Research and Reports

13 The exception to this is the first discovery: content placement. This data comes from three years of observations. Such a large time span was important to find a sufficiently large number of items to draw inference from. We looked at over 2,500 items that got various placements over this time period for this analysis. 83 APPENDIX METHODOLOGY us one way to use sufficiently convergent data to draw con- clusions. 13 In quite a few of the analyses, we look at "views per item" as a measure of effectiveness and success. There are a few rea- sons for this. Most importantly, dividing by the item inventory count is a good way to control for differences in content mar- keting size. Larger companies tend to have larger content inventories, and it is difficult to make comparisons between larger and smaller web pages. When we normalize content marketing outcomes by the item inventory, we make con- tent marketing efforts compa- rable, large and small. Further- more, it is possible to think of this as a measure of content marketing return on invest- ment: Your content inventory is your investment, and the ex- posure to your audience is the return that you get out of this investment. As such "views per item" is a good way to quantify how much you are getting out of your existing content. To keep our visuals more ac- cessible for a larger audience, we stuck to presenting aver- ages, and we chose to keep standard deviations out of our visuals. However, internally, we observed other distribution parameters such as the me- dian and the percentiles, and we checked for a reasonable amount of statistical signifi- cance. We share with you the results that passed a reason- able test of skepticism.

• Cover