Where would you like to start?
Guide to Market Research
Market research is about three things:
Typically, these three things occur as a part of a fairly standard process.
We estimate things. What is our market share? How satisfied are our customers? What impact will a 10% price cut have on sales? Second, we explain things. Why do people buy iPhones? What makes customers satisfied? Why did our last product launch bomb? Why do sales increase by 30% when price is cut by 10%? What can we do to rejuvenate our brand?
Lots of market research studies are focused entirely on estimation (although this technical term is not commonly used by practitioners). Market sizing studies work out how many people are in a market. Usage and attitude studies (U&As) measure what people think and do. Governments around the world conduct census studies, estimating the number of people in their countries.
When people think about market research they often think about questionnaires, asking people things like:
Thinking about your last flight on Qantas, were you...  Very satisfied  Somewhat satisfied  Neither satisfied nor dissatisfied  Somewhat dissatisfied  Very dissatisfied
However, there are lots of other ways of estimating facts about markets. Peoplemeters sit on top of TVs and count who is watching what at any given time. Point-of-sale scanners in the supermarket measure what has been purchased. Toy manufacturers give children rooms full of new toys and observe which ones are played with. Some researchers even sort through garbage to find out what people really eat and buy. In focus groups and in-depth interviews, qualitative researchers discuss topics with people and form conclusions about what people think; these conclusions are also estimates (although qualitative researchers would not use this term!).
Estimates can relate to history, such as how many people purchased a product last month. Estimates can be predictions, such as the revenue impact of a price increase or the degree of cannibalization that will result when a new product launches. We can even have estimates of things that did not happen, but could have happened, such as the amount of sales that we would have had last year if we had spent an extra million dollars on advertising. Predictions can be quantitative forecasts, such as forecasting that a new ice cream will be purchased by 8% of the market. Predictions can also be more general conclusions about cause and effect, such as concluding that a change in the packaging design of a product will result in an increase in sales.
Estimates are often derided, with good reason, as being mere facts and figures. The real need is often for understanding. Is Microsoft the dominant software brand in the world because it is excellent, because it has blocked its competitors or for some other reason? Why do people buy Coke? What does a new bank have to do in order to encourage customers to switch to it? What can a government do to make people more willing to pay their tax?
The core of all good research is the creation of models. Models explain how things work. Any explanation of how a market works is, by definition, a model. We cannot estimate anything without models. Think about an estimate of new sales for a product. The only way we can create such an estimate is by taking into account the factors that determine how many people will buy the product. Less obviously, even a simple estimate, such as the proportion of people that are women, can only be obtained by the application of numerous models (we return to this later).
The term ‘model’ can conjure fear in the minds of the uninitiated. Some people imagine that models are complex sets of equations that can only be understood by statisticians and mathematicians. This is a misunderstanding. The technical meaning of the word ‘model’ is identical to the understanding you had as a child. Consider a model of Copenhagen’s harbor, built out of Lego. This is a model in exactly the same way that we use the term ‘model’ on this site: it is something designed to look like the real things, but it is not actually the real thing.
A model is a simplification of reality. Good models of markets capture the key aspects of how markets work. Bad models misrepresent reality.
We can build a model out of wood, Lego, metal, clay and any number of other materials. Models can be expressed in words:
The amount of advertising determines the level of sales.
Models can be combinations of images and words:
And as equations:
$Sales = 121 + 4.1 × $Advertising expenditure
Each of these examples of models differs in their precision. The equation is generally the most precise type of model (although the precision will be misleading if the equation is wrong). Models also differ markedly in terms of their complexity. The models we have seen so far are all simple, involving only a few variables. In real-world applications much more complex models are often required, such as the model shown below, which is discussed in Laddering).
Although models are at the heart of research, the term model is a technical term and few people doing market research use the term when describing what they are doing. Sometimes people will instead talk about theories. A theory is just another name for a model.
Sometimes people talk about hypotheses. A hypothesis is a tentative part of a theory. Often research is conducted to check a particular hypothesis. By checking and refining a model, we better understand how a market works. For example, if somebody says “we have a hypothesis that we need to reduce the price of our product to below $4 in order for people to notice our product”, they are saying that a stakeholder believes that this may be the case and they want to check if it is true. Similarly, the hypothesis that “price is more important than brand” is a speculation about the strength of the relationship between variables in a model (i.e., they are speculating about estimates).
Most often, users of research talk about wanting to conduct research so that they can find insights. An insight is a conclusion about how a market works that is useful. As discussed, a model describes how a market works, so an insight is a description of a model, or a part of a model. However, to qualify as an insight, the conclusions about how a market works need to be useful. Thus, an insight is a conclusion about how a market works which results in a company doing something that they would not have otherwise done. Research that does not lead to some improved outcome, such as a better strategy or policy, has thus, by definition, failed to produce insight.
Consider the 4P model of marketing, which says that the success of a product is determined by price, product, promotion and place:
This is a popular model in introductory marketing courses (particularly once ‘promotion’ is defined to include advertising). Many students in introductory courses know little about the subject, and as a result, find this model useful. For some, this simple framework is an insight. However, this model would never be a source of insight to an experienced marketer.
To qualify as an insight, we need to have something that is a statement about how a market works which is non-obvious to the end-user of the research. A conclusion like “if we advertise at the same time as offering discounts, we get twice as many sales as when we offer discounts without advertising” may qualify as an insight, provided that this is correct and not already known by the end-user.
Throughout this site we will discuss lots of quite technical aspects of research. However, it is important to never forget that they are just means to an end, and the end is insight with potential to create competitive advantage. Research that provides no insight is useless. The challenge of obtaining insight is at the heart of the tension that exists between research consultants, and their clients, as illustrated in this joke:
A man flying in a hot air balloon suddenly realizes he’s lost. He reduces height and spots a man down below. He lowers the balloon further and shouts to get directions, “Excuse me, can you tell me where I am?”
The man below says: “Yes, you’re in a hot air balloon, hovering 30 feet above this field.”
“You must work in market research,” says the balloonist. “I do” replies the man. “How did you know?” “Well,” says the balloonist, “everything you have told me is technically correct, but it’s of no use to anyone.”
The man below replies, “You must work in marketing.” “I do” replies the balloonist, “But how’d you know?” “Well”, says the man, “you don’t know where you are, or where you’re going, but you expect me to be able to help. You’re in the same position you were before we met, but now it’s my fault.”
How to find insights
The best chance of obtaining insights from research is to:
- Create models.
- Use the right type of data.
- Work out what is known before attempting any research.
Use appropriate data
The data that is used needs to be capable of containing insights. Although it is certainly the case that some people are more gifted than others at extracting insights from data, it is also true that some forms of data are more useful in helping to find data than others.
Remembering that an insight is a new conclusion about how the world works, it follows that our best chance of finding insights is to use data that we have not previously examined. The more varied the data we examine, the better our chance of finding insight. If we have gained most of our understanding about how a market works by analyzing sales data, when trying to crack a difficult problem, we should focus on using some completely different type of data (e.g., qualitative research, surveys). Similarly, if a firm has typically used qualitative research, the likelihood is that it will gain better results from a quantitative survey.
Identify what is known before doing any research
If we do not know what models are currently being used to make decisions, it is extremely difficult to design research which will uncover insights. When we analyze data, there are a thousand and one results that may appear to be interesting. As most people do not have the patience nor the capacity to review thousands of findings, researchers focus on presenting the findings that seem to best explain how the market works. If these are already known it is almost inevitable that nothing ends up being found out from the research.
Unfortunately, many people who commission research fail to understand this last point and, when asked for their theories and hypotheses, say things like “I want the research to be objective … I don’t want to taint the research with my prejudices … you just do the research, I’ll work out it means.” However, these people fail to recognize that in doing so they are maximizing their chance of getting useless research (in other words, finding out they are in the hot air balloon).
Standard Research Process
Briefing: the client articulates what they want to know (i.e., their objectives for the research). Most of the time, they will either be conducting research to gain one or more of specific estimates (e.g., market share, advertising recall, predictions) or insights. This briefing can be done verbally or in writing; the written document is called a brief in England and Australia and a request for proposal (or, RFP) in the US.
Project scoping: the researcher works out what the client needs. The project scoping needs to find one or more of: all the estimates the client requires; any hypotheses or theories of the client, and the behavior(s) in the market that the client wishes to change (most commonly, brand choice).
Research design: the plan by which estimates are going to be computed and models refined. Research designs are sometimes referred to as research methodologies. The 6Ws is a good aid memoir for making sure that all the detail of the research design has been thought through:
Who should be interviewed? What information should be obtained? When should the data be collected? Where should the data be collected? Why should the data be collected? What way should it be collected?
Proposal: this describes the research design to the client, along with pricing and timings. It is primarily a sales document. It is typically about six to twelve pages in length, but it can be as short as a one line email or as long as several hundred pages.
Instrument design: questionnaires are written for quantitative research, data capture specifications for observational studies and discussion guides for focus groups and depths (in-depth interviews). See Basic Questionnaires and Advanced Questions and Questionnaires for more information.
Data collection: interviewing and/or observation (e.g., recording where people go in a shopping center). This will typically involve qualitative research, conducting a quantitative survey with a questionnaire or a both of these activities.
Data processing: erroneous data is identified and either deleted or corrected (cleaned), and text data is turned into categorical data (coded).
Weighting: the distribution of the sample on key variables is compared to what is believed to be the correct distribution; if the distributions differ, a weight is computed to correct the sample (this is discussed in more detail in the next chapter).
Analysis: estimates are computed and models are refined. This can take many months, in the case of large strategic pieces and government reports, or may be as limited as a bit of reflection in the shower for a fast turnaround qualitative study. This is discussed in more detail in Basic Data Analysis and Advanced Data Analysis
Reporting: sharing the results with the client. Most of the time this is done with a presentation and a PowerPoint document, and perhaps an online dashboard.