Mobtown Labs

Image via Wikipedia

I got into a discussion with @durana today about doing some simple calculations on real-time display ad data.  I pointed out to him that if we are looking at a real-time flow of data coming in of say display ad impressions and clicks and page views and we were to calculate a click-thru rate or conversion rate that the data would look really messy and lack integrity.

My fear was that a media buyer is used to looking at daily conversion rates and click thru rates and we are watching a 24 or 48 hour build of data flow in before our eyes, then rates could be inflated or deflated.  He then pointed out that in fact this would be more accurate.

Now, I know we need to consider re-set cookies and new cookies and all of that but….I mean do you really?  If you are truly looking at real-time data flow in the door and you have a real-time bidding engine to take advantage of that and you can look at users flow through specific sites and audiences from Impression, to Click, and/or to page view, and then to sale and cascade from audience to audience…….then you don’t need to do as much sampling and data validating.  You just need to know the cost that you are incurring in real-time and how much that would go up or down based on the click-thru rate and conversion rate that you are looking at and estimate the impact. 

Also, if you are looking at the effect of your change in real-time then you can make up for any inaccuracies so quickly that it doesn’t matter in the first place considering accuracy is only valuable if time passes between when you take action and receive the data on the outcome of your action.  But you are acting in real-time and can get the results in real-time, then why so accurate???