There's a lot of talk about data and marketing these days. And rightly so. Data has transformed business, creating and eliminating job titles, and expanding marketing budgets and responsibilities. But let's take a minute to acknowledge one thing: data can lie.
Without careful attention to some of the ways data can be misused, we run the risk of acting on those insights with potentially damaging outcomes; a scary prospect for a job market facing a shortage of data analysis experts.
Identifying mistakes individuals and organizations make when dealing with data is important not just to data analysts and decision makers, but to the public too.
1. Testing Until You Get the Answer You Want
There is a prevailing notion among marketers that if you can't back your ideas up with numbers, then you don't deserve a seat at the table. This is a noble, if misguided attempt to legitimize marketing efforts. The problem with this assertion, is that it fails to acknowledge that data can be manipulated to support or refute ideas.
Simply adding or removing data points, excluding variables, or presenting some results and not others, can drastically change the story that the data tells. Even repeated analyses of a data set can increase the probability that a result is based on chance alone.
This is not to suggest that data is as fallible as guesswork. Rather, it is extremely important that both analysts and decision makers within an organization be knowledgeable about statistical best practices. Doing so keeps everyone's ideas in check, and reduce the chances of bad insights leading to bad decisions.
2. Failing to Recognize Bias
We all fall victim to personal biases at some point, and these can also make their way into research through study design.
For example, you may have pre-conceived notions about what makes a social media campaign a success. You believe that the most important thing is media budget, and that influencers don't really have as meaningful an impact. So in your analysis of campaign engagement, you ignore influencer data completely, confirming your assumptions. With no analyses of influencers, you have no way to disprove your hypotheses.
Time should be spent before analyzing a problem to identify potential biases and their outcomes.
3. Too Much Data
More is not always better, though the big data revolution might have you thinking otherwise. In terms of variety, smaller, better quality, and relevant data is much better than more, lower quality data sets.
In terms of size, not all problems are suited to larger data sets. Larger data sets also have their own limitations from an analysis point of view.
Knowing which data sets and variables to analyze will vary from problem to problem and question to question. Exploratory analysis beforehand can help you identify which are worth including and which are not.
4. Misleading Visualizations
Just as data can be manipulated to present misleading results, so too can data visualizations. Depending on which data you present, and how you present it, visualizations can tell a very different story. Several publications have been accused of this over the years, including the New York Times, Globe and Mail, and many others.
As a consumer of data and data visualizations, it's important to have a critical mind when viewing them. Ask what could be missing, how data might have been presented differently, and what effect that could have on the information that the presenter is trying to convey.
5. Ignoring the Data
That said, if there's one non data-related factor holding back organizations, it's the failure to actually act on insights. This is particularly difficult when your idea is already in action, and the incoming data says that a major component isn't going as planned.
If data is to truly revolutionize how organizations operate, an adherence to best practices, along with a willingness to change direction will be the difference between success and failure, both at a micro and macro level.
So let's make data our friend, and treat it with the sensitivity, time, and care it deserves.
ALSO ON HUFFPOST: