Top Six Marketing Analytics Mistakes

In my experience there are several top marketing analytics mistakes that are common across different industries. Unfortunately, I have seen these problems repeated over and over, in many companies, during my career. If you want to be successful in analytics, you have to be particularly careful to avoid these mistakes. Here is a short list:

• Not adequately influencing the decision maker

It is a completely wasted opportunity for the company if you are unable to influence the decision maker using your data-driven analytics insights. In some cases the issue is the lack of storytelling skills. Analytics professionals need to improve their communications skills and be able to explain the stories that the data is telling them in a way that business decision makers will easily understand. Sometimes, the problem is the lack of sufficient effort in convincing all necessary stakeholders. This is particularly challenging in large organizations where you need to convince a sizable number of stakeholders before any decisions are made. I know how frustrating it is to spend time on an analysis, only to see it rapidly discarded; to ensure that doesn’t happen to you, improve your internal “selling” skills. You and your organization will benefit tremendously.

• Not focusing on the most relevant business problems

Many times, the most relevant problems from a business-impact perspective are also the most complex and difficult to solve from an analytics point of view. That means analytics professionals might steer clear and head straight for the easy problems. “Easy problems” are, for example, situations with readily available data or questions about marketing activities that are simple to track. But who cares if you have fantastic results from an advertising campaign test, if that campaign can only potentially affect a tiny fraction of the company revenue? If your analytics professionals fall into the trap of solving the easy problems rather than tackling the most relevant problems, you might have to get them back on track. When you are trying to decide where to focus your analytics efforts, look for the biggest potential revenue impact, margin impact or cost reduction opportunities. Then start your analytics there.

• Lack of a holistic view

Many times in marketing analytics, the lack of a holistic view becomes a problem. This is also called a problem of “missing variables” in econometric modeling. For example, a few years ago, when I was working at a high tech company, a friend of mine who was running paid search advertising campaigns came to me, excited that his campaigns had hugely increased their return on investment (ROI) over the past week. When I asked what he’d done differently, he said that he had not changed anything. But it just so happened that we had spent $10 million on a TV campaign that same week, which resulted in the improved ROI in his paid search campaigns. This problem is particularly common in digital marketing analytics, which tend to overlook influences from offline media that are harder to track. Conversely, in another company, I saw that improving the customer experience in tech support calls generated a much higher chance of repeat customers, cross sells, and upsells, and thus higher long- term customer lifetime value. This was outside the control of the Chief Marketing Officer of the company, but it proved to be so important for our revenue that she brought it to the attention of the Chief Operating Officer to improve the quality of the calls. Be sure you are looking at all aspects of the business, not just your marketing department.

• Using the wrong data

This problem encompasses a broad range of mistakes like using the wrong data, using data that has noise, or using data that is irrelevant. Your analysis is only as good as the data you’re using to generate it. If you are using the wrong data, then all your analytics conclusions based on that data are going to be inaccurate. There is a lot to be said about the importance of the process of cleaning data before starting any analysis. This typically involves looking at the data to identify noise patterns, remove outliers, or complete any missing parts. For example, if you are using data from two different tracking systems, an old one you used last year and a new one you are using this year, you’ll need to normalize the information to be able to join both sources of data for analyses and comparisons.

• Lack of statistical significance

Unfortunately I have heard too many times the expression that some information was “directionally correct” to justify using test results that are not statistically significant. Let me be clear: if a result is not statistically significant, that means you really don’t know anything, so you cannot make a decision based on that information. This also applies to results of statistical models that have components that are not statistically significant.

• Limitations of methodologies

If you don’t understand the limitations of the analytics methodologies you are using, you will be tempted to take results at face value and act upon them, instead of accounting for those limitations. One way to approach these limitations is to cross validate results using several techniques. For example, you can use a controlled in-market test to confirm the results of an econometric model. A common problem to avoid is to use the same tool to try to solve all situations. For example, some companies or professionals specialize in certain techniques like applied neural networks, agent- based modeling, or multi-linear regression. In this case, every problem that they encounter they try to solve with that same tool. Each analytics methodology has pros and cons, and being aware of those will help you avoid plenty of mistakes.

In sum, in order to avoid making marketing analytics mistakes, you need to adequately influence the decision maker, focus on the most relevant business problems, apply a holistic view, use the right data, make sure that your results are statistically significant and, crucially, be aware of the limitations of the methodologies you use.