Quantitative Research: Measuring What Matters with Surveys You Can Rely On

What quantitative research is for

Quantitative research turns opinions and behaviour into numbers that can be compared, tracked and acted on. How many people know your brand? What share of customers would recommend you? Which of three concepts is most appealing? How much would people pay? Which factors drive satisfaction most?

It relies on structured questionnaires answered by enough people to allow statistical analysis, typically hundreds or thousands. Done well, it gives decision-makers confidence: not just a sense of what people think, but how widespread each view is and how it varies across segments.

This guide explains how quantitative research studies are designed and delivered, and how to judge whether the numbers deserve your trust.

Common types of quantitative study

Many of these work best alongside qualitative research, which explores the reasons behind the numbers.

• Usage and attitude (U&A) studies: the full picture of a category, who buys what, how and why
• Brand health and tracking: awareness, consideration, image and usage over time
• Concept and product testing: which ideas, products or packs perform best
• Advertising pre- and post-testing: whether communication works
• Customer satisfaction and NPS: experience, loyalty and drivers
• Pricing research: willingness to pay and price sensitivity
• Segmentation: grouping customers by needs, attitudes and behaviour
• Market sizing inputs: incidence, penetration and purchase volumes

Who to survey: defining the population

Every quantitative study starts by defining the population it represents: for example, “adults aged 18–55 in metro and tier-1 cities who bought packaged snacks in the past month.” This definition determines who qualifies, which screening questions are needed and how results should be described.

Being precise prevents a common problem: results presented as representing “consumers” or “businesses” in general, when the sample actually represents a narrower group. Clear definitions also make trackers comparable over time.

Sampling and sample size

The sample must reflect the population on the characteristics that matter, usually through quotas on age, gender, region and other variables, followed by weighting to correct small imbalances. Sources include online panels, customer lists, telephone samples and face-to-face fieldwork; our sampling & recruitment team advises on the right mix.

Sample size depends on the precision needed and the subgroups to be analysed. For a simple percentage, around 385 responses give a margin of error of about ±5% at 95% confidence, and around 1,067 give about ±3%. These figures assume random sampling and apply to each subgroup separately, so a study comparing four regions needs a robust base in each region.

Choosing a data collection mode

Our guides to online data collection and offline data collection explain each mode in more depth.

Designing the questionnaire

The questionnaire is where most quantitative errors begin. Questions must be clear, neutral and answerable; response options must be exhaustive and not overlapping; and the order must avoid priming people before key questions.

• Ask about specific, recent behaviour rather than general habits
• Avoid leading wording and double-barrelled questions
• Use consistent scales and label them clearly
• Include “don’t know” where people genuinely may not know
• Rotate lists and concepts to avoid order effects
• Keep it as short as the objectives allow, especially on mobile

A pilot or soft launch with a small share of the sample checks timing, understanding and routing before full fieldwork. It is the cheapest insurance in research.

Quality control during fieldwork

Quality is protected throughout fieldwork: daily quota monitoring, removal of speeders and straight-liners, attention and consistency checks, open-ended review, and for interviewer-led modes, supervision, call monitoring and back-checks. Rejected interviews are replaced so the final sample meets its design.

A trustworthy partner reports these outcomes openly: how many interviews were removed and why. Frequent Research is certified to ISO 20252 for market, opinion and social research, and you can download our Quality Process.

Cleaning, coding and weighting

After fieldwork, data is cleaned according to agreed rules, open-ended answers are coded into themes, and the data is weighted where needed so that the sample matches the population on key characteristics. Weighting should be modest; large weights signal that some groups were under-represented and increase uncertainty.

Our data processing & analytics team delivers clean datasets in SPSS, Excel or CSV, with documentation of every cleaning and weighting decision.

Reading the numbers: significance and meaning

Tables usually show results by key subgroups with significance testing, which indicates whether a difference between groups is likely to be real rather than random variation. But statistical significance does not mean importance. With large samples, tiny differences can be significant yet irrelevant to the business; with small samples, meaningful differences may not reach significance.

Good analysis looks at the size of differences, the consistency of patterns across related questions, and what the numbers mean in practice. Small bases should always be flagged.

Advanced analytics

Some questions need more than tables:

• Driver analysis: which attributes most influence satisfaction, preference or loyalty
• Conjoint analysis: how people trade off features and price, and which product configuration maximises appeal or revenue
• MaxDiff: ranking many items, such as claims or features, by importance more reliably than rating scales
• Price sensitivity methods such as Van Westendorp and Gabor-Granger
• Segmentation using cluster or latent class analysis
• TURF analysis: which combination of products or flavours reaches the most people

These methods require careful design at the questionnaire stage, so they should be planned from the start rather than added after fieldwork.

Pricing research in practice

Pricing is one of the most valuable and misunderstood uses of quantitative research. Asking “how much would you pay?” directly tends to produce unreliable answers. Better methods present realistic choices: conjoint exercises where respondents choose between product options with different prices, or structured price-acceptance questions that reveal ranges rather than single points.

Results are most useful when combined with business economics, such as costs, margins and competitor prices, to model revenue and profit at different price points, not just preference.

An example: a category usage and attitude study

A food company entering a new category wants to understand who buys, how often, through which channels, which brands they use, what they value and where gaps exist. The study surveys 1,500 category buyers across six regions, online in cities and face-to-face in smaller towns, with quotas by age, gender and region.

The questionnaire covers purchase behaviour, brand awareness and usage, attribute importance and brand ratings, with a MaxDiff exercise on product benefits. Analysis produces a needs-based segmentation, a map of brand positions and a driver analysis showing which benefits matter most for each segment. The result is a clear picture of where the company can compete, and with whom.

Reporting that leads to decisions

Quantitative reports should lead with answers, not methodology. A short executive summary states what was learned and what to do; the main report supports each conclusion with clear charts; and an appendix covers method, sample and detailed tables. Dashboards are valuable for trackers and large studies where many teams need to explore results.

Our reporting & insights team builds reports and dashboards around the decisions they support, and presents them in a debrief with the client team.

Common quantitative research mistakes

Most are avoided with careful design and experienced review before fieldwork starts.

• Vague objectives, producing long questionnaires with little focus
• Samples too small for the subgroups that matter
• Leading questions and unbalanced scales
• Asking people to predict behaviour they cannot predict
• Treating every statistically significant difference as important
• Changing questions between tracker waves

Customer satisfaction and NPS done properly

Satisfaction and Net Promoter Score (NPS) programmes are among the most common quantitative studies, and among the most often misused. A single score tracked over time says little on its own. The value comes from understanding what drives the score, which customer groups differ and which actions improve it.

Good programmes survey customers at meaningful moments, such as after onboarding, a support interaction or renewal, keep questionnaires short, include an open question asking why, and connect results to operational data. Driver analysis then shows which aspects of the experience matter most, so improvement effort goes where it counts. Closing the loop, following up with dissatisfied customers, turns measurement into action.

Trackers: consistency over time

Tracking studies measure the same things repeatedly, monthly, quarterly or annually, to show how a market, brand or customer base is changing. Their value depends entirely on consistency: the same questions in the same order, the same sample definition and quotas, comparable sources and similar timing.

Changes to a tracker should be rare and deliberate. When a change is unavoidable, running old and new versions in parallel for a wave allows results to be bridged. Dashboards that update as each wave completes help teams spot movements quickly and link them to campaigns or market events.

Choosing scales and question types

Small design choices affect results. Five- and seven-point scales are common; the key is to label points clearly and use the same scale consistently across related questions. Balanced scales, with equal positive and negative options, avoid nudging respondents. Ranking long lists is hard, which is where MaxDiff helps; and open-ended questions, used sparingly, explain the numbers and often surface surprises.

Getting started

Share your objectives, target population, markets and timeline, and we will recommend a design, sample and mode, with a clear cost. Read our complete market research guide for the wider picture, or request feasibility.

Frequently asked questions

• Can surveys include images, videos or product concepts?

  Yes. Online and CAPI surveys can show images, videos and interactive stimuli; telephone surveys rely on clear verbal descriptions instead.

• How long does a quantitative study take?

  Simple online studies can take two to three weeks from questionnaire to report. Multi-country, mixed-mode or advanced analytics studies usually take four to eight weeks.

• Can you weight results to match the population?

  Yes. We weight to reliable population profiles where appropriate, keep weights modest and document the approach so results can be interpreted properly.

• What is quantitative research?

  Research that measures opinions and behaviour numerically, using structured questionnaires answered by enough people to allow statistical analysis and comparison.

• How many respondents do I need?

  It depends on precision and subgroups. Around 385 gives about ±5% at 95% confidence for a simple estimate, but each subgroup you analyse needs an adequate base too.

• Should I use online, telephone or face-to-face interviews?

  Online suits connected audiences and speed; telephone reaches people without internet; face-to-face gives the widest coverage. Mixed mode combines them when needed.

• What is conjoint analysis?

  A method that shows respondents realistic product options with different features and prices, revealing how they trade off attributes and what they are willing to pay.

• What does statistical significance mean?

  That a difference between groups is unlikely to be due to chance. It does not mean the difference is large or important, which must be judged separately.

• Can you run trackers over several years?

  Yes. We keep questions, sample definitions and quotas consistent across waves and provide dashboards that update as each wave completes.

• Do you provide raw data?

  Yes. Clean, weighted datasets in SPSS, Excel or CSV with a codebook and documentation of cleaning and weighting decisions.

Still curious? Ask the assistant Talk to a specialist →

Explore our Market Research services