In our careers as marketers we are often presented with problems that require some kind of statistical analysis.  One of the most frequently-faced issues is that of content or quality ratings.

Let’s say your company produces 5 different widgets.  You ask 100 of your customers to rate these widgets and ask them to rate all five when they do so.  Assuming they all provide feedback, you now have 100 ratings for each widget – it’s easy from here to take an average and identify which of your widgets is the most popular.  Unfortunately, the world never works this way.

In reality, your best customers for widget 1 might not even know widget 5 exists.  Take this into account, and your distribution ends up completely different.  You might have 80 ratings for widget 1, 40 ratings for widget 2, 25 ratings for widget 3, 30 ratings for widget 4, and 5 ratings for widget 5.  You can still take an average and claim you know which widget is the most popular, but your numbers here would be very misleading.

Your averages in this case are not comparable to one another.  If you put them side-by-side, you’re claiming the five ratings of widget 5 each carry more weight than the eighty individual ratings of widget 1.  If you were to add a sixth widget to your offering and just one person weighted it, their single vote would count as much as eighty votes.  It blows your ranking system out of proportion and makes it impossible to convey accurate information to stakeholders and future customers.

Thomas Bayes There is a way to correct this problem, though.  Using theories written by an 18th century statistician, Thomas Bayes, we can actually figure out what the correctly weighted average will be for each widget given the information we know about the entire product offering.  Then you will truly know which of your products is the most popular and can act on that knowledge accordingly.

Over the next few days I’ll explain a little more about how Bayes’ method works.  Then I’ll give you a practical example so you can see it in action.  The goal is to enable you to use Bayes’ method in your company – I guarantee the extra information you gain from knowing exactly where your products sit in relation to one another will be invaluable!