My apologies if the headline hits a raw nerve. The mass availability of powerful personal computers and code have made data science techniques like Machine Learning available to any company, offering the promise of better business decisions in multiple areas, and at high speed. Increased business investment in this area can put downward pressure on traditional research spend. However, please don’t view traditional research and data science as rival methods.
Machine learning is not Data Science, it’s a subset of it. Market research is also a subset of it. All the term Data Science does is to re-frame the disciplines of “research” and “statistics” in light of the new influence of computing power in making decisions from data.
Consider Drew Conway’s venn diagram explanation of what Data Science is…
He positions traditional research as a keystone of a wider discipline, both for its understanding of how to apply statistics, but also for its ownership of what he calls “substantive expertise”, i.e. knowing the subject area, and coming to the analysis with reasonable hypotheses to test.
Conway’s view also makes it clear that with the simple addition of “hacking skills”, traditional researchers can transform the work that they do. Methods such as ML are only the visible tip of the data science iceberg. The full toolkit supporting it is available to you as a market researcher, to make your job easier, your turnaround faster, and your projects deliver better returns. For example:-
- It can eliminate long-winded and semi-manual data preparation work, saving days of time per month (hence, the “put yourself out of a job”)
- It allows you to repeat the same processes and calculations with absolute precision every time they need doing.
- You can use it to access any dataset, not just Excel and SPSS but also SQL, JSON, and datasets too big for Excel that have traditionally been outside your grasp.
- You can use it to apply quant-style techniques to qual-style data like interview transcripts or open-ended question responses.
- You can build engaging, interactive apps which can communicate your data far more immersively than PowerPoint or Excel ever could.
- You can transform your data into formats that make them accessible to the company’s ‘data lake’ and programmatic use.
At least one of the items in the list below is likely to have piqued your interest. So how do you go about exploiting data science in market research jobs?
What Conway describes as “hacking skills” basically means learning code. Researchers already have the inquisitive minds required to be a good “hacker”, and nowadays learning code is not hard. More than this, it’s going to be a primary job skill for researchers in the very near future. How many colleagues do you know who resisted the need to become expert in Excel or SPSS?
For data science, you have a couple of fantastic options; Python and R. These are increasingly interchangeable, but Python would be first choice if you want to take the plunge into Machine Learning, R would be first choice for pretty much everything else.
Neither of these languages are just for advanced maths. As opensource languages they evolve constantly, creating new methods to solve problems across the full data science/market research range of work. Importantly, both use simple logic and functions (instructions) that look very much like plain English.
So how to get started? The Media Research Group offers one-day courses to get you up and running with R for everyday media research data work (disclosure: course delivered by Culture of Insight!). Online learning platform datacamp also offers a wide range of courses in Python and R, and the first modules are often free.
Comments