Online Surveys and Online Data Collection: How to Get Fast Research Data You Can Actually Trust

Why online data collection became the default

Twenty years ago, most quantitative research meant interviewers with clipboards or telephone banks. Today, a large share of surveys worldwide are completed online, often on a phone, in a few minutes. Researchers call this CAWI: computer-assisted web interviewing.

The reasons are easy to see. Online surveys can reach thousands of people in days, across many countries, at a fraction of the cost of fieldwork. Respondents answer when it suits them, which tends to produce more honest answers on sensitive topics. Data arrives in a structured form, ready for analysis.

But speed has a downside. It is just as easy to collect thousands of poor-quality responses as good ones. This guide explains how online data collection works when it is done properly, and what to check before you trust the results.

When online surveys are the right choice

Online works well when your target audience is online and reachable through panels or your own lists, when you need quantitative answers (how many, how much, which option), and when timelines are short. Typical uses include:

• Concept, product and packaging tests
• Brand awareness and tracking studies
• Customer satisfaction and NPS research
• Usage and attitude studies
• Pricing research, including conjoint and Van Westendorp
• B2B surveys with professionals and decision-makers
• Healthcare professional surveys, through verified panels

Online is weaker when your audience has limited internet access, low digital literacy or is very hard to verify online. Rural households, older citizens and some trade audiences are examples. In those cases, a mixed approach that adds offline data collection such as telephone (CATI) or face-to-face (CAPI) interviews gives a fairer picture.

Where the respondents come from

The single biggest factor in online data quality is the source of respondents. There are three main options.

Your own customer list is ideal for satisfaction research, but it only tells you about existing customers. Research panels are communities of people who have agreed to take surveys in return for rewards; good panels profile members in detail, verify identities and remove poor-quality participants. River or social sampling recruits people in the moment through ads or website intercepts, which is useful for hard-to-find groups but harder to control.

Frequent Research runs its own panels, Turn2Opinion and MyFrequentRewards for consumers and business professionals, and MyHealthOpinions for healthcare professionals, patients and caregivers. Owning the panel means we control recruitment, profiling and quality scoring instead of relying entirely on third parties. For multi-country studies, we blend our own sample with vetted partners and manage quotas across sources with our Sampledge platform.

How many responses do you need?

Sample size depends on how precise you need the results to be, and on how many subgroups you want to compare. For a simple percentage estimate from a large population, the standard formula is n = z² × p(1 − p) ÷ e², where z is 1.96 for 95% confidence, p is the expected proportion (0.5 if unknown) and e is the margin of error.

At 95% confidence and a ±5% margin of error, that gives about 385 completes. At ±3%, you need about 1,067. Most importantly, those numbers apply to each group you want to report separately. If you need reliable results for four regions, you need a robust base in each region, not just overall.

Strictly speaking, these formulas assume random sampling, which most online panels are not. Treat margins of error from online quota samples as a guide, report them honestly, and use weighting to align the sample with the population. Our sampling & recruitment team helps set these targets before fieldwork begins.

Designing a questionnaire people will finish

A well-designed questionnaire is the cheapest quality-control tool there is. Long, confusing surveys produce bored respondents, and bored respondents produce bad data.

• Keep it short. Ten to fifteen minutes is a sensible maximum for most consumer surveys; less on mobile.
• Design for phones first. Many respondents will answer on a small screen, so avoid huge grids.
• Use plain, neutral language. Avoid leading questions, jargon and double-barrelled questions that ask two things at once.
• Screen carefully at the start, so only the right people continue.
• Put sensitive and demographic questions later, after trust has built.
• Always include a “don’t know” or “prefer not to say” option where appropriate.
• Test the survey on real devices and run a soft launch before full fieldwork.

Survey programming turns the questionnaire into a working survey with routing, piping and quota logic. Errors here are expensive, so every script should be checked against the questionnaire and tested with dummy data before launch. Our research design process builds this review in.

The quality problem nobody likes to talk about

Online research has a real fraud and attention problem. Some people take surveys only for rewards and rush through them. Some try to qualify for surveys they do not fit by guessing the screener. Automated bots and organised survey farms exist. Industry bodies such as ESOMAR have published guidance on assessing online sample quality precisely because the issue is widespread.

The answer is not to abandon online research, but to build layered quality control:

• At the panel: identity and contact verification, deduplication, and ongoing quality scores for members
• At entry: digital fingerprinting to spot repeat devices, geolocation checks, and bot detection
• During the survey: attention checks, trap questions and consistency checks
• After completion: removal of speeders, straight-liners and nonsense open-ended answers
• Before delivery: manual review of verbatims and final logic checks

Removed responses are replaced during fieldwork, so the final dataset meets its quotas. A trustworthy supplier will tell you how many completes were removed and why. If a supplier reports a zero rejection rate, ask questions.

Online vs offline: choosing the mix

Many of the best studies combine methods. An online survey covers urban, connected respondents quickly, while telephone or face-to-face interviews fill gaps with rural or older audiences. The questionnaire is adapted for each mode so results can be combined and weighted.

From raw responses to decisions

Collection is only half the job. The data then needs cleaning, coding of open-ended answers, weighting where needed and tabulation. Our data processing & analytics team delivers clean datasets in SPSS, Excel or CSV, with banner tables and significance testing.

Finally, reporting & insights turns tables into a story: what did we learn, how confident can we be, and what should the business do next? The best reports are shorter than you expect and specific about actions.

Questions to ask any online data collection supplier

Clear, specific answers are a good sign. Vague reassurance is not. Frequent Research is certified to ISO 20252 for market, opinion and social research and is a corporate member of ESOMAR, and we share our quality process openly. You can download our Quality Process document.

• Where exactly will respondents come from, and how are they recruited and verified?
• What share of the sample comes from your own panel versus partners?
• What quality checks run before, during and after the survey?
• How many completes do you typically remove, and will you report removals?
• How do you handle incidence and feasibility for niche audiences?
• Can you combine online with CATI or face-to-face if needed?
• How is respondent data protected, and are you aligned with recognised industry codes and standards?

What is different about online research in India

India is one of the most interesting markets for online research, and one of the most demanding. Smartphone use has grown enormously, so online surveys now reach far beyond large cities. But access, language and comfort with online forms still vary widely by region, age, income and gender.

In practice, that means a few adjustments. Surveys should be offered in Hindi and relevant regional languages, not just English. Quotas need care, because an unmanaged online sample will over-represent younger, urban, higher-income men. Mobile-first design matters even more, since many respondents will never see a desktop version. And for studies that must represent the whole population, online usually needs to be combined with telephone or face-to-face interviewing.

None of this is a reason to avoid online research. It is a reason to design it thoughtfully, with a partner who understands where online coverage is strong and where it needs support.

Respondent experience and fair incentives

Respondents are not data points; they are people giving their time. When surveys are long, repetitive or broken on mobile, good respondents drop out and the remaining sample is skewed towards those willing to put up with anything, often for the wrong reasons.

Fair, clearly explained incentives, honest length estimates, a progress bar, and a thank-you at the end make a measurable difference to completion rates and data quality. Panels that pay reliably and treat members with respect keep better members over time. That is why our panels publish clear reward thresholds and payment methods, and why we push back when a questionnaire is too long for its audience.

A worked example

Imagine a consumer brand that wants to test three new packaging designs with women aged 25 to 45 who buy the category monthly, across four metro cities.

First, feasibility: we check how many qualifying panel members exist in each city and estimate the incidence of monthly category buyers. Next, design: a monadic test, where each respondent sees one design, with about 150 completes per design per region group, so differences can be tested. The questionnaire is kept to about twelve minutes, programmed for mobile, translated where needed and soft-launched with 10% of the sample to check timing and logic.

During fieldwork, quotas are tracked daily, speeders and inconsistent responses are removed and replaced, and open-ended comments are reviewed. At the end, the client receives a clean dataset, tables with significance testing, and a short report recommending which design to take forward and why. Total time from brief to report for a study like this is often two to three weeks.

Protecting respondent data

Online surveys collect personal information, sometimes sensitive information, so data protection is part of quality. Respondents should know who is running the research, how their answers will be used and that they can stop at any time. Identifiable details should be kept separate from survey answers, and results reported in aggregate.

Our approach is explained in our Privacy Policy and Research Ethics pages, and supported by our information security certification.

Getting started

If you are planning a survey, the fastest way to get useful advice is to share your objective, target audience, markets and ideal number of responses. We will come back with feasibility, a recommended method and a cost, usually within one working day.

You can request research feasibility or read our complete guide to market research in India for the wider picture.

Frequently asked questions

• What is online data collection?

  It is collecting research responses over the internet, usually through online surveys completed on a phone or computer. Respondents typically come from research panels, customer lists or targeted online recruitment.

• How fast can an online survey be completed?

  For common consumer audiences, fieldwork can finish in a few days. Niche B2B or healthcare audiences, or multi-country studies, usually take one to three weeks. We confirm timelines after checking feasibility.

• How do you stop fake or low-quality survey responses?

  We use verified panels, device and location checks, attention and consistency checks, removal of speeders and straight-liners, and manual review of open-ended answers. Removed responses are replaced during fieldwork.

• Can online surveys reach doctors and business decision-makers?

  Yes, through verified professional panels. Our healthcare panel verifies healthcare professionals, and our business panel profiles members by role, seniority, industry and company size.

• What sample size do I need for an online survey?

  About 385 completes gives a ±5% margin at 95% confidence for a simple estimate, but you also need enough responses in each subgroup you want to report. We help set sample sizes based on your objectives.

Still curious? Ask the assistant Talk to a specialist →

Explore our Market Research services