If marketing researchers disregard primary research, they’re going to be punished. Data, data, everywhere, Nor any drop to drink… Experienced statisticians and data scientists have all had the experience of analyzing masses of data every which way from Sunday without finding the answers they were seeking. I should be honest and confess that data dredging is a statistical sin… but everybody does it. See Stuff Happens for some of the ways it can put your analysis, and perhaps your job, in peril.
Forgiving this shameful behavior for a moment, the reason why this often happens is because the data we’re analyzing were collected for reasons unrelated to our research objectives. In extreme cases data dredging has been likened to mining a landfill for gold. Sometimes we do strike gold and not all data are landfills, but the risks that we won’t find what we’re looking for are not trivial. There is also plenty of fool’s gold in them thar landfills. However, with so much data now at our fingertips, we’re easily tempted to believe that if we seek we will find…
Primary research is often the better path. Wikipedia offers a concise definition of it (happily, free of goofy mixed metaphors):
Primary research involves the collection of original primary data. It is often undertaken after researchers have gained some insight into an issue by reviewing secondary research or by analyzing previously collected primary data. It can be accomplished through various methods, including questionnairesand telephone interviews in market research, or experiments and directobservations in the physical sciences, among others.
Again, from Wikipedia:
Secondary research (also known as desk research) involves the summary, collation and/or synthesis of existing research rather than primary research, in which data are collected from, for example, research subjects or experiments. Care should be taken to distinguish secondary research from primary research that uses raw secondary data sources. The key of distinction is whether the secondary source used has already been analyzed and interpreted by the primary authors.In a market research context, secondary research is taken to include the reuse, by a second party, of any data collected by a first party or parties. Sometimes, secondary research is required in the preliminary stages of research to determine what is known already and what new data is required or else to inform research design. At other times, it may be the only research technique used.
The Purdue OWL provides more detail as well as tips and guidelines for certain kinds of primary research. An online search will lead you to many other useful documents, including full presentations on primary and secondary research that cover the advantages and disadvantages of each.
Cost and speed often favor secondary research, but not always. In my line of work, marketing science, it usually makes more sense to think of them as complementary rather than competing methodologies. As noted above, the distinction between the two kinds of research is not always sharp since original analysis of existing data may be classified as primary research. Examples of this include analyzing social media data and customer transaction data to address specific marketing business objectives.
Most of the time, however, when we speak of primary research we mean research that is designed to address specific questions and requires the collection of original data for those purposes. Focus groups, in-depth interviews and consumer surveys are typical examples of primary marketing research. Secondary research, as mentioned earlier, often provides essential background for the design of primary research and the analysis of primary research data. Frequently, in the social and behavioral sciences, of which marketing research is one kind, primary qualitative research is conducted to aid in the design of primary quantitative research. The quantitative phase, in turn, may lead to further qualitative research or additional secondary research.
Marketing research projects can stretch on for months because designing the research, analyzing the data, and absorbing and interpreting the results requires careful thought and frequently the input and involvement of many people. Only parts of the entire process can be automated. When a project has clear and simple objectives, or is being repeated – for instance, the second and subsequent waves of a tracking study – automation is more practical. Many marketing research projects are one-off customized studies in which this is less cost-effective or simply not feasible, however. Those who claim otherwise are revealing that their knowledge of marketing research is limited.
Why do we conduct marketing research in the first place? Here are some of the questions it tries to answer:
- Who uses our product category? How do they differ from those who do not?Who knows about our brand?
- Who buys our brand? Are there segments of consumers with different purchase patterns, demographics and things they want from the category?
- What media do they use? How best to reach them?
- How do they shop our category? Is the purchase for mostly themselves or for others? Is it mainly impulse, autopilot or planned?
- Where do they shop?
- What other brands in the category do they buy? Where do they buy them?
- Why do they buy our brand more (or less) often than competitors’ brands? What do they like most and least about our brand…and the competition?
- Is our brand frequently confused with other brands, for example, because of package or brand name?
- Are there groups (segments) of brands in our category that compete more directly against each other than with other brands?
- Do the brands in our category have distinct images that are associated with purchase patterns?
- What about price? Is our brand seen as low end, high end, as providing value for money? Does this match the positioning we’ve tried to create for it?
- How much would our sales change if we raised or lowered our price?
- Are there missing SKUs in our lineup? Conversely, are there SKUs or variants we should drop?
- How do consumers define the category? Is it different from the way we do? What other categories do we compete against?
- Do we understand the ways people actually use the products in the category, e.g., how and for what occasions? Do purchase and usage differ by occasion?
- Is our product difficult to understand or use?
- If we change the formulation of our product, will consumers notice? How will they react?
- What do consumers want that is not now currently offered by any brand in the category? What do they think of our ideas for new products?
- How well is our marketing working? Should we reallocate spend to achieve better ROI?
- Do we need to adjust our mix of marketing channels?
- Do we need to adjust our mix of distribution channels?
- Do some kinds of people respond more (or less) to our marketing? How is our marketing (and competitors’) marketing changing consumers’ expectations for the category and their purchase behavior?
- What role does seasonality play in our category?
- How do in-store promotions and shelf placement come into play?
- Sales are rising/declining/flat – do we understand why?
- Which of these candidate ads will work best?
Some of these questions can be addressed satisfactorily through secondary research or primary research of existing data, but most require primary research in addition. There’s no way around it – all the data we’ll need are not just out there waiting to be analyzed. Research – including the analytics and deliverables – must be specifically designed to answer these sorts of questions, in many cases.
Attitudes towards shopping, uncovering what different consumer segments want from a product category, their detailed demographics and other very specific data often must be purposely collected and tied together statistically to give us the answers we need. Very general psychographics and canned demographics designed by third parties, let alone the imputed data found in typical data mash ups, will seldom provide a picture that is sufficiently detailed and accurate to answer our questions adequately.
When marketers and marketing researchers talk about big data, machine learning, automation and AI they are often referring to direct marketing, CRM, media planning, attribution modeling and other marketing-related activities that have not historically been considered marketing research. These applications typically utilize secondary data for predictive analytics. Marketing research can inform these activities but cannot be replaced by them. There seems to be quite a lot of confusion regarding this, which originates in large part from people selling technology and from the marketing-research-is-dead crowd. Many of their criticisms of marketing research concern things we do not actually do. I used to turn the other cheek and ignore them but, in many cases, I find it is now necessary to respond to these criticisms directly and sometimes forcefully.
In the sections of my company library labeled Research Design and Causal Analysis, Sampling and Survey Research and Qualitative Methods, I’ve listed a number of books about, or related to, the design of primary studies. If you are very new to research, it can be helpful to prepare a brief mock-up of the Executive Summary and Business Implications section of a mock report. I still recall doing this myself many years ago.
See Combining Smart Design with Smart Analytics, Why Segment? and Taking Quantitative Marketing Research to a Higher Level for some thoughts on how marketing research is typically conducted and various ways I feel it can be improved. Many more of my pontifications regarding marketing research and data science can be found on my LinkedIn profile under articles.
I hope you’ve found this interesting and useful!
___________________________________________
Article by channel:
Everything you need to know about Digital Transformation
The best articles, news and events direct to your inbox
Read more articles tagged: Featured, Marketing Analytics