3.3.2 - Market research and sampling

Sources and forms of market evidence

The value of research depends on whether it answers the business's actual decision problem rather than merely collecting information.

Primary methods include questionnaires, interviews, focus groups, observation and trials. Illustrative case: a repair company is deciding whether to offer evening appointments. It could ask potential customers about their working hours, interview people who failed to book and trial an evening service. A questionnaire can obtain comparable answers from many people; an interview can explore why a customer hesitates. Observation records behaviour rather than relying entirely on memory, although the researcher may not know the reason for the behaviour. A trial measures actual purchasing in a limited setting, but its results may not transfer to a different location or season.

Primary research can be tailored to the precise product, price and target market. It may also be costly, slow and badly designed. A leading question such as “Would you welcome our convenient new service?” encourages agreement. Asking “Which appointment times could you attend?” is more neutral and yields more actionable information. Stated intentions are not the same as purchases: somebody willing to try a free service may reject its eventual price.

Secondary sources include the business's transaction records, published industry research and official population or economic data. Internal sales records can reveal which items sell together and when customers return. Existing material is often quicker and cheaper to use, but its definitions, coverage and age must be checked. An industry report for the whole country may tell a local shop little about residents within walking distance. Sales records include existing customers, not necessarily the people the firm wants to attract.

The Office for National Statistics reported that 18.6% of England and Wales's population was aged 65 or over in Census 2021, compared with 16.4% in 2011. An accessible-transport business can use that dated secondary evidence to investigate an ageing market. It still needs local information about travel needs and willingness to pay: the national age proportion is not a forecast of its own bookings. ONS Census 2021 population estimates.

Quantitative data are numerical: the number of orders, prices paid or the proportion selecting an answer. They support calculations and comparisons. Qualitative data concern meanings, reasons and experiences: an interview might show that customers find a website confusing. Qualitative research can explain a numerical pattern and uncover an unanticipated need; a few detailed interviews cannot establish how common that need is in the whole market.

These distinctions are separate. A primary survey can contain both numerical ratings and open comments. A secondary published interview study can contain qualitative evidence. Businesses often combine them: use secondary data to identify a possible segment, interview it to understand needs, then quantify likely demand with a carefully designed survey or trial.

Selecting a representative sample

Sampling studies a smaller group from a population so that research can be completed within the available time and budget.

The population is the full group relevant to the question, such as all households in a proposed delivery area. A sample is the subset actually studied. Researching every household may be impractical, so the business hopes that the sample provides a useful picture of the population. Representativeness concerns how well it reflects relevant characteristics and behaviours; a large sample of the wrong people can still mislead.

In random sampling, each member has a known chance of selection, commonly an equal chance in simple random sampling. Researchers can select randomly from a complete customer list rather than choose whoever is easiest to contact. This reduces deliberate selection bias, but requires a suitable sampling frame: a customer list excludes non-customers. Randomly selected people may also fail to respond, creating non-response bias if their views differ from those who participate.

Stratified sampling divides the population into relevant groups, or strata, and selects randomly within each. Proportional stratification reflects each group's population share. Researchers must know the population's composition, choose relevant strata and access suitable lists. Proportional representation on one characteristic does not eliminate bias on every other characteristic.

Applying known population proportions

A population is 60% commuters and 40% non-commuters. The planned sample is 200.

1. Convert each percentage to a decimal and multiply by sample size:

1. Check the group total: $120+80=200$.
2. Select randomly within each group; choosing the nearest 120 commuters would not produce a stratified random sample.

Deriving proportions from group totals

A delivery area has 900 flat households and 600 house households. A sample of 150 is required.

1. Find the total population:

1. Apply each population fraction to the sample:

1. Check $90+60=150$, then select randomly from both household lists. This ensures the intended composition, while questionnaire and non-response bias remain possible.

Quota sampling sets required numbers for specified groups, but interviewers select respondents without random selection within those groups. An interviewer might be told to obtain answers from 120 commuters and 80 non-commuters near the station and town centre. It can be quick and avoid the need for a complete sampling frame. However, the interviewer may select approachable people or people present at particular times. Meeting the quotas does not establish a random or fully representative sample.

The distinction between stratified and quota sampling is therefore not simply that both use groups. Stratified sampling includes random selection within the strata; quota sampling uses non-random selection to fill the required group totals.

A larger well-designed sample can reduce random sampling variation, but cannot repair a leading question, missing group or systematically biased selection. The right method depends on decision stakes and resources. A national launch may justify a carefully stratified study; a small shop choosing between two window displays may gain sufficient evidence from a limited, honestly interpreted trial.

Using research in marketing plans

Research reduces uncertainty most effectively when managers connect the evidence to a specific choice and examine how the evidence was produced.

Illustrative case: a small kitchenware producer is considering a dishwasher-safe lunch container. Interviews suggest that easy cleaning matters to customers. A survey estimates interest at two possible prices, while a limited selling trial records actual purchases and returns. This evidence can guide product design, pricing, choice of target segment and the scale of the launch. Operations can assess whether the material meets the promised standard; finance can compare likely sales with the development and production costs.

The manager should check relevance, reliability and timeliness. Were respondents likely buyers rather than convenient volunteers? Did they see the actual price and product specification? Are competitor offers different now? A transparent, recent study of the correct market may be more useful than a larger old study with unsuitable definitions. Research has an opportunity cost: the money and time could have funded another activity, and delaying a launch may let a competitor establish its product first.

Digital data can make research faster. Website analytics show where users abandon a purchase; search patterns can reveal interest; transaction records can identify repeat purchasing. A controlled comparison of two webpage designs can help isolate the design's effect if other conditions are similar. Yet clicks are not sales, online reviewers may have unusually strong opinions, and customers who use a website may differ from those buying through shops. Algorithms can reproduce bias in the data used to train or target them.

Ethical choices affect both evidence and trust. People should understand how their information will be used. Collecting more personal detail than the decision needs can increase privacy concerns and undermine participation. Managers must avoid treating a convenient digital measure as permission to exploit sensitive information or manipulate customers. This lesson uses ethical principles rather than assuming that every available dataset is appropriate to use.

Combining sources can reveal contradictions worth investigating. If a survey promises high interest but the trial records few purchases, the sensible response is to investigate the price, sample and trial conditions rather than average the two claims mechanically. A plan can use a staged launch, a spending limit and a review date so that new evidence changes the decision before the full budget is committed. Research informs judgement; it cannot remove uncertainty about competitors, future incomes or customer behaviour.