How to overcome deceitful data


Tweet about this on TwitterShare on FacebookShare on LinkedIn

goat with its head in the fence

Lara Lee and Daniel Sobol recently shared an extremely considered and thought provoking article explaining what data can’t tell you about your customers.

They discussed some examples where data analysis either led analysts to draw false conclusions on consumer motivations or was simply unable to explain ‘why’ consumers acted the way they did.

Many other journalists and data scientists have reported on the misinterpretation of data. One example you might look back on now with a bit of a chuckle is the Forbes article Predicting that Romney would crush Obama in the last election based on poll results. Perhaps it’s unfair to single out this article. It is, after all, just one of an infinite number of miscalculations when it comes to data analysis.

I was recently discussing motivations behind the consumption of marketing materials with one of my colleagues and I explained one specific scenario which I believe epitomizes this notion that data alone cannot always explain our motivations or behavior:

I have been a subscriber to retailer A’s newsletter for many years now. I’m also a religious consumer of the content they share within these. I’m a frequent opener and more often than not will click through to whatever promotional landing page this retailer has shared with me. I’ve also navigated through to the shopping cart numerous times.

So retailer A wouldn’t be too far off the mark in classifying me as an Engaged User or noting my purchase intent, no? Perhaps they have concluded that I use their online resources for research and purchase in store. Afterall, why not? I continually consume their content.

In this case retailer A would actually be very far off the mark.

See, I am a loyal retailer B customer. I probably have a relatively high lifetime value for retailer B and purchase from their stores or online at least a couple of times annually.

Retailer A could definitely be forgiven for misinterpreting my behavior and motivations. I have given them plenty of reason to do so. While I should perhaps be considered an Engaged User as I do interact with their content frequently, my purchase intent is extremely low.

You might be asking, then why do you continue to engage with Retailer A? It’s simple. Although I’m a loyal B customer, I (somewhat vainly) do need reassurance that my loyalty is not misfounded. I often compare my brand to that of its competitors to reinforce my ideas.

Maybe you think this sounds strange, or you might have had an anecdotal similar situation. In any case, that’s totally irrelevant. The long and short of it is that data alone would have likely misled analysts in this scenario.

How can we overcome such pitfalls of data analysis?

1. Be extremely careful when talking about causality!

Correlation as we all know does not necessarily mean causality. Before even thinking of dropping this word, you may wish to consider the actions below.

2. Talk to your customers

This can range from a quarterly project with a leader in the field, to a simple conversation. Surveys, no matter of what size can open your eyes to the reasons why consumers behave the way they do, and provide you with actionable feedback to improve your offering and increase revenue.

3. Share ideas internally

We once saw an example where a team of US-based analysts relatively quickly saw a fall in website traffic in Sweden. They spent a couple of hours trying to analyse the reason for the drop. They ran numerous website performance tests and contacted IT to check their tagging.

After wasting a significant portion of their day, a Swedish colleague from a different department who walked by and overheard the conversation casually asked if the drop was related to a specific regional holiday that day.

Leverage internal resources as much as possible by creating conversation around your data to diversify the way in which you think. Not only could it potentially save you a lot of time but help you find creative new ideas and insights into the minds of potential consumers.

How do you make sure you aren’t misinterpreting your customer behavior? What are your safeguards against acting on poor data-based conclusions?

Tweet about this on TwitterShare on FacebookShare on LinkedIn

Megan Wilcock

VP of Business Development for Sweetspot. Responsible for strategic brand development, marketing and business development. BA/BComm graduate from the University of Melbourne. My passion lies in finding creative solutions and encouraging collaboration.

Add a comment

Try Sweetspot today!

Not Another Dashboard.