A couple Olives from the content team recently attended a Social Media Breakfast in Minneapolis. During the event, we explored the truth about marketing data alongside other local social media enthusiasts. Speaker, Nick Rosener, shared his statistically driven insights in his talk The Data N̶e̶v̶e̶r̶ Lies: How to Get the Truth from Marketing Data. You can check out his slides below or read on for 5 key takeaways from the presentation.
1. We Face Two Types of Problems
Marketers are problem solvers. We spend our days analyzing our clients’ needs and implementing strategic solutions. Along the way, we encounter two types of problems: marketing problems and statistics problems. Marketing problems are the strategic, creative, tactical challenges we face. We turn to data and analytics to help find solutions. Statistics problems are often rooted in that data.
2. We Don’t Always Know When We Have a Problem
The catch with statistical problems, is that we won’t know when we have a problem. Sometimes, we don’t have enough data to draw truthful insights, but it’s hard to know when our data pool isn’t large enough. We may think a data experiment produces useful conclusions, but if the experiment isn’t designed properly, the results are null. The good news is, we can take steps to ensure we’re drawing reliable conclusions from our marketing data.
3. Ignore Meaningless Data (Sorry, Impressions and CTRs)
If we hope to uncover trustworthy insights, we must draw conclusions from the proper data. Social media impressions and clickthrough rates often look impressive but have very little impact on your ROI. You may do well to ignore these variables when calculating the cost-effectiveness of your Facebook-promoted ad. Instead, you need to look at the cost of the ad, your conversion rate, and how that translates to your gross profit margin.
4. Don’t Rely on “When To Post” studies
Designing our own marketing studies takes time and it’s easy to want to turn to outside studies for quick strategy and tactic tips. Social media marketers often consult “when to post” studies to optimize their own posts. Do not adjust your strategies based solely on those results. Why? The audience studied likely doesn’t reflect your unique audience, and past events do not necessarily reflect current events…
5. Correlation Does Not Equal Causation
…and correlation does not equal causation. Keep this idea in mind when analyzing results of “when to post” studies, but also your own retrospective marketing experiments. It may look like your data is sending you a message but don’t make a leap to causation just to make sense of the data. You’ll have to design a randomized experiment that measures multiple variables to account for the lurking variable that may impact your data.
We walked away from last week’s talk understanding that properly designed marketing equations and experiments can produce reliable data, but we have to be careful about the conclusions we draw from that data. It’s tempting to scour pages and pages of analytics reports for data trends, but fishing for answers can lead to false conclusions. If we take the time to design air-tight experiments with adequate sample sizes, we can discover data that can power our marketing.