I went to Omniture Summit in Salt Lake City, UT last week. If you’re not into marketing and analytics you wouldn’t have had a very good time there (unless you’re a fan of The Killers which was the main act at the conference party) and you’ll definitely find this blog post very boring.
The main keynote speaker was a star only outshined by The Killers: Seth Godin presented nuggets of wisdom from his latest book called Linchpin. Citing Mr Godin: How can you market water better? By saying it’s wetter? He encouraged everyone to think of themselves as creative geniuses; but geniuses in a very pragmatic rather than the nutty professor kind of way (citing him again: Einstein gave this genius thing a bad name). His quest for innovating and making a difference is not novel but I must say he’s an excellent story teller and he does a good job of invigorating one. He did in my case, anyway. I’m half-way through Linchpin in audiobook which is the second best thing if you don’t have a chance to see him live.
Next up Wired founder John Battelle made an interesting point that e-commerce and online marketing wise we’re at “fish with feet” stage in evolution. The days of socially isolated purchases that are tied to desktop or good old shoppe on the corner are over (ie. we’re not relying on swimming only anymore) but we haven’t quite figured out what to do with mobile web, location data and people expressing their intentions in a social fashion. In 5-10 years we’ll do things differently, just that no-one knows how exactly.
Rest of the summit was much more practical and took place in specialized tracks (and this is where it gets really boring if you’re not in online marketing, and really interesting, if you do). So if you haven’t yet then stop reading now, what follows are mostly notes to self.
The sessions I picked revolved around increasing conversion and I can roughly categorize my takeaways into three categories: getting started, making sense of data and recommendations on site.
Views were bipolar on whether to start optimizing near the top or bottom of the funnel. All in all it’s a simple mathematical equation so it shouldn’t matter where you do it, if you improve conversion by 10% somewhere this will increase sales (or some other measure) by 10%. However, the closer you are to completing the purchase the clearer the intent in and thus more valuable any increase is. This presents the case for starting at the bottom of the funnel, a view supported by Robin who responded to my related tweet.
Process wise consensus was that there shouldn’t be a separate process for setting up web testing and related analytics. Rather such optimizing should be build into existing processes for creating and updating web content.
Then it’s good to keep in mind that testing web content doesn’t always increase revenue. Sometimes you want to evolve your design or make the user experience cleaner, without necessarily increasing sales. AB test acts as a hedge that you haven’t lost revenue.
And last but not least – start simple. There are many reasons for that but Jared Spool’s 300 million dollar button case illustrates this best.
Making sense of data
It’s old adage that testing and analytics get you the WHAT, user tests and focus groups get you the WHY and that for best results you should use both.
What I hadn’t thought about before is that there is a ton of data available that greatly helps to come up with hypotheses on which bits of your particular landing page are important and thus should be tested. This can be summarized by sort of “existential questions”:
Where did I come from?
Top external search terms, Top banner ads, Email campaigns pointing to it, Referrers, Geo-segmentation if applicable, Technology of visitors, etc.
Why am I here?
Top entry pages, Bounce rate of top entry pages, ClickMap, Top internal search terms, Onsite promotions on landing pages, Page layout, Navigation, Usability, Above-the-Fold Analysis, etc.
Where am I going?
Next page flow from entry pages, Conversion rate of entry pages, Campaign stacking, Internal Keyword Stacking, Top exit pages, Previous page flow of top exit pages, etc.
(The above was almost word-for-word from a presentation by Brig Graff of Omniture)
Recommendations on site
Recommedations can be quite wide if not random for the users in the exploring more, such as visitors on your front page. If you get closer to the end of the purchase funnel either make recommendations super relevant or lose them completely. Example from Argos: if users come in and search for a specific product based on catalogue ID then recommendations actually decrease conversion.
When you’re creating personalized recommendations to customers, why not tell customers that they’re personalized recommendations, and perhaps even add why. Think Amazon.
For content recommendations timeless entertaining recommendations perform better than breaking-news style news that get updated often.
In summary a good theme to keep in mind is Make it work. Then, make it better. (and hey, this perhaps applies in other parts of life, too, not just online marketing.)
Additional reading: 2010 Trends and Challenges in Web Analytics by Econsultancy.
Reading this back this definitely is the most professional and boring blog post ever. I guess I should write the next post about sexual escapades to maintain some level of balance?