'm often asked for advice on selecting an analytics tool, and the virtues of one program vs another. (Usually the 'another' is Google Analytics.) While I have my preferences and reasons why I've made that decision, it still isn't easy to summarize these reasons or convey them to someone else, and of course every unique situation impacts the 'right' decision.

To develop a better and more comprehensive answer to this question, I've recently taken to carrying around a little notebook and writing down the questions that I have for my analytics software. My theory is that over time I'll be able to take this list and then try to answer these questions in a variety of packages and by documenting that exercise I'll have a real world solution to the problem of both recommending (and justifying) certain solutions.

My goal is not to build a 'checkbox' style capabilities list, but rather to provide a few sentences of context as to the possibility, simplicity, caveats, and context of how each package is able to provide 'answers'. In a perfect world, the answers would clump in ways that would make it easy for someone to decide which tool is right for them.

My lists are grouped by subject. Questions about traffic acquisition, questions about visitors, questions about the site structure and content, questions about the economics of the business. Many in some ways correspond to standard KPIs, but I'm purposely trying to put these into the form of business issues/questions not simple titles of information types.

So if you'd please, comment here with one or three of the questions you want your analytics program to answer. Common ones or vexing problems, everything is welcome.

I'll compile the list, organize it as best I can, and publish it back out. Perhaps even set up a wiki where experts in the various packages can contribute answers and examples. Thanks in advance.

Posted by Craig Danuloff at 3:37 PM | Permalink | Comments (1)

I'm starting to feel a bit like the pre-'99 Susan Lucci, but honored none-the-less to nearly make the 'Top Web Analytics Blogs' list for the second time. The esteemed Mr. Kaushik does much to promote and assist the analytics community, and his blog-roll is no exception.

I can only suggest that you take his advice, link here, and get this blog onto the top 10. It took Ms. Lucci 19 tries to make the big time - so making it anytime before about June 08 will be considered a victory.

Posted by Craig Danuloff at 8:17 AM | Permalink

Our friend Shane Atchison of ZAAZ has added his 2007 web analytics predictions to the mix. In some ways it summarizes and feels like a very good wrap up of the other thoughts we've seen, but Shane adds some very wise observations.

I particularly like these two: "innovation will come from 3rd parties" and "human capital will be the largest issue".

And Shane is a man who knows, so I would take stock when he starts calling winners as he does in noting the progress of Omniture (OMTR) and Visual Sciences.

Now if Shane will just explain to all those of us who live a few years behind him what Proto-Analytics is? Is that the next next big thing?

Posted by Craig Danuloff at 1:38 PM | Permalink

Reading all those predictions got me thinking about some of my own:

1. Google will release a major upgrade to Google Analytics which will really put the heat on the major vendors. This will be a good thing for everyone (except the other non-major vendors) as the competition will force the competitive vendors’ technical and marketing teams to really move their products, messages and support materials forward.
   
   
2. Avinash’s 90/10 rule will become broadly accepted and even the vendors will begin to admit that analytics software is a tool not a solution.
   
   
3. A new metric – the ROWA – will be born. “Return on Web Analytics” will track the ROI of this newly recognized increase in the cost of analytics (software + people). Shouldn’t web analysts be able to document its’ own contributions?
   
   
4. By the end of the year there will be at least two more analytics consultancies in the U.S. (or depts. within larger agencies) with at least 20 people devoted to web analytics. (I think the analytics group at ZaaZ is that large already but don’t’ believe there are any others today. Please comment if I’m wrong.)
   
   
5. By the end of the year at least major one analytics vendor will demonstrate beta software, or announce the upcoming release of, a completely revitalized web analytics application that isn’t pageview-centric but instead starts with the broader goals of the business and its traffic sources, website, content and people (as suggested a few days ago).

In other words, they’ll stop making us twist our heads to fit our questions to their data, and will instead start presenting data (and more) that answers our questions.

Posted by Craig Danuloff at 10:57 AM | Permalink | Comments (1)

A bunch of the smartest web analytics bloggers took part in the New Years tradition of making predictions, and someone on the WebAnalytics Yahoo Group took the time to compile links to all of them. Since I hadn’t quite got around to making mine yet, I’ll comment on those I found the most interesting:

• Mr Kaushik says “2007 will be a banner year for Web Analytics education (books, blogs, seminars, conferences).”
  There is no doubt that more and more people will wake up to the need to understand web analytics, so there will be more books, blogs, and conferences. But I don’t think the real crazy-go-nuts boom will hit until ‘08 or ’09. By then emetrics will be nearly as large as SES was in ‘06. (I hope Jim Sterne is ready).

• Avinash also says that “Pay Per Click (sometimes called SEM) analytics will continue to be important but cede mind-share and control to Search Engine Optimization analytics.” How I wish this were true (and said so just the other day). But the $$ and the development effort is all over the paid search world, and I expect both the explosion of paid search reporting and bid management tools to continue and only a relative trickle of better SEO tracking to emerge.

• Lastly AK suggests “More Analysts and Web Decision Makers will realize the futility of torturing click-stream data and embrace qualitative measurement options”. Some will but not enough. As long as the analytics tools present menu after nested menu full of reports, the majority of users (especially new ones) will look at these reports and think they’re supposed to be able to learn something from them.

• Mr Peterson thinks “Smart people will stop freaking out about how "Web 2.0" is going to be measured and will begin to develop rational and reasonable models for tracking emerging Internet technologies and business models.” Again I agree there will be some progress, but I don’t think the ‘freaking out’ will even slow down, nor will the pace of web 2.0 changes which keep putting pressure on both tools and analysts, and I predict that the ‘How to track web2’ sessions at emetrics ‘07 and ’08 will be standing-room-only.

• Eric also mentions that “those with the right tools to identify click fraud may not particularly like what they find.” The ability to document a significant amount of click-fraud across a wide range of PPC campaigns has the potential to be the most important story of 2007. If there is a ‘web 2.0 bubble’ nothing would deflate it as fast as this. I’m just not sure yet if the nature of this fraud (which I do instinctively believe is very significant) lends itself to being detected given the limitations of the infrastructure on which the whole system rides.

• Mr Angel thinks “agencies will finally emerge as a significant factor in the web analytics space... and there are going to be lots who will screw it up pretty royally.” He’s very right on both of these counts.

• Gary also provides which I think is the most interesting of all the predictions with “The pendulum will swing back – at least a little way – from ASP solutions to internal solutions for web analytics.” I don’t know about that but find it very intriguing.

• The GA Experts say that “Microsoft will release ‘DeepSoft’ or whatever they rename Deepmetrix, and it’s likely to be free.” And like everything MS has done in this space for the last three years, it will have absolutely no impact on anything.

• GA also says “web analytics terminology will simplify and become more business orientated and less technical.” I really don’t think this will happen – jargon doesn’t go away as adoption increases – or we wouldn’t have RSS, trackbacks, SERP reports, CPC or ROAS. The old guys don’t want to give it up and the newbies want to join the club.

UPDATE: I added my own predictions.

Posted by Craig Danuloff at 12:08 AM | Permalink

The page view doesn’t have to die. It just needs some help. One little metric can’t carry the weight of the internet anymore. (Sure hits did it for a while, then stickiness then eyeballs and even clicks, but those were the old days…)

The problem is that we need to understand at least four basic aspects of life online:

• Traffic - The flow of visitors in and out of our sites is critical in a world where dollars are spent on ads and paid search, word of mouth programs and organic search optimization. We need to know where these visitors come from, what they do after they arrive (this visit and in their ‘lifetime’), and then calculate the ROI of our investment in acquiring them.
  
  
• Sites – While we can increasingly syndicate content and distribute widgets, the majority of our time and energy still goes into building and maintaining web sites, and we need to know how those sites and (yes) the web pages that comprise them are being used. What’s new is that we can no longer assume the page is a static element – it can now contain dynamic content, elements syndicated from other locations, AJAX and other levels of interaction and progressive or reactive disclosure, and significant personalization or contextualization. So we need to know not only what pages visitors saw in what order and how many times (and for how long) but also what content was on those pages when they saw them and how they interacted with that content in that environment.
  
  
• Content – Chunks of content live free and independent lives these days. A chunk of text can appear on many different pages of a single site, and/or be syndicated across any number of sites. Widgets and videos enable important interaction at the content level (which we need to know all about) but can simultaneously live in lots of places both on our site and across the web. We need to be able to look at both the interactions with any widgets we’ve imported (the widget owner cannot withhold that data) and the aggregate interactions with any widgets or content we’ve distributed. Ads are just chunks, so this applies to them too.
  
  
• People – The elements and constructs above only exist to serve people, whose money it is all of these web pages are really chasing. Yet the current crop of analytics software and conventional metrics only let us see glimpses of these people in the shadows of clicks and page views. We need to be able to segment users not only by where they come from and what pages they see, but by inferred or explicit personas, psychographics, scenarios, intents, and the type of content they interact with and the way in which they interact. We then need to clearly tie these folks to content, sites, traffic sources, and success events.
  
  

If I were developing a new analytics platform I’d consider the needs of each of these four aspects separately, and design both the data collection and reporting for each based on their unique needs. Current systems start with the click and then the page and try to spread this very thin data across all of them. It doesn’t work and as both the industry and people working in it become more mature and sophisticated this is becoming clearer and therefore less tolerable.

The increasingly loud complaints about the page view are the sound of a market opportunity for the next wave of web analytics products. For all of our sakes I hope someone jumps on this quick.

Posted by Craig Danuloff at 10:13 PM | Permalink

2006 has been the year of the death of the page view. The page view was a simple number that has been used to represent the amount of activity or accomplishment a person executed on a web page. But with new technology like AJAX and widgets, the number of times a person clicks, or the number of times the page reloads or refreshes has become completely irrelevant to how substantial or how successful a user session has been.

To further complicate the matter, a slightly deeper consideration of how people and ‘pages’ interact tells us that some page designs and content structures are simply more efficient at getting things done – so it might take one site four pages to ‘give’ the user what another site can give them in one page. Has a site with a three-page shopping cart accomplished more than a site with a one page shopping cart if they both get the sale?

Bryan Eisenberg first delivered the death notice with the even bolder statement that the ‘Web Page is Dead’. Steve Rubel then focused on the use of page views as the counter for advertising sales & revenue. But this week the concept hits the mainstream, because comScore Media Metrix announced that MySpace overtook Yahoo as the ‘king of the internet’ and Yahoo responded by saying that PageViews are now an irrelevant way to count. Ironically, MySpace had previously been critiqued for having a woefully inefficient design that unnecessarily inflated page-views.

Web analytics users and vendors need to take this death march very seriously. Both are far too page-view-centric and switching over to the new reality is going to take some time.

For vendors, the changing nature of a page impacts not only the centrality of page views but also the role and content of fallout reports, page overlay activity reports, and more. And more important than what’s broken from of the old is what’s missing from the new. How do we measure ‘content chunks’ which only appear when requested and get used in multiple places around the site as well is in syndicated locations? How do we see the path within a page much like we used to see the path between pages? How do components on a page influence each other and desired outcomes?

For users, beyond demanding that the vendors start both working on these solutions and talking to us about them, the most important need is for a new concept to replace the page view as the ‘center’ of activity reporting. Eric Peterson has been talking about the concept of ‘engagement’ and it sounds like the right concept. There are some nasty details to work both in the definitions and implementation, but the discussions and work should begin.

I’d personally like to see something like an engagement point system. One point for showing up, one for staying more than 3 pages, one for watching the entirety of an online video, two for revisiting a product twice in one session, two for putting anything in your cart or signing up for something (email subscription etc.), and on and on. A user defined system that then allows me to look at users and traffic sources and campaigns and page chunks by the segments user points.

It’s just one on-the-fly idea. The farther I get into web analytics the more I think it needs a ground-up re-think. Let’s use the death of the web page to revitalize both the software and the way we use it.

Update: Google's Matt Cutts weighs in on AJAX, Yahoo, and Page Views.

Posted by Craig Danuloff at 9:39 AM | Permalink

We don't really ask that much of our web analytics software. Just tell us what's happening on our websites and what we should do about it. Count what people do, tally it up, create pretty little tables and graphs for us, and then offer a suggestion or two about how things could turn out better tomorrow. Is that too much to ask?

Apparently it is. For some reason to use website analytics software is to be frustrated, confused, and more than a bit unhappy your choice of web analytics software. In fact, website analytics packages are like airlines - the one you use the most is almost certainly the one you hate the most. Unless of course the memories of the one you tried before that haven’t completely faded…

There are plenty of reasons for this. Blame to go all around. So let’s start assigning some:

1. Welcome to my delusion. The web analytics vendors make claims and create impressions that would make a weight-loss marketer proud. Visit their sites, download their whitepapers, get chatted-up at the trade show booths, and you could somehow get the idea that installing any of these packages is going to nearly magically solve your online business problems.
   
   
2. Help is not a specific request. Analytics buyers and even users, in my experience, have a vague and simplistic notion of what they want, generally along the lines of the completely unrealistic requests I used to start this post. They are complicit in the way vendors market and even prioritize features because they ‘ooh’ and ‘ahh’ at all the right times in the demo and walk away believing that they really want/need lists of the most popular pages and the ability to see what percentage of their visitors came from Google.
   
   
3. There is no ‘Brain Surgery for Dummies’. Measuring and understanding the interaction between web pages, site visitors, marketing activities, and the economic attributes of a business is massively complex stuff. Vendors who pretend it’s going to be easy or users who expect it to be are setting themselves up for failure and disappointment.
   
   
4. Rome wasn’t built in a day. Realizing that web analytics isn’t going to be fast or easy allows us to begin accepting that it’s going to be slow and hard. We can consider the phases that a real-world adoption will require and stop to ponder a conceptual framework for what it is we’re really trying to accomplish, while at the same time enjoying the new and broader perspective that our analytics software is just one tool we’ll use along the way. In other words, knowing that this is a tough process which is going to require significant time and energy, we’d be wise to figure out what we really want and exactly how we’re going to get it.
   
   
5. Geez, this is a crappy hammer. Analytics software starts with data and tries to produce information. They pride themselves on the reports they can give you and the formatting and delivery methods and triggers. Marketers need actionable insights and/or understanding and these are rarely provided by reports no matter how well formatted. I believe this is what’s underneath all the dissatisfaction with web analytics software – they’re starting with this relatively modest pile of clickstream data and trying to stretch it and shape it as many ways as they can and we’re thinking about the complexities of software and business systems as well as human behavior and asking questions which they can’t begin to answer.

So to summarize: vendors lie about what they can do, users really don’t know what they want, this is much more complicated than most of us can imagine or are willing to admit, it’s going to be a long hard process to really get anything useful, and no matter how you look at it the current crop of analytics software really kind-of sucks.

/// Lots of thoughts to expand on here, and I hope to in the comments or over the coming days. And yes it's a little tongue-in-cheek vendors, we don't really hate you. But we sure could love you more. :-)

{This post inspired by Robin's report from SES about retailers and their thoughts on analytics software}

Posted by Craig Danuloff at 10:00 AM | Permalink

Does it matter how many pageviews your site serves? Of course not. What matters is how much you sold, how many people signed up, how informed or persuaded they became, or how engaged visitors were (and for how long) in whatever it was the site is offering. Yet “page views” is a common, prominent, and frequently discussed metric in website analytics.

The death of page views is being discussed (here too), giving a good reason to shine a little light on what it tells and doesn’t, and how we can perhaps better use this common metric.

Page views are a popular measure for three reasons:

1. They give some idea of how many steps a visitor took, how engaged they were, in the process or information that your site is providing.
2. It’s something of a universal, and therefore provides a way to compare one site to another – how many page views did your site get last month? How many averages pages per visit?
3. Analytics companies tend to set their fees based on page views.

The argument that page views are flawed cannot be denied. The problem is that a page view is not a consistent and universal measure of progress or anything else. A good site design, well written copy, and clear navigation all lower page view counts. Ajax programming techniques and other java widgets provide functionality that would have required a page refresh, without a page refresh, and thereby also lower page view counts while at the same time increasing what can be accomplished on a ‘page’. RSS syndication takes the content from the page and puts it somewhere else where it is seen - without being counted.

One of the reasons this came up is that someone decided that MySpace has a lousy design which requires way more clicks than necessary, thereby inflating its page view numbers, and then took the trouble to estimate how much lower it would be if the pages were better designed. (see the large drop after Q1, 2006?) And it turns out that if this were the case the $580 million the site changed hands for might have been a substantially lower number.

But assuming you’re not looking to pawn your site off on Rupert Murdoch, or aren’t selling banner ads for your primary source of revenue, what does this mean to your use of the Page Views metric? I think it just suggests that you should think of it as a very relative measure.

Continue to watch it as a quick barometer of activity. Given your current site design and content, page views do provide a sense of how visitors are consuming the site from one day (or week) to the next. You’ll obviously want to know the ‘internals’ (number of visitors and average pages per visit) too.

Force yourself to think about why you’re driving that number of page views (and pages per visitor). Take it to heart that driving down the page view count by improving your design, content, or navigation is a good thing. Calculate and track your average revenue per page view, and try to drive it up by both improving page content and navigation and by eliminating links and paths that your analytics show to be generally unprofitable.

And remember three things:

• Page views are a lousy way to compare disparate sites.
• You don’t sell page views, so there’s no pride in having a lot of them.
• You do pay for page views (with most analytics) so you might as well make each one count.

Posted by Craig Danuloff at 10:19 PM | Permalink | Comments (3)

Apparently China is blocking RSS traffic coming through FeedBurner. And Dave Winer rightly points out that this is a big problem with having RSS centralized through one source.

Feedburner is a cool service that provides analytic reporting on RSS feeds. They started when RSS was a little tiny baby, have innovated consistently, and have been rewarded with huge marketshare and success. But is insane that I or anyone with a serious analytics package still has to use them.

RSS isn't some exotic experiment anymore. It's just about as fundamental to online success as a home page. It's a significant method of communication and client/user interaction. It's directly connected to our web sites. Why is robust RSS analytics support not available in Omniture, HBX, Google Analytics, or IndexTools?? (These are the ones I use and know don't support it, please comment on the others that should be taken to task.).

When I look at how many visitors my site had today, or how many times a page (particularly blog post or other feed-like tidbit) was displayed, wouldn't including RSS numbers make that data far more accurate and useful? Isn't that what these packages are supposed to do?

I'm sure I have no idea how hard or easy adding this support would be, nor do I really have any appreciation for the ten-mile-long feature wish list these developers are wrestling with. I'll also admit that I really don't care. I'll instead point out what is becoming a recurring theme in my analytics posts; these packages have to start to be judged (and start judging themselves) based on how fully they solve real world problems. Analytics data that excludes RSS is a partial solution.

An example of the RSS readers-per-day summary provided by FeedBurner.

Posted by Craig Danuloff at 7:50 PM | Permalink

Spending some more time in Google Analytics has generally been improving my view of the tool, but there are two glaring omissions that would need to be fixed before I could strongly recommend it. Surprisingly, both of these issues have to do with missing links.

The first is referring URLS. Google Analytics has a referring domain report, but the only way to get the full referring URL is to use a rather buried sub-menu "Refering Souce > Analysis Options > Cross Segment Performance >> Content" and even then the referring URL is not a link but rather you’d have to type in the domain and then copy-paste the balance of the URL.

(This workaround was found by JustAddWater.nk who describes it and offers some background and other thoughts.)

Could that be any more frustrating and inconvenient?

The second is inbound search links! Google Analytics has a CPC vs Organic Keyword report which shows keyword that generated inbound traffic by engine, but there appears to be no way to actually click out and see the search engine results page. I commonly do this to see who is ranking above/below and try and get a sense for how to improve ranking on that keyword. With GA it’s necessary to go manually rerun the search, then ‘find’ yourself in the results.

Neither of these functionalities is complex, and I don’t know that I’ve ever seen an analytics package without them. Google Analytics is clearly being enhanced on an ongoing basis – I wonder why these features are left out and where they stand on the ‘to-do’ list?

BTW: ROI Revolution has a great blog with ongoing Google Analytics tips.

Posted by Craig Danuloff at 2:42 PM | Permalink | Comments (1)

The importance of ‘conversion rate’ as a core analytics metric is being debated in blog posts from Avinash Kaushik, Matt Jacobs, and LunaMetrics.

Avinash makes the point that conversion rate as generally defined is a vastly over-rated metric because it focuses only on ‘sales’ and thereby forgoes all users who came with an intent or goal other than buying something. Matt says that conversion rate doesn’t discern the quality of the revenue. And Luna responds by saying that these are both true but somewhat advanced concepts and probably a bit more than many online marketers can wrap their heads around at the moment. (These are obviously very brief summarizations, please read the original posts for the full thrust and subtlety of each writer.)

On the whole I have to side with Luna on this one. Most web sites have a clear primary purpose (sell, sign up, download, etc.) and this is invariably tied to business revenue, and therefore I think that the percentage/ratio of site visitors that ‘convert’ is vitally important and needs to be prominently tracked. We’ve also found that high visibility of the abysmally low conversion rate can really help get apathetic executives engaged, either intellectually or at least in approving resources for site redesign, persuasion architecture, A/B or other landing page tests, additional content, etc. Without this number they thought the site was ‘good enough’, or more realistically just had no way to think about its efficacy.

Having said that, if the business and site can reasonably discern other goals beyond the ‘conversion event’ for certain site visitors (customer service, presale research, etc.), then tracking their success to ‘Task Completion Rates’ makes great sense. The major analytics packages should enable this at a core level, making it possible to use custom-defined funnels as ‘completion rates’ that can be connected to inbound traffic, marketing programs, and other data just as fully as (traditional) conversion rates.

It is undeniable that businesses have all kinds of additional complexities not reflected in a conversion rate. One time vs repeat customers, high margin vs lower margin items, high-ticket and complex purchases which happen only after a long list of intermediate milestones are reached, post-purchase issues and customer service, on and on. Of course we’d like to capture as much of these issues in analytics as possible, to provide clear and actionable reports. But the need for more segmentation and granularity doesn’t minimize the importance of the more base metrics. Luna is entirely right that most online marketers are barely crawling and so asking them to run and do backflips just isn’t reasonable.

Beyond the distinctions in priority that are really the core of this discussion, the issue highlights how far the current crop of analytics tools are from what we all really need. My fellow bloggers are pointing out that the conversion rate metric lacks sufficient context and levels of granularity. If we have web sites to achieve results, and analytics software is supposed to measure performance and report on results, shouldn’t they at least be able to cover the concept of conversion well?

Posted by Craig Danuloff at 11:54 PM | Permalink | Comments (1)

Another bit of must-see-internet-tv. In this one Avinash Kaushik, who is one deep thinker when it comes to Web Analytics, presents his ideas of how analytics should be used and what they can and should provide to web marketers and companies doing business online.

This is not a review of charts and graphs and statistics. It’s as relevant to any online marketer or business manager as it is a challenge to any web analyst. This is forward and aggressive thinking that is ultimately less about web analytics as its now generally practiced and perceived and more about taking the core of how analytics should be practiced and incorporating it into an integrated approach to doing business online.

In fact, not long after it begins Avinash proclaims web analytics as dead. And then he builds a case for its replacement, something he calls Web Insights which attempts to focus on actionable information as opposed to interesting data. He goes on to describe how he executes against this goal at Intuit (where he is Senior Manager for Research and Analytics) and share some very impressive examples.

It will be interesting to see which, if any, of the current crop of analytics vendors can take the challenge Avinash suggests and connect their data (and charts and tables) to a broader set of fundamental business issues. Right now they all seem justifiably buried in wish-lists for more powerful tools on the analytics side, and it’s hard to see anyone really moving towards insight (not that they don’t all already love that word).

Posted by Craig Danuloff at 1:46 AM | Permalink

In the latest post in his Excellent Analytics series, Avinash Kaushik demonstrates how one slight change in the labeling of the data series in a report can transform it from being cold and impersonal into something meaningful with which business managers can 'connect'.

His example is changing the row label for 'Visitors of 3 pages or less' to the more descriptive 'Flirters' which helped the people working on his web site (he works for Intuit) to better envision the situation and more actively focus on potential improvements. (Read the whole post for more details and a better description of the results.)

Two of the dirty-little-secrets of the analytics world are that the distance between interesting and actionable is often great, and that half-a-dozen charts and tables can make anyone's eyes glaze over and yet most report-sets deliver dozens. Avinash wisely sets it as the analysts job to make the data relevant and his choice of the word 'connect' as the goal is perfect.

Bonus Link: Avinash recently interviewed Matt Belkin, VP of Best Practices, at Omniture.

Posted by Craig Danuloff at 11:03 PM | Permalink

Ranking highly on Google for your top keywords is a priority for any search engine optimization effort. But what percentage of your efforts, budgets, and resources should you apply to that task?

Jakob Nielson provides strong anecdotal evidence in his new UseIt column, showing that his very popular website gets by far its largest chunk of traffic from Google, BUT the other sites that drove traffic combine to provide 35% more traffic than Google does.

Similarly, the top 10 search words and phrases account for a full 10% of his traffic, but 83% of the 110,399 terms that did drive traffic were only used once during eight weeks. And these single-time queries accounted for 3X the traffic of the top 10 terms.

The importance of multi-word terms and phrases is also highlighted, by a great chart that shows the distribution of 1-2-3-4-5 word searches. Short (one and two word) searches dominate the most active searches, but 'long tail' is filled with 3, 4, and 5 word phrases that in aggregate generate an enormous amount of traffic that would be missed if the site is not text-rich and at least somewhat optimized to rank for longer terms.

Look at your on analytics history to see if your results match those of UseIt.com. The lessons from his numbers are clear:

• Don't think about optimizing your site for just a small number of terms. That isn't how people search and it isn't where all of your traffic opportunities lie.

• Use extensive and expressive language to describe the subjects you cover. Searches will find ways to use even more varied language, and you have no chance to capture them if you stick with narrow 'approved' terminology lists.

• Work hard to rank well in Google, but also work hard to gain visibility and links at other search engines and at all kinds of other sites too.

I strongly recommend reading the entire column for more interesting and instructive information.

[via inside analytics]

Posted by Craig Danuloff at 3:05 PM | Permalink

Today is Web Analytics Weds, which means that there are get-togethers for the analytically-minded in cities all over the globe (Analytics Weds List). I'll be trying to attend the New York City meeting tonight, but coincidentally (I hope it wasn't intentional) the first Omniture user group meeting is also tonight in NYC (W Hotel in Union Square, 6pm) so not sure yet how I'll allocate my time.

We've been Omniture SiteCatalyst users for about two years, and prefer it over the other packages because of the level of customization it provides in terms of both data collection and reporting. We just moved our own site and this blog onto SiteCatalyst yesterday, after previously using Google Analytics for about six months, and IndexTools for about 18 months before that. I still use Google Analytics on 3 other sites, and still have other personal IndexTools accounts which I've used for years.

My impressions of Google Analytics have always been pretty poor. I had some experience with Urchin (the company and technology that became Google Analytics) and found it servicable if not impressive, but am constantly confused by the Google Analytics interface - I could create dozens of posts for thisisbroken.com on the interface foibles. In addition, it always seems like I can't get there from here - find basic data that I'm used to getting from an analytics package. Like if I view the referer list I see domains but not the specific URLs which I'd like to follow back to see the inbound links. Or if I look at Daily Visitors I can't figure out how to see visitors by the hour. I could go on and on.

So I was surprised yesterday to find two things. First, the Unofficial Google Analytics blog, which in tone and content sure makes GA sound impressive, and this post from Aninash Kaushik that basically says if the right person use GA properly you don't need higher end packages (like Omniture).

At the moment, I'm still not convinced because I've spent a pretty fair amount of time trying to like GA and am just entirely unimpressed. But I'm going to spend some time reading and playing and see if these two can convince me or if I can write an authoritative contrarian post as to why the cost and features of Omniture are a good investment.

Whatever the package is, it's clear that web site analytics needs a lot more time and attention from online marketers. Nearly every client and prospect we talk to is under-utilizing the tools and capabilities, which means making decisions on less than all the available information. I was glad to find the list of analytics bloggers on Eric Peterson's site (although his list leaves off the great Matt Belkin) and have added them all to my RSS reader and look forward to reading (and writing) more on this subject in the very near future.

Posted by Craig Danuloff at 10:34 AM | Permalink | Comments (2)

There's always danger in accepting things as they are, especially in the world of charts and graphs. The data can easily be made to look like something, even if it isn't exactly (or even remotely) what it appears. On a web site the number of variables that impact most situations are numerous, and yet web analytics software generally delivers relatively simple reports that attempt to boils the situation down to just two or three variables. This many of your visitors converted into sales. That many people who saw the ad clicked on it. Very often that information should only be taken as a clue - an indication of something much larger and more complex.

Omniture's Matt Belkin challenges one of the cornerstone metrics in website analytics, in a series of recent posts, where he argues rather persuasively that 'unique visitors' should be passed over in favor of the more accurate and perhaps relevant 'visit' metric. Even better, Matt documents 15 reasons why the Unique Visitor metric is untrustworthy - cool stuff coming from the head of the Best Practices Group at the largest Analytics vendor.

It isn't that there is (or would be) anything wrong with unique visitor analysis, as Matt explains, it's just that there is no good way to accurately get that number. Despite that little fact, vendors (including Omniture) deliver oodles o'reports as if they could. But even more interesting, I think, is Matt's contention that visits represent opportunities. That's value-added thinking. Every time someone returns, they are trying to learn or accomplish or buy something. It doesn't really matter if they've been here before or not, right now you've got a chance to persuade and convert them.

Taking those visit numbers apart - figuring out how many of your chances were converted into not only final sales but also 'micro-conversions' as Matt suggests, improves the texture of your information dramatically.

Posted by Craig Danuloff at 1:02 AM | Permalink