Online Stat of the Year?

Marketing Research, Web Analytics, ,

Over on the Rimm-Kaufman Group blog was a report on what Forrester’s Carrie Johnson had to say at Shop.org’s Marketing Workshop. There are quite a few interesting tidbits, but here’s the pair that blew me away:

Correlation between Google Gross US Revenues to US E-Commerce Growth: .96.

Correlation with Yahoo Display Ad Sales and US E-Commerce Growth: -.04

Now, I understand that Correlation does not imply Causation but at some level when you get directional spreads like this you have to sit up and take notice.

One explanation is this: e-Commerce sites do not buy any Display to speak of, but we know that’s not true – don’t we?

Other questions:

1. Another conclusion would be Yahoo matters very little to e-commerce activity. Sure, less than Google, but to this degree? If in fact Display enhances Search performance, you would think Yahoo would have more of an effect. Perhaps folks see Display on Yahoo and then Search on Google? Wouldn’t that be a trip…

That scenario would really provide a whole new twist on the measurement of view-throughs.

2. Google gross rev’s include AdSense, of course. So we’re not really comparing PPC to Display here, though one could argue AdSense is more targeted than Display. So what we are discussing here is the relevance of ads, not PPC versus Display.

3. Does Yahoo Display include Travel ads triggered by selection of Location? Auto ads triggered by selection of Model? Etc. Etc. You could argue those ads are really “Search” if you look at it from a behavioral (customer) perspective.

Sure would like to find the source on this, and see what we are actually talking about here.

Other questions you would ask / data you need to make a judgment on this? How about wild speculations on what this data means, if anything?

eMetrics Toronto

DataBase Marketing

I’m on my way back from the show, lots of enthusiastic and bright folks up there in Toronto.  Damn cold too – at least for this Florida-based guy.  But I managed to hang in there with a coat I bought for a trip to Edmonton with Bryan Eisenberg back in January of 2001 or so it wasn’t nearly as cold in Toronto as it was for that trip though!

Anyway, great show as usual and thanks to Jim S. and Andrea H. for inviting me to participate in so many ways.  Even more thanks are probably due to Matt and Fanny because they are behind the curtain and are doing all the work! 

Wizards they are.

The biggest news from the show I was exposed to was the awarding of a patent to Joseph Carrabis of NextStage Evolution for his technology enabling any “programmable device” – phone, web site, car, etc. – to look at user behavior and know “how that person is thinking”. This could be a good thing.

Not sure exactly where this goes, but clearly if this idea works out, services like Touch Clarity will be considered “dumb” by comparison.  I mean, parsing a referrer is one thing, knowing how I am thinking is quite another.  “How” meaning (to simplify) what cranial modes the user is in as they are interacting with the device. 

I won’t attempt further explanation, see here and pay particular attention to the C, B/e, M matrix thoughts.

I’ve known Joseph for some time now and he’s an incredibly bright guy – and very entertaining to talk (OK, argue!) with.  Now that he has a patent, I’m sure we will start to hear more details on what is required for implementation and (some) idea of how it works.

Any comments on this development?

On Monday, I taught the DAA BaseCamp Intro Level Course to a pack of 40 fine people who wanted to learn more about the basics of web analytics.  To any of those people who are now blog subs, thanks for the treat of being your instructor and I very much appreciate your kind words on the “user experience”.

Then as everybody else rolled in, there was of course the opportunity to have spirited (in several ways) discussions about everything web analytics until 2 AM.  Great to see Braden, Hosam, Judah and Raquel from the DAA Education Committee team as well as all the usual show management and speaker suspects. 

Tuesday I presented a session on making web analytics invaluable to C-Level people.  One of the issues web analytics faces with C-Level folks is they don’t really care much about history; they get paid for thinking about the future.  So unless you change your mindset and start doing some prediction – like the BI folks do – then it’s likely the C-Level folks will continue to be under-whelmed by your reports.

What does this kind of future-oriented reporting look like? 

Well, for those of you with some experience under your belt, it looks like this, though I dumbed it down a bit for a general audience presentation to a simpler model closer to this.

Comments on any these ideas? If you were at the Toronto show, what did you think of experience?

Perfect Google-Click World

DataBase Marketing, Measuring Engagement, Web Analytics

So, what would a universal cookie across both Display and PPC give us?  What could we look forward to, what’s the wish list of the online “ideal marketing world” we could live in when we really understand how Display and PPC interact? 

I’ll give it a start, feel free to add to this wish list…

1.  From a macro web advertising perspective, all available Google-Click “space” is capable of being optimized for performance – whatever your definition of “performance” is.  That means an end to the idea of the space being attached to a pricing model – for any given space, you might see either a PPC ad or a Display ad. 

Hopefully, the advertiser would have some control over this allocation, deciding if / when which pricing models are used in which spaces, similar to the controls over pricing model existing in AdSense today.

I realize running Display units in Search inventory may seem counter-intuitive, but the key is pricing control.  The web desperately needs a more effective way to expose people to ideas they have never heard of in context.  Running random display units across opaque networks is not a particularly good way to do this; running targeted display units based on search history – a more advanced form of behavioral targeting – would do the trick.

Likewise, running PPC ads in tightly segmented display spaces can lead to big payoffs, as it did in the Lab Store example.

2.  A real gift would be some cookie-based sense of where the visitor is in the funnel, probably based on the search phrases they are using. 

“Level 1” would be no prior interest, “Level 2” would be “uses single word searches” on the topic, “Level 3” would be “uses multi-word phrases” on the topic, and so forth.  Visitors in the unknowing or shallow knowledge Levels would be exposed to cheap Display units – both in Search and in Display inventory.  Those expressing active interest (Engaged, if you will) would be exposed to PPC units, again, both in Search and in Display inventory.

You’d need some kind of history control on this, because the data set would probably get too huge.  Say for example, trailing 30 days interest, so you make sure the visitor is still Engaged with the topic.  As active searching on the topic dropped off, you’d kill the data off, because it’s no longer relevant.  If it starts up again, so be it, but a restart is a new profile.

Other ideas from the crowd?

Given cross-site tracking already exists for Display, do you think there will be any Privacy problems if Search History was used to target outside of Search?  People are pretty used to seeing ads based on searches in the search space…why not bring that over to Display?

Do you think Google-Click will give us a “universal cookie”, or will they keep the Display side in the dark for the sake of Brand-oriented folks who only care about impressions?