Articles Comments


» Google Adwords, Paid Search in General, PPC Analysis & Case Studies » Understanding Variation in PPC Data | Principles of Variable Data Analysis

Understanding Variation in PPC Data | Principles of Variable Data Analysis


Variable Data in Paid Search Advertising“LOOK EVERYWHERE and you will see variation” is the first sentence in Ellis R. Ott’s 1975 classic Process Quality Control: Troubleshooting and Interpretation of Data. That sentence could not be more valid when considering management and optimization activities of paid search advertising. In the world of quality, process and statistics Ott may not be as well known as Walter Shewart, W. Edwards Deming, Joseph Juran or Donald Wheeler but in my opinion he is equally important.


Ott starts off Process Quality Control stating some general principles recognized by those troubleshooting variability in data. Those principles include:

Rule 1. Don’t expect many people to advance the idea that the problem is their own fault. Rather it is the fault of the raw materials and components, a worn-out machine, or something else beyond their own control. “It’s not my fault!”


Rule 2. Get some data on the problem; don’t spend too much time in the initial planning. (An exception is then data collection requires a long time or is very expensive; very careful planning is then important.)


Rule 3. Always graph your data in some simple way – always.


These principles, or rules, are pretty straightforward and apply as much today as they did back in 1975 – even in the world of paid search marketing.


If you think about it, each keyword, search query, ad creative, competitor and customer adds more and more variables to the mix. The factors are countless but that does not mean managing and working to identify patterns and trends in the data is hopeless. It should be clearly understood that successfully managing variability in paid search means looking at the entire account over longer periods of time. With variables like cookie duration settings and other outside factors, hour to hour and week over week analysis is less effective than a broader analysis of month over month or year over year data. Managing day by day or week by week can inadvertently cause reactions to data that is not true. This false sense of information can open up the door for more variability in performance than should be expected. This is why knowing the tolerances of variability of your paid search accounts on an intimate level should allow you to focus more on the process of expansion and growth an less on flash results.


Once patterns and trends of variability in clicks, cost and conversions are established and you have a sense of the accounts’ takt time, you should be able to report on your findings effortlessly. As Ott states, communicate the analysis in clear and simple graphs or tables. Paid search is complicated enough and flashy charts are not required, however adding context to your reports and dashboards is critical for precise communication. Two of the best resources for analytics and dashboards, Avinash Kaushik and Stephen Few would echo this same basic principle. It is also helpful to understand variability and trends by campaign, ad groups, keywords, categories, styles, etc. as long as the information is actionable. Anything more is considered waste (over production) and should be avoided at all costs.


With the high level of variables in paid search it is essential to have the right enabling processes established in order act quickly on analysis so that performance targets are met. Next time you experience a random spike in traffic, follow the rules above – understand the root cause, collect the data to support your findings and communicate those findings in a way that is simple, effective and actionable.

Filed under: Google Adwords, Paid Search in General, PPC Analysis & Case Studies · Tags: , , , ,

Leave a Reply