Google Analytics: Tracking Online and Offline Worlds
For some time now, Google Analytics have offered the ability to track both website visits and television commercial impressions (views), if the commercial was bought via Google.
The graph has always shown a peak of what multi-channel analytics can look like: a single dashboard to see everything from TV commercials to mailed newsletters influences your website traffic and online sales. But TV Impressions v. Website Visits graph also begs the question:
What’s the correlation between TV Commercials and Website Views?
Understanding correlation and causation -essentially, understanding your media mix – is the holy grail and end goal of Multi-Channel Analytics. It’s one thing to track something and say “Gee golly, I can see my radio audience reach and PPC campaign on one sheet!” and it’s another to solve Wanamaker‘s question: “Half the money I spend on advertising is wasted; the trouble is, I don’t know which half.”
Solving “How does this offline channel affect my website?”
For very easy graphs like the Google Analytics graphs, the equation is easy for any statistician to derive (yes, marketers need to re-read their college stat books):
- Total Website Visits = ~1.90 * TV Impression + 340,566.1 website visits
And hereâ€™s equation applied against my version of the Google Analytics graph:
Not bad, huh? Unfortunately, the real world is different with a lot more variables: multiple marketing channels, geographic differences, seasonality, public relations efforts, multiple cross-channel campaigns and more introduce a host of variables that effect a websites traffic and online sales.
Huge companies like Proctor & Gamble have entire teams dedicated to media mix modeling with some even using something called “agent based modeling” (used by the military) to determine how their massive multi-million dollar and cross-channel marketing campaigns effect sales.
So the question is: how will analytics companies like Omniture and CoreMetrics be able to extend multi-channel marketing tools into solutions that can solve the question:
“What types of marketing campaigns mixes generate the most sales?”
Where did I get the data?
I carefully recreated the television data based on the 4,444,444 impressions over 30 days and created website data that would fit the Google Analytics graph. See below:
How accurate is your analysis?
In real world applications? Probably not very. The equation works if we assume the relationship between the television commercial impressions and website visits is overly simplified: if you get X many commercials impressions, you’ll get Y amount of website visits.