There are many times when I'm wrong ... I go into a project with pre-conceived notions of what I expect to happen, I run a series of queries, and then something unexpected happens.
Our instinct, of course, is to immediately torture the data, trying to slice and dice it in a way that yields the answer we're looking for.
Here's one example - I looked at customers with exactly three purchases in 2013, trying to measure how much customers spent in 2014, based on the number of channels they bought from in 2013, and based on the number of merchandise divisions the customer purchased from in 2013.
I didn't have much faith in the omnichannel thesis (more channels = better future value). I had tremendous faith in the merchandising thesis (more departments = better future value).
What do you see?
Channels add value if the customer only purchases from one merchandise division.
Merchandise divisions add value if the customer only purchases from one channel.
Lowest-value customers historically bought from 1 Channel / 1 Division, or bought from 3 Channels / 3 Divisions.
Lesson: Every business is different. Customers exhibit unique behaviors. You have to analyze your own data, controlling for various factors, and find the truth for your own business. In this case, my hypothesis was wrong - and instead of beating the data into submission, it's just better to acknowledge that, in this case, I'm wrong. And that's ok.
I ran a poll on Twitter late last week: When do you think business will return to normal at your company? 13% = May 25. 20% = July...
It is time to find a few smart individuals in the world of e-mail analytics and data mining! And honestly, what follows is a dataset that y...
Say you manage a paid search program. Last month you spent $100,000 and the following happened. Cost = $100,000. Clicks = 200,000. Co...
Two weeks ago I ran a poll on Twitter, asking if users calculated the profitability of their marketing efforts. 32% said "no"...