Is Your Data Accidentally Lying to You?
Imagine sliding a report across the desk to your boss outlining how customer acquisition counts are through the roof and the cost to acquire a new customer is at an all-time low − only to find out that you were actually paying to acquire a second or third email address for your existing customers!
We all know data tells a story. However, when big data is reviewed in little silos, it can sometimes tell a rather inaccurate story. Hidden information, misclassified information, False Positives, data collection bias, missing data, bad data entry, subjective data interpretation, cherry-picked data, sample collection bias, small sample size, logical fallacies: these can all result in off-target decision making. According to a recent insideBIGDATAreport, poor data across businesses, organizations and the government contribute costs of up to $3.1 trillion a year to the U.S. economy. To make matters worse, 14.9 percent of marketers claim they do not know what big data is, let alone how to use it.
At Listrak, we've seen merchants interpret a LOT of data over the years. And we've seen a few common themes where giving our merchants the ability to explore their own data could have headed off missteps in their data analysis. As a result, we've spent time developing a customizable data visualization layer that helps our CRM merchants explore their own data beyond a single data point to really understand what's going on under the hood in their marketing ecosystems.
Here are four suggestions on how to use data exploration to look beyond a single data point to extract more holistic and accurate signals from your data:
1. Exclude outliers and test data
One common request we get from clients that utilize our Abandonment Solution is to remove outlier carts that tend to skew average abandoned value and other metrics available in our default Abandonment Dashboard. We also see our fair share of test data. Outliers creep into any data set for a variety of reasons, but their presence can make it difficult to understand your average customer.
The Fix: When exploring your data, add filters to exclude outliers or test data points that fall outside of expected ranges.
2. Clarify what you want to know & improve your data collection accordingly.
Order source is another hot issue for multi-channel retailers. We commonly see orders bucketed into two groups: Online or Offline. However, having only two options, orders that are made online but shipped to store or shipped between stores may create confusion for a retailer hoping to analyze their orders by channel.
The Fix: First, clarify what you're trying to track. Second, if you wish to track both purchase source and pickup location with one field, consider creating additional source names in your data to reflect unique shipping situations. For example, instead of only tracking Online or Offline, decide where each shipping situation applies and add sources to bucket that data. For example, use sources such as Online, Online Ship-to-Store, Offline and Offline-StoreTransfer.
3. Explore your data to rule out other possibilities or events.
A simple change to your website can have a ripple effect on your entire marketing ecosystem. When viewed in a silo, individual dashboards don't always provide enough context or nuance to uncover the reason behind sudden shifts in data. When you encounter a shift in data, use that as an opportunity to look deeper.
The Fix: Explore potentially related data points. For example, if repeat buying appears to be down, review your new subscriber acquisition metrics. Did the website team add an irresistible new promotion for new subscribers that resulted in loyal customers entering a brand-new address to snag that first-time-buyer discount?
4. Pick the right data visualization type and add labels
The beauty of data exploration is that you’re in the driver's seat for finding answers to your company's unique questions. However, if you've ever seen deceptive political advertising (...and who hasn’t?), you know that the way data is presented is critical. Since you undoubtedly work with a team, you have great power to help others analyze the information you've compiled.
The Fix: Help your coworkers and boss understand the custom reports you generate by selecting a data visualization type (chart, graph, scatter chart, etc.) that most effectively illustrates your data. Adding labels and a description will also help.
Hands-on data exploration is a great way to overcome data silos and ensure that your data is telling the full story. Have you found creative ways to overcome reporting silos in your organization?