As a high performing marketer, data should be at the centre of each decision you take to bring your business forward. You collect data across your activity, analyse it, put some conclusions forward, and base your next step on your observations. But an easy mistake to make when you are analysing a set of data is confusing Correlations and Causations. Without a clear understanding of the two, you could be making important decisions based on the wrong data. Let’s take an example to illustrate the two:
You are the marketing chief for a thriving e-commerce store that sells wine bottles. Business is going well, but you decide to send a 25% limited time offer discount code available only this month via email to your customer base to boost sales. At the end of the promotion period, you review your numbers and see that 60% of customers have claimed your discount code. You give yourself a high five and go brag to your colleagues at the coffee machine about all the orders you think you generated. But was your sending of the coupon the cause of all these purchases, or are the two just correlated?
A Correlation is simply a relationship or mutual connection between two variables. However, a Causation refers to one event being the result of another event. For example, an (extremely basic) example causation would be: clicking on an ad caused me to land on a website. Going back to our example, if you only have a quick look at your sales data, it’s easy to make the assumption that all the people who redeemed your coupon bought because of the discount you offered them. The assumption is that the coupon caused the purchases (causation). Based on this assumption, you may be tempted to blast your audience with regular discount offers every month to boost your sales. But if you dig a bit deeper into the data, you may find that most of the people who claimed your coupon are regular buyers that place at least one order per month. You review their purchase history and quickly find out that the purchases with the coupon are not significantly higher that the purchases you would have generated anyway. In other words, the people who claimed your coupon would have bought no matter what, but since they got a discount, they decided to use it.
The conclusion is that instead of boosting your sales, the coupon simply cut your revenue on many offers by 25%! So without the coupon you thought was bringing you extra business, you would have actually made more money. In order to measure the causal effect of your promotion, you need to use a control group:
By isolating a sub-set of your audience and not exposing it to your promotion, you can then compare the number of purchases the exposed group makes vs. the control group.
Without adequate testing of your data, it is remarkably easy to confuse a correlation with a causation and take entirely wrong strategic decision. As a little illustration to conclude this article, if we simply had a look at these stats without thinking twice, it would be easy to blame Nicolas Cage for pool drownings and margarine consumption for divorces…
If you have any comments or questions, feel free to contact Hyphen now.