the need for and limitations of sampling

3.2 Market Research – Sampling

3.2.1 Purpose of Market Research

Market research supplies the information a business needs to make informed decisions about:

• Market size and growth potential
• Key competitors and their strategies
• Customer characteristics, needs and buying behaviour

Sampling is the technique that enables these objectives to be achieved without having to study every single member of the market.

3.2.2 Primary vs. Secondary Research

	Primary Research	Secondary Research
Definition	Data collected directly from respondents for the specific research project.	Data already published or collected for another purpose (e.g., reports, statistics, company records).
Advantages	Tailored to the exact research question More up‑to‑date and specific Allows measurement of attitudes, intentions and perceptions	Cheaper and quicker to obtain Large volumes of data often available Useful for background and benchmarking
Limitations	Higher cost and longer time‑frames Requires design of questionnaires or observation schedules Sampling decisions affect reliability	May be outdated or not directly relevant Limited control over data quality Potential copyright or access restrictions

Sampling is a core component of primary research because it determines which members of the population are actually surveyed.

3.2.3 Why Use Sampling? (Need for Sampling)

• Cost efficiency – surveying the whole population is often prohibitively expensive.
• Time saving – data can be collected and analysed more quickly from a sample.
• Practicality – some populations (e.g., all potential customers worldwide) are impossible to reach in full.
• Manageability – smaller data sets are easier to code, analyse and interpret.

3.2.4 Types of Sampling

Category	Method	Key Features	Typical Business Use	Primary‑Research Scenario
Probability Sampling	Simple Random	Every element has an equal, known chance of selection.	When results must be statistically generalisable.	Randomly selecting 500 customers from a national database to test a new loyalty card.
	Systematic	Select every k‑th element from an ordered list.	Large, well‑ordered frames (e.g., customer list).	Choosing every 10th name on an alphabetical mailing list for a product‑trial invitation.
	Stratified	Population divided into homogeneous strata; random sample taken from each.	Ensures representation of key sub‑groups (e.g., age, region).	Dividing a market into age groups (18‑24, 25‑34, 35‑44) and sampling proportionally for a new smartphone launch.
	Cluster	Whole clusters (e.g., stores, schools) are randomly chosen; all units within selected clusters are surveyed.	Reduces travel costs when the population is geographically dispersed.	Randomly selecting 20 retail outlets from a national chain and surveying every shopper in those stores.
Non‑Probability Sampling	Convenience	Samples are taken from respondents who are easiest to reach.	Pre‑testing questionnaires, exploratory work.	Distributing an online survey via the company’s Facebook page to gather quick feedback on a packaging redesign.
	Judgment (Purposive)	Researcher selects respondents believed to be most representative.	Expert panels, niche markets.	Interviewing senior chefs to understand trends in high‑end restaurant menus.
	Quota	Samples are filled to meet pre‑set quotas for key characteristics.	Ensuring gender, age or income balance when random sampling is impractical.	Recruiting 100 males and 100 females, each split across three income brackets, for a brand‑awareness study.
	Snowball	Existing respondents recruit further participants.	Hard‑to‑reach groups (e.g., specialised professionals).	Starting with a few senior engineers and asking them to refer other engineers for a survey on new CAD software.

3.2.5 Limitations of Sampling

• Sampling error: The difference between the sample estimate and the true population value. Expressed as a margin of error (ME).
  Simple random‑sample formula:
  $$\text{ME}=z\;\sqrt{\frac{p(1-p)}{n}}$$ where z = confidence‑level multiplier, p = sample proportion, n = sample size.
• Bias: Systematic distortion caused by non‑random selection, poorly worded questions, or interviewer influence.
• Non‑response bias: Certain groups are less likely to reply, making the sample unrepresentative.
• Coverage error: The sampling frame does not include all elements of the target population (e.g., an online panel excludes people without internet access).
• Limited depth: Small samples may miss rare but strategically important segments.
• Sampling‑frame quality: An out‑of‑date or duplicate‑laden list creates coverage error and inflates bias.
• Margin‑of‑error assumptions: The simple‑random‑sample formula above is not valid for stratified, cluster or quota samples; those designs require design‑effect adjustments or separate calculations.

3.2.6 Reliability and Validity of Data

• Reliability – the extent to which a measurement yields consistent results.
  • Test‑retest reliability: repeat the same questionnaire after a short interval.
  • Internal consistency: use Cronbach’s α for multi‑item scales.
• Validity – the degree to which the instrument measures what it intends to measure.
  • Content validity: questionnaire covers all relevant aspects of the construct.
  • Construct validity: questions relate logically to the underlying concept.

3.2.7 Mitigating Limitations

1. Develop a comprehensive, up‑to‑date sampling frame that truly reflects the target market.
2. Choose an appropriate sample size; larger samples reduce sampling error.
3. Apply randomisation (or stratification) to minimise selection bias.
4. Use follow‑up contacts, incentives or mixed modes (online, face‑to‑face) to lower non‑response.
5. Weight the data to correct for over‑ or under‑represented groups.
6. Pre‑test the questionnaire to check wording, order effects and reliability.

3.2.8 Choosing and Interpreting Visualisations

Effective visualisation helps students interpret market‑research data and satisfies the Cambridge requirement to read tables, charts and graphs.

• Bar / column charts – best for comparing frequencies or percentages across categories (e.g., brand preference by age group).
  Key points: label axes clearly, start the y‑axis at zero, use consistent colours.
• Pie charts – useful for showing parts of a whole (e.g., market‑share distribution).
  Key points: limit to 5–6 slices, label each slice with percentage, avoid 3‑D effects that distort perception.
• Line graphs – ideal for displaying trends over time (e.g., sales growth month‑by‑month).
  Key points: use regular intervals on the x‑axis, plot data points or markers, include a legend if multiple lines are shown.
• Stacked bar charts – show how sub‑categories contribute to a total (e.g., revenue by product line within each region).
  Key points: colour‑code sub‑segments, ensure totals are readable.
• Always add a concise title, source note and clearly labelled axes or legends. When interpreting, comment on the direction, magnitude and any notable outliers.

3.2.9 Interpreting Quantitative & Qualitative Data – Checklist

• Verify the sample is representative (compare demographics with known population data).
• Check confidence intervals and margins of error for key percentages.
• Identify outliers or inconsistent responses; decide whether to retain or exclude them.
• Cross‑validate findings with secondary data where possible.
• Present quantitative results using the appropriate visualisations (see 3.2.8).
• Summarise qualitative insights (themes, quotes) in tables, mind‑maps or short narrative paragraphs to complement the numbers.

3.2.10 Example Calculation

Suppose a company wants to estimate the proportion of customers who would switch to a new product. A simple random sample of 400 customers yields 180 favourable responses.

Estimated proportion:

$$p = \frac{180}{400} = 0.45$$

At a 95 % confidence level (z = 1.96), the margin of error is:

$$\text{ME}=1.96\sqrt{\frac{0.45(1-0.45)}{400}}\approx0.048$$

Confidence interval:

$$0.45\pm0.048\; \text{or}\; 40.2\% \text{ to } 49.8\%$$

Interpretation: The company can be 95 % confident that between 40 % and 50 % of all its customers would consider switching.

3.2.11 Real‑World Example of a Sampling Limitation

Convenience sampling is quick and cheap, but it can produce a bias that over‑represents tech‑savvy respondents. For instance, an online survey posted only on a brand’s social‑media channels may miss older customers who rarely use the internet, leading to an inflated estimate of interest in a new app.