What Every Marketer Needs to Know About Data Science

This week our motorcycle-riding data scientist, Wils Corrigan, Ph.D., returns to share what marketers need to know about data science. While the Anametrix platform automates tasks assigned to data scientists – allowing more time to actually analyze the data – it’s up to Wils and his team to answer clients’ really specific marketing questions. Give Wils and his team a marketing question and he will roll up his sleeves, dive deep into the data and use his predictive analytics modeling and statistical power to deliver the answer. We caught up with Wils to find out what he thinks every marketer should know about what he does.

I am a statistician and a geographer, which are well-defined terms; however, at work, I’m called a data scientist, which is still a bit ambiguous. It’s highly likely that if you ask a dozen different “data scientists” to provide a definition for data science you will get twelve different answers. Most of the answers will include math, statistics, computing with data, machine learning and data visualization. Some will mention big data, computer programming, messy data, creation of data products and communication skills. Then a few will have some odd items on the list, claiming that to be a data scientist one must know six sigma, have advanced spreadsheet skills, know rules-of-thumb but not theory, think models are bad or have a penchant for storytelling. The term data science has become one of those nebulous buzzwords that make statistics or data analysis seem interesting and sexy but gets added to the list of confusing industry jargon (e.g. data mining, big data, predictive analytics, etc.).

Since even the definition of the term is so unclear, you, as a marketer, might be wondering what data scientists do, how we provide value to a marketing department and why data scientists are in such high demand.

Let’s start with what data scientists do. My working definition of a data scientist is someone who engages in a multi-disciplinary approach to data analysis intended to assist clients with making decisions in the face of uncertainty. I’ve often heard questions similar to, “does the data analysis just look in the rear-view mirror?” Usually in these situations the “rear-view mirror” referred to previous experience with analyses or applications that only used simple graphs and descriptive statistics, which barely scratched the surface of what was possible to learn from the data. Now, while I do like driving, and especially riding, it truly is a poor analogy for data analysis. So instead of Mario Andretti, let’s use the example of Sherlock Holmes.

1. Sherlock first consults with the police (consultation with client)
2. Then he investigates the crime (data gathering)
3. He sifts through the information to determine what is reliable and what is unreliable (data sublimation)
4. Sherlock deduces what happened (exploratory data analysis and inferential statistics)
5. He decides upon where and how to catch the criminal, hopefully before another crime is committed (inferential statistics and predictive modeling)
6. Finally, at least ideally, based upon his experience in numerous cases, Sherlock assists the authorities in figuring out how to reduce the chance that similar crimes happen again (optimization)

These are just some of the steps a data scientist can take when solving a problem, and note that each step can be every bit as important as any other to the overall outcome. For example, data of unknown quality leads to results of unknown quality. As I mentioned in my blog interview with Anametrix CEO Pelin Thorogood, the only thing worse than having no data is having bad data.

Now, let’s dive into some of the terms used above in the Sherlock Holmes example:

• Exploratory data analysis is examining the data to improve understanding and enhance subsequent analyses and modeling. Data visualization is a good example of this technique. Through exploratory data analysis, data scientists can identify outliers, patterns and trends, as well as indicate a need for additional data.
• Inferential statistics tests hypotheses and make inferences about populations (potential customers) based upon samples (current customers). Techniques include regression trees, ANOVA and linear and nonlinear models, to name a few. This process allows data scientists to draw conclusions about the importance and effectiveness of different strategies and tactics.
• Predictive analytics refers to the application of statistical models and machine-learning algorithms to create predictions or forecasts. A predictive model uses a set of observed values and its relationship with multiple predictors to estimate which outcome is the most likely to occur given a particular set of circumstances. Data scientists can then tell marketers the most likely results of proposed plans and increase the amount of information available for decision-making.
• Optimization takes predictive analytics a step further by providing recommendations to improve performance or revenue, based upon the expected outcomes from the predictive models. Techniques include linear, quadratic and integer programming and global optimization heuristics.

Using the processes and techniques listed above paired with good quality data, data scientists can provide value to a marketing department and answer questions such as:  “Which campaigns are working?” “What factors affect churn?” “What campaigns will perform best during the holiday season?” Or even “How should I allocate my marketing spend for next year?” I’m guessing the data scientist’s ability to answer these difficult questions is why the field is in such high demand.

So what does every marketer need to know about data science? While it is a new field with an ambiguous definition, data science builds upon decades of research in several fields. Data science allows marketers to get great value from data by transforming data into information and removing the guesswork in the answers to those difficult marketing questions. But start early. Data scientists can be more effective if they are involved in the beginning stages and can assist with determining which data to collect and use, the design of data collection methodologies, the quality of existing data assets and the scope of the marketing questions.

Thanks Wils for sharing what marketers need to know about your field. To find out more about how the Anametrix platform automates data scientist tasks, click here to schedule a demo. If you’d like to work with Wils and his team to answer a specific marketing question, please outreach sales@anametrix.com.