I was invited to spend the day with our Search Engine Marketing agency today (free lunch!), and got to hear about all the latest and greatest going on in the world of Search and on the web in general. I am fascinated by the whole thing, and energizes me to be involved with it. I also love that the whole thing runs on the back of analytics. Optimization, etc happens soley on metrics. But one key problem we’ve run into (and I’m told others have as well) is that most web analytics solutions have a limitation of picking up the full string of tracking codes visitors pick up along the way, making it hard to do true keyword analysis or even full scale marketing effectiveness across different media. Within Omniture for instance, there is a limitation of only picking up 2 tracking codes for linear attribution. Let me give you a scenario…
Visitor types in ‘laptop’ in Google and clicks on the link to go to Lenovo, doesn’t buy anything.
A week later types in ‘thinkpad’ in Google, clicks on link to Lenovo, still doesn’t buy anything.
A week after that visitor types in ‘t61’ in Google, clicks on link to Lenovo, and then buys something costing$1500.
With linear attribution (we have expiration set on 60 days), we can attribute credit to the keywords equally, since they were all vital in the final result, a purchase. The problem is Omniture (and I think other solutions) doesn’t capture the full string of keywords. In the example above, Omniture would only capture the last two, so $750 would be attributed to ‘thinkpad’ and $750 would go to ‘t61’. And the word ‘laptop’ would be left without anything. In the data it would look like it was an underperforming keyword and possibly get its allocation cut since it wasn’t providing any return on investment. However, we know without metrics that we need to buy that keyword as it plays some role in ultimately causing visitors to gravitate towards ‘branded’ keywords such as ‘thinkpad’. In fact, we’ve been told that a visitor usually searches 4.9 times on average before a purchase. So we’re missing 2.9 of the searches!
So what to do about it? As of now there isn’t much out of the box. I am hoping that the bright folks in Orem are dilligently working on a solution as this effects more than Search. I’d like to see the interactions between all online media and see the patterns of media that visitors encounter, and to see the true lifetime value of all of these media. For instance, being able to look at the progression of media (search, banners, email, affiliate, etc) until a success event occurs and beyond.
Ok, that was my stream of consciousness. enjoy the ramble.
Anyone out there with more suggestions for solutions?