Notes from ad:tech San Francisco

Finishing up and closing down after some successful days at ad:tech San Francisco – and with this event being one of the biggest shows in the interactive world and the SF event is doubtless one of the premier locations for the ad:tech guys – it did not disappoint. However; the sessions seemed a bit “thin” on content and quality – perhaps because I compare it to Emetrics speakers, who are usually very deep and highly qualified. BUT there was one session which I think had a very valuable question:

“How to attribute conversions to campaigns?”

Most Analytics Vendors and Analysts works with a concept of either:

• Direct Source Attribution
• Original Source Attribution

Where you will see that most SEM (Search Engine Marketing) companies use an “Intelligent” attribution model, whereas if the visit comes from a paid campaign, the direct source is used and if the visit is organic they will trace it back to the original paid campaign source if any. This is of course all cool if you are in the SEM business – BUT in all fairness, this is not correct and attributes far too much value to your PPC campaign efforts. There have been a lot of research done, reliable as well, which clearly indicates that the key points of influence is not necessarily the searched for keyword only – but a mixture of touch points before the final clickthrough from Banners to Blogs to ...

But how to attribute then?

I think that is a post by itself :-) – so let me hit submit and get started on my journey back to Europe. Yes! It is a Journey getting from SF to Budapest. Cheers.

Beyond basic Cost Per Action (CPA) measurement – looking towards CPnA (Post 1/2)

Cost Per Action (CPA) can be defined within the following capacity; “An advertiser that only pays for an advertisement when an action has occurred. An action can be a product being purchased, a form being filled, etc. The desired action to be performed is determined by the advertiser.” – and thus calculated as follows:

CPA = Total Cost of Advertisement / Number of Actions

As an example; let me set up a simple Custom Report to show us the result from an IndexTools retailer, where the Action is “Sale” and the campaign cost is from paid search:

Where you see that on 8th April we have a SALE CPA = $12.27 ($3,117.35 / 254)

As another example; setting up the same simple Custom Report to show us the results from the same IndexTools retailer, but where the Action is “Email a Friend” and the campaign cost is from the same paid search campaign:



Where you see that on 8th April we have a Email a Friend CPA = $135.53 ($3,117.35 / 23)

This is all good and well when looking at basic Cost Per Action (CPA) measurement. However; we cannot expect that:

• Retailers only have “SALE” as a successful action for CPA measurement
• Retailers have single independent actions
• Campaign clicks result in only one positive action

We must expect that most retailers have multiple valuable actions that can all occur within the same campaign visit. That said – the "picture" looks different if I show both at the same time:



Where you notice that the total campaign cost on 8th April (rightfully so) stays at $3,117.35 – BUT more exciting is that we still have:

• SALE CPA = $12.27 ($3,117.35 / 254)
• Email a Friend CPA = $135.53 ($3,117.35 / 23)

I think you see the problem. That I could add in the other 25 something actions that this retailer measures - and in the end the total cost of all CPA combined would be greater than the total campaign cost. Not even that, in sessions where we have multiple positive valuable actions, the CPA for the individual positive actions is not decreased/changed.

I call this challenge: Measurement of CPnA

This is crucial in Affiliate Marketing – which is ALL about Conversion Tracking (Action tracking) – so now comes the question about how to solve or deal with this challenge. Which is now ever more important with the launch of e.g. Google Pay Per Action (PPA), which essentially is a standard affiliate Cost Per Action (CPA) marketing program. I will deal with this is my next post: “Beyond basic Cost Per Action (CPA) measurement – looking towards CPnA (Post 2/2)”

I would love to hear what some of you guys are doing about this fun challenge? – even if you just discard this fact and do not think it holds the same weight as I.

Emetrics Summit Düsseldorf.

Just finished a fantastic set of authentic Sushi at an all Japanese speaking restaurant – not than I can complain about the good food at the Emetrics Summit Düsseldorf, which was super. However; this German version of the summit is still somehow limited by, the relative to the other summits, low attendee number.

On the fun side - I talked to Aurélie Pols from OX2 this morning’ish and after consulting Matthew I think I can conclude that we will be the only 2 persons, who will attend all 4 Emetrics in 2007.

Jim!! – bring out the medals, because Aurélie and I have to be the most “well educated people” out there. :-)

Top 10 Web Analytics Blogs

Avinash did an April update to his super Top 10 Web Analytics Blogs – which is the one I use for reference, making sure I read the ones I am supposed to read :-)

The list goes like this and you will notice that I am on his recommended list – Juuuhuu, thank you very much Avinash!

# 1: Occam’s Razor by Avinash Kaushik
# 2: Web Metrics Guru by Marshall Sponder
# 3: Google Analytics Blog by Jeff Gills
# 4: Web Analytics World by Manoj Jasara
# 5: Eric T. Peterson’s Analytics Weblog by Eric Peterson
# 6: Increasing your website’s conversion rate by Robbin Steif
# 7: Unofficial Google Analytics Blog by Michael Harrison
# 8: Lies, Damned Lies… by Ian Thomas
# 9: WebAnalytics.be Blog by Aurélie Pols
# 10: Web Analysis, Behavioral Targeting and Advertising by Anil Batra

Avinash’s Personal Recommendations:
# 1 Web Analytics & Affiliate Marketing blog by Dennis R. Mortensen
# 2 Visioactive by Ian S. Houston

I read every single one of them and if I should add anything to it, that would be that you read Gary Angel’s blog as well.

...And now on to a real post after three softer ones. ..Titled: "Beyond basic Cost Per Action (CPA) measurement – looking at CPnA" - more about this later.

Comments from Search Engine Strategies New York 2007

Just finishing up and closing down after four great days at Search Engine Strategies – New York ... Waiting to go to the Airport in Newark (how did I end up there?).

I think my one (actually two) takeaways from this show would be that:

• Use of Analytics to perform in a very competitive Affiliate Marketing market is crucial!
• Use of affiliated advertising on a web property should not begin before after hitting the 1000 unique visitors a day

I could not agree more and I am happy to see this confirmed by a super affiliate like Jeremy Schoemaker a.k.a “ShoeMoney” – especially his strong emphasis on the use of Analytics – which also verify the need of a blog like this; with a subject: “Dennis R. Mortensen's blog about how to increase publisher revenue through analytics!” :-)

Bryan Eisenberg’s did his always exciting Webmaster Radio Blog Buzz show, live at SES. I participated with this group of sharp analytics focused people that Bryan introduced as the "guru's of the space":

• Avinash Kaushik
• Gary Angel
• Marshall Sponder
• Dennis R. Mortensen
  

DOWNLOAD/PLAY: Blog Buzz SES NY (MP3)

Duration: 10.28 min. File size: 10MB

...And I am thrilled that I got to tell Gary Angel that there indeed is people (ME to begin with) that read his bright posts in detail – however long and advanced they are. They are a perfect fit for my one hour commute. :-)

Equally nice to meet up with Marshall Sponder in person, after reading I do not know how many hundreds of his posts.

Back to Budapest and off to Emetrics in Düsseldorf.

Emetrics London Video Interview: Dennis R. Mortensen

I had the chance to sit down with David White from weboptimiser and comment on Emetrics and the Web Analytics industry general.

David’s comments:
“Dennis from IndexTools explains why you should visit the next Emetrics Summit, He tells us what is in it for him, what is in it for you, why visitors enjoy the event and gives us one or two of his famous top tips.”

Length: 3.42 min.
Link: Dennis from Indextools (Google Video)
Entertainment level: high :-)



I think David has done some fantastic editing...

Visits vs. Unique Visitors vs. Authenticated Users

I read an interesting post (Authenticated User IDs vs. Unique Visitors) on the Web Analytics Forum on Yahoo today, that I would like to comment on – the post goes like this:

"Hey Everyone, I have a challenge I hope you could help me out with. For one of our sites, we pass the Authenticated User-ID, e.g. aberlinger, via sProp1 in Omniture. EVERY (yes every) user must log-in access this site. Check out the stats below for Authenticated ID's vs. Monthly Unique Visitors for March 2007:

30,710 unique visitors
19,708 authenticated user ID's

How is this possible?

I realize people can remove cookies, but certainly not 30%! I also understand that "Unique Visitors" and sProp data are different reports. However, you there is nothing more unique than our User-ID's. Has anyone else out there dealt with this issue regardless of the tool they use? I would appreciate the forums input.Cheers & Happy Holiday, Adam"

The answer to the first question: “have anyone else dealt with this issue?” – would be; I think everybody deal with this if they try to compare their visitor metrics, but I do not see this as an error, merely a missing agreement between web analyst and analytics vendor on what to expect from the following metrics:

• Visit
• Unique Visitors
• Authenticated Unique Visitors

Doing a similar metric check on a IndexTools client, where I filtered out all other activity than “signed-in”, we get the following for April 1st:



678,316 - Visits on April 1st (100% of all visits)
451,219 - Daily Unique Visitors on April 1st (66% of all visits)
418,764 - Authenticated Unique Visitors on April 1st (61% if all visits)
(Cookie’s Disabled: 1.32%)

This shows that the Authenticated Unique Visitors is 7.2% lower than the Unique Visitors (and have in mind that this percentage is of course higher should we look at a longer period) – let’s have a look at March, which is the period Adam looked at:



19,142,005 - Visits in March (100% of all visits)
12,644,738 - Daily Unique Visitors in March (66% of all visits)
7,860,645 - Monthly Unique Visitors in March (41% of all visits)
6,524,020 - Authenticated Unique Visitors in March (34% of all visits)
(Cookie’s Disabled: 1.19%)

Which shows that 17% of the unique visitors in March were NOT truly unique, agreeing for a second that a "sign-in" is as close as we can get to a true unique visitor.

Dealing with the second question: “30,710 unique visitors vs. 19,708 authenticated user ID's - How is this possible?”

Here Adam has 36% of his unique visitors in March that were NOT truly unique. Looking at the above example (and I am not concluding that this is a viable benchmark, but I think it is a good indication that Adam’s numbers could be correct). Then it is the definition of the metrics that are the issue. The direct answer is that, this is possible because:

1. User Reject Cookies
2. User Delete Cookies
3. User login from different computers (Home, Work, Hotels etc.)
4. Multiple Users use the same computer with different sign-in’s
5. ...

CONCLUSION
When the Analytics Industry - vendors like me :-) use the term: Unique Visitors, we added a bit of sugar and marketing to it, because Unique Visitors is more accurately described as “Unique Computers”. With that said, it becomes unfair to compare unique visitor and authenticated visitor metrics as they simply do not represent the same aggregated total.

And amidst all this uncertainty (about the definition of the Unique Visitor metric) beware about the vendor that says they can provide you with a even more accurate unique visitor sum than your current vendor.. because that is utterly nonsense. I wish I could tell you that, by shifting from Omniture to IndexTools I would be able to present you a number twice as accurate, but I cannot, and my metrics are very likely going to be no worse or no better than Omniture’s.

None of the presented metrics are flawed or inaccurate as such – they just represent different visit totals and perhaps it is time we rename the “unique visitor” metric to something more realistic :-)

Personally; I am a bigger fan of the Visit metric. But that’s me.

Have a great Easter!

My comments to Emetrics London 2007

I attended Emetrics in London last week (..and just returned to the office) - and must say that this was definitely worthwhile. I think the biggest take-away from this show was the presentation by Vincent Kermorgant from Nokia, titled: “Implementing Web Analytics the Nokia way”.



What I liked about this presentation, was that it did not focus on the web analytics tool at all, it focussed entirely on the process of deploying and maintaining web analytics within a larger corporation. BUT there are a lot of good ideas that can be used at any level. Find Vincent’s whitepaper here: Implementing Web Analytics the Nokia Way: a Customer’s Methodology (PDF). If you do not have the time to read it, you will still be a lot wiser, just skimming the above picture. It is really good to see that focus turned towards processes and FAR away from features.

Friends and Contacts from the show:
As always; VERY nice meeting you Avinash – sorry for not being able to make it to the Dinner. I was happy to get the opportunity to meet up with the smart people from ox2 (Rene Dechamps and Siegert Dierickx) and I will see you guys (.. and Aurélie) again in San Francisco shortly. Jesper Lindhardt, congrats on the NEW Omniture business cards :-)... And Simon Gibbard shifted to VisualSciences, good to see you and congrats on your new business cards as well.. Sorry for bashing “you” on Price on the vendor panel. He he. Greg Kelton, Congrats on the GM role for Optimost!!

Finally... for those of you attending the US event - see you at Emetrics San Francisco in May.