If someone uses terms like Big Data, Machine Learning, or Cloud Computing in the context of optimizing Google AdWords accounts, they really must know their stuff. Either that, or they are trying to wow everyone with popular terms from media to impress everyone how trendy and cool they are ;) Next time someone pitches Big Data to you within your AdWords program, here is something you should know...


Big Data (http://en.wikipedia.org/wiki/Big_data) is defined as data sets so big that normal data base management tools or traditional data processing applications are not adequate. The largest AdWords accounts we have seen have had millions of keywords in them with beyond complex setups, and they were fine using Excel spreadsheets managing their account before we took over, and we were just fine using open source database tools. Reality: The problem with AdWords accounts is not too MUCH data; it is too LITTLE data.

The big lie about Big Data and AdWords

Too little data is a problem, because while you may have 100,000 keywords in your account, each one of them has a lot of variables attached to help you gain a better understanding of how it is behaving in terms of the following:

  • By match type
  • Day of the week
  • Hour of the day
  • Geo areas
  • By device

Depending on how you configure and set up your AdWords account, you can capture all of this information in a way that lets you act on it easily. These variables do not stand alone; they work together (for example, exact match clicks in a certain geo area between 14:00 and 15:00 for a specific keyword) -- and here starts the challenge. For you to DO something with this information (vs. a broad claim of doing hourly bidding), there needs to be a difference in conversion rates and/or amount in the shopping cart to justify bidding higher or lower than during other hours of the day. To make this decision with a certain degree of confidence and to make a positive change, you need a certain amount data to support it.

The vast majority of advertisers will have very few keywords with this type of information. If you act on it for only the keywords/ad groups where you have enough data, it will likely improve performance. If you act on it for all of the keywords/ad groups, there is a high probability that it will fail, because you make random bidding decisions without the support of data. There may be an exception, but most online advertisers are not facing a Big Data problem -- but instead are facing a not-enough-data problem.

Big Data is a hot topic these days, but the concepts and technologies behind it are irrelevant for AdWords programs.