Knowing how TV attribution works is first step in the right direction. However, a true analytics ninja’s search doesn’t stop there. As an advertising channel, TV is still the by far the leader in its space. More users are starting to consume TV while surfing the web simultaneously and we as marketers need to be able to measure its impact in real time.
The big questions is, how does this all come together? Well, let’s dive into it!
How do I get stated?
There is one major pre-requirements to get started with it – You should either know some level of coding or be BFF with a colleague who does. Since we need to get the data down to the minute, we have to tap into the API of your web analytics tool. There are a few plug-ins out there that can do this for you, but that would require admin access to the site (a.k.a making more BFFs across the organization).
That being said, here are the three major steps involved in doing this analysis
1) Gather the data logs of accurate airing times of your TV spots
Getting this right is key, as this determines the outcome of your analysis. Media plans tend to have estimates airing time but in reality the actual airing time is slightly off. You MUST have the actual airing time to get this right. Ask your agency partner for this or if you’re on the agency side, ask your media team they would be able to get this to you. It should look something like this
The two key fields highlighted above are the date and time columns, which help you identify the time when the spots ran. Additional information in the form of network, program, time period, creative, etc will provide you ample ways to slice and dice your data. For example, if you see that people watching Jersey Shore are more likely to visit your site, you have get a pretty good picture of who your target audience is 🙂
2) Compile your site metrics
This is where your coding ability comes into the picture or in my case my BFF’s. To explain this, I am going to use google analytics (GA) as my web analytics platform. GA offers a great API library to export any type of data in the coding language of your preference (or your BFF’s). My BFF helped me with this and the code can downloaded from the link below
The code might look daunting, if you are someone like me. All you should worry about is the output which is a bunch of dimensions and metrics needed for the analysis. I used the ones listed below which served my purpose.
dimensions = 'ga:date,ga:hour,ga:minute,ga:deviceCategory', metrics = 'ga:newUsers,ga:sessions,ga:transactions,ga:transactionRevenue,ga:organicSearches,ga:bounces').execute()
3) Aggregate the data & tie it back to the commercial airing
The last step in this three step process is to create two groups of data. We will need to use the commercial airing as the timestamp and create data sets of ‘pre’ and ‘post’ for before and after the commercial airs. The ‘pre’ helps in establishing the baseline for your analysis while ‘post’ helps in identifying the incremental impact due to TV. Once again, my BFF helped me with a python script to aggregate the data.
You can find the script here –> Code for data aggregation
You could use either 5 minute increments or 10 minute increments to define the two groups. Your resulting data set should look something like this.
Voila, you have all that you need for the analysis. It is time to put on your analytics ninja hat and start nerding out.
Closing thoughts: I’ll be honest, this is not easy. If you are able to navigate through the hurdles, the outcome is worth it. I assure you, not everyone in your organization knows their way around this which means it is yours for the taking. Even better, you could save your company some valuable $$ by not opting in for that TV module your boss was considering while making a case for your promotion #WINNING (psst. Click it)