What is data science in market research?

Data Science covers a range of different disciplines that are to do with the collection, analysis and reporting of data. Chris Cook, Managing Director, PAL Stats, offers his expertise to discuss what data science in market research is

The disciplines that are needed depend on the requirements of the project but include:

• Market Research
• Mathematics
• Statistics
• Computer Science
• Machine Learning &
• Artificial Intelligence
• Information Science.

Once you have identified an issue, you need to decide how you are going to solve it.

Below are the broad steps that need some thought:

• 1. Specify the problem
• 2. Deciding on the research needed for a solution
• 3. Collect the data:

a) Identify the audience

b) Run qualitative analysis (focus groups, depth interviews)

c) Construct a questionnaire

d) Fieldwork (getting questionnaires completed by the target audience)

e) Ensuring data is in a form you can run qualitative analysis on (i.e. in digital format)

• 4. Run the analysis
• 5. Interpret the results
• 6. Communicate the results to the stakeholders
• 7. Make a plan, communicating this to your teams
• 8. Implement the solution to your problem
• 9. Monitor, review the results of your plan
• 10. Assess and refine your thoughts to see how you can improve things further. Then start again at point 1 changing the problem based on the evidence you have obtained.

I will be writing a series of articles covering the data science subject areas (points 1-5). I will do this from a conceptual point of view and try to keep the complexity of the mathematics to a minimum.

1. Specifying the problem

Phase 1 is the most critical phase as Albert Einstein said: “The formulation of a problem is often more essential than its solution, which may be merely a matter of mathematical or experimental skill.”

If you don’t get this correct, you could be wasting money and resources, implementing a poor solution as the problem was not specified correctly. Consult with all stakeholders to get an understanding of the problem. Try to identify what result you are trying to achieve.

Assess the impact of resolving the issues and the benefits that would derive from the solution. Determine a budget to resolve the issue and what the potential cost savings/benefits the solution may have. Assess the value for money of the implementation would bring budgeted costs vs cost savings. If the budget allows, bring in professional help. If it is not value for money and you can work with the existing deficiencies, it may well be better not to embark on the project. Try to identify any issue and risks that the project may cause and try to mitigate these when specifying the problem.

2. Deciding on the research needed for a solution

You need to decide how you are going to solve the problem.

• What data are you going to collect
• What information you need to collect that will give you the answers you need
• From where and who do you need to collect data
• Decide what methods of data collection you are going to use
• Decide how you are going to analyse the data
• Decide who is going to interpret the results and then action them
• How are you going to communicate the results to the stakeholders?
• How do you assess whether the plan has worked or not?

3. Collecting the data

a. Identify the audience

Specify who your audience is, do you have access to every member of that audience? Ideally, you should. Everyone in the audience is called the population.

Try to identify any holes in the audience and see if you can reduce those holes, perhaps see if there are others with whom you can partner. Ensure that the audience has given their permission to be contacted so that you are GDPR-compliant.

If you ask every member in the audience and they reply it is called a census.

If you take a subset of the audience from the population, it is called a sample. The larger the sample, the more accurate the results are. However, depending on the specific case, a sample of 500 is usually good enough. Every person in the population should have an equal chance of being picked (this is known as the uniform probability model). If they don’t have an equal chance, then bias can be introduced. Any bias can distort the results.

b. Run qualitative analysis

Often this an important phase.

Depth interview

Asks a lot of why questions to understand the topic areas providing descriptive data about people’s behaviours, attitudes and perceptions, helps to understand complex processes, from this it will help you construct relevant questions in a questionnaire.

Focus Groups

A focus group discussion is an interaction among one or more experts and more than one individual. The intention is to gather data. In a focus group, discussion investigators interview people with common qualities or experience for eliciting ideas, thoughts and perceptions about subject areas or certain issues associated with an area of interest. It is particularly useful in segmentation studies as, when the participants take different views on a topic, there is a possible segmentation to be had. If the topic has consensus, then the topic is not discriminating and should not be asked in a survey that’s aim is to segment the population.

Next time I will cover the other elements of “collect the data”.

Please note: This is a commercial profile

Chris Cook

Managing Director
PAL Stats Ltd
Phone: +44 (0)1908 921 000
Website: Visit Website