Learn from my Web Analytics implementation mistakes – Campaign Attribution

For those that are about to embark on the adventure of implementing a web analytics solution, hopefully you’ll take your time with some of your decisions. I can’t mention enough that you need to take your time and think things out as some of these decisions can definitely impact future decisions and functionality.

The one that still haunts me from my Omniture implementation over 2 years ago is the way I set up campaign attribution. It still actually really bothers me every day. My thinking at the time was that I needed to be fair in how I attributed campaigns and not get stuck in the trap of only giving credit to the last tracking code/tactic as this could cause you to kill campaigns that are upper funnel. Because some campaigns are not going to be the last thing a visitor clicks on before buying I wanted to at least let the tactic get partial credit. To allow for this I set my campaign tracking codes to linear attribution.

Additionally, I wanted to allow for campaigns to have a longevity to them as we had evidence from CNET and other publications that the buying process for a PC buyer takes like 60 to 90 days. So I gave the tracking code a 60 day cookie. If you are keeping score I have it set as a linear attribution with a 60 day expiration window. This sounds ok on the surface and somewhat fair to the various tactics that contribute to an ultimate sale.

But…here is where I went wrong. First, I didn’t look at the buying cycle of people on our site before deciding. Just because your average buyer takes 60 to 90 days to make a decision, doesn’t mean they are doing all that shopping on your site. In fact, most of the research is being done on other sites such as Comparison Shopping Engines, our friend Google, and online review sites like CNET. By time they get to our site a lot of the time has already gone by. So tip #1…understand your buying habits of your customers before setting your expiration cookie. Even if you do get it wrong, within Omniture this is an easy one to remedy as you can change this easily in the Admin settings in about a minute. Say you wanted to change from 60 days to a week, no problem.

The bigger mistake I made was the first thing I mentioned regarding linear attribution. Within Omniture, this isn’t easy to fix. If I changed from linear attribution to something else like most recent tracking code, all the campaign data in your reporting will basically vanish. It doesn’t actually get deleted, but is hidden from your reports as a new database table is essentially created and used for campaigns. If you were to go back to linear after that, your old data would once again become visible though the most recent stuff would then get hidden. I learned that one from another misstep. In fact, I would caution you from ever using linear attribution for that very reason.

So why do I want to change it anyways? Well, to bring in outside data like product cost, cancellations, or even call center stats that you want to line up with your campaign data you can’t do it with a linear attribution as it creates a one to many data relationship. It will only work when revenue or orders are tied to a single tracking code. This is a big problem for me as I want to enrich the data to see ROI and cancellations by campaign.

Also, the thing I was trying to compensate for, meaning giving attribution to upper funnel tactics, can be done with other tools such as campaign stacking and Discover.

Now what? Well, I wish that Omniture had a way to basically copy the data table over from the historical campaigns and place onto the new data table for campaign attribution, but I don’t think they have any out of the box solutions for that. Might be something I have to engage Engineering about which means $. In any case, I don’t want others to fall into the same trap I did, so hopefully I can help someone out there in blogistan by learning from my mistakes. In fact, I could probably write an entire book on this stuff. Hindsight is always 20/20 though, you just keep tweaking it over time.

Advertisements

2 responses to “Learn from my Web Analytics implementation mistakes – Campaign Attribution

  1. Thank you for sharing this. I have recently ran across this issue with the company I am implementing SiteCatalyst for. I am wondering if you can point to some well established solutions with the SC platform.

  2. I wanted to allow for campaigns to have a longevity to them as we had evidence from CNET and other publications that the buying process for a PC buyer takes like 60 to 90 days.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s