When a global computer reseller wanted to make sales teams more efficient by focusing on deals more likely to close, it built an in-house model to score leads and shortened sales cycle time. Data science played a key role in giving the sales teams an edge.
When a digital commerce fashion brand wanted to improve conversion rates on its websites, it built a recommendation system based on past purchase data. Again, data science was a key partner in delivering ROI.
As these real examples show, data science is now aimed at many marketing problems — which means that marketers must understand the fundamentals of data science methods and how best to engage data scientists with their teams and projects.
“Many marketers are called upon to have a grasp of basic data science methods to communicate with data scientists and analytics teams, evaluate projects and assess opportunities,” noted Martin Kihn, research vice president, Gartner for Marketers.
Mr. Kihn provided the following overview of data science fundamentals to help marketers improve collaboration and outcomes with their data scientist partners.
Data science basics
Data science for marketing is a hybrid discipline that draws from statistics, mathematics and computer science that can encompass elements of network theory, geography and even neuroscience. Data science exists to solve business problems and, in a marketing context, it requires intimate understanding of business needs. This technique is how marketers get useful insights from data of any size and form. This differs from big data, which focuses on infrastructure, storage and processing.
What data scientists do
In general, data scientists explore data, make predictions and find structure. Typical tasks for a marketing data scientist to deliver this information include:
- Measure: Determine the impact of marketing efforts and ad campaigns
- Optimize: Recommend changes in tactics or spending to improve results
- Experiment: Design and execute tests to isolate causes
- Segment: Identify groups and subgroups of customers and prospects
- Model: Build predictive computer models to improve response rates
- Storytelling: Communicate messages derived from data to inspire better decisions
Data scientists work within three domains.
Data exploration is when a data scientist uses statistics and visualization techniques to find patterns in data. Visualizations, charts and dashboards are what exploration looks like for a marketer.
Data experimentation is applying experiment design methods to develop and test hypotheses under controlled conditions. A/B and multivariate testing are examples of data experimentation.
Finally, machine learning is applying algorithms to build models and make predictions. Predictive models and clustering are examples of machine learning in action.
How to work with data scientists
Keep in mind that quantitative skills, some industry knowledge and especially curiosity are what makes a good marketing data scientist. Provide opportunities for data scientists to learn on the job and explore speculative ideas that tap into their curiosity. Don’t “drag-and-drop” data scientists into a problem. Marketing analysts need help defining precise business needs and trade-offs.
Don’t “drag-and-drop” data scientists into a problem.
Develop an open and collaborative relationship with your analysts and data science experts, both in-house and agency-side, to ensure they understand the problems you are trying to solve. In a world that is increasingly reliant on data-driven decision making, pay attention to the caveats made by your analysts. They are often in a position to know how reliable their recommendations are — or are not.
“Continue to deepen your appreciation of the potential and limits of data science,” noted Mr. Kihn. “It will improve the effectiveness and efficiency of your programs.”