If you have a few years of data from your PPC ad campaigns, you can develop a prediction model that is categorical. Such as; which keywords are profitable or have the potential of profit each month?

Let’s say your business has over 300 keywords, and you want to create a calendar based on when to use certain keywords based on certain conditions.

A model could be developed to classify either yes or no for each keyword with in a time series.

-All keywords that have consumers who purchased AND a 20% profit margin window.

-Keywords with Add to Cart Behavior

-Keywords with Blog Behavior that converted with in re-target windows of 30 days AND a 20% profit margin.

etc…

Whatever you would need to classify profitability custom to your business should be defined within your classification model.

Out of the 300 keywords we have categorized yes or no based on the above conditions by month for each keyword.

Some keywords may answer as yes for certain months and no for others.

Prior to this point you could develop calculations to show over the years if the keyword has fluctuated in performance. If for example a keyword shows 5% decline every year you could factor that in that may classify a keyword as “no”.

What are your thoughts?