The basics of internet-based market research are similar to most market research surveys. Decide who to talk to. Find out how to get hold of them online and then design mechanisms to find the information required.

Online-surveys allow for more sophisticated surveys, more interesting question types than could be carried out either on paper or via the telephone. Unfortunately many researchers still stick with old-style methods that look boring and disengaging to modern audiences.

The speed of turnaround and availability of large sample sizes also make iterative, agile and longitudinal research possible, but there are still occasions when phone or face-to-face are better options.

Online samples for market research

Online surveys are the main means of collecting quantitative market research data on the Internet either via research panels - pre-recruited consumers who have signed up to take surveys - or via email invites, or messaging to already known contacts. Technology also allows pop-up surveys - so called 'river sampling' and event-based sampling to catch individuals in the 'flow'.

Using customer lists is easy, but might miss buyers of competitive products, so often panels are used to ensure a full market coverage.

Sample quality

For any lists or panels, a major factor to consider is the quality of the panel or list. Good panels have procedures for opt-in, consent for data protection, and processes for countering bots, hackers and professional survey takers to ensure the sample quality represents the market being investigated.

An unfortunate problem with panels is that they have become something of a honeypot for fake participents aiming to earn some money as a 'side-hustle'. While they might think they are being helpful, in practice they dilute the quality of the panel. Always check the quality of the responses and put in checks to try to weed out fake responses.

Difficult to reach groups

While panels are typically large enough and broad enough for consumer-level surveys, tight targeting such as for a specific buyer group, or to small geographic area means even large panels can struggle to have enough available contacts. For tight-groups big panels will often pool resources with other panel providers to cover a difficult to reach group.

What you pay for sample will depend on how difficult it is to reach your target rate, based on the 'incidence rate'. Low incidence groups will require more screening to find the right people and so are more expensive.

B2B panels

In business-to-business sectors, panels are difficult to build and manage and typically only exist for certain broad categories. People change job regularly (around 25% of a b2b contact list is out of date after a year). Specialist B2B panels do exist, often focused on well-researched groups like IT managers. However, high incentives draw in participants who 'role play' for the incentives available. Always check how individuals are recruited, and put in place strong quality checks to counter 'click-farms'.

For customer lists, or where quality is absolutely paramount or the subject matter is too detailed for general panels, telephone research may be the best option as experienced interviewers are able to navigate to the right contact and may be better than online panels.

Healthcare

Healthcare is a specialist area, typically with good panels able to access to different medical specialisms, but typically at high cost. As with B2B panels, check the data quality and be aware that healthcare ethics and the law over data protection may be much stronger than for other groups.

Using a panel - checks

All panel providers should provide clear estimates of the completes they believe they can get, and provide information about how the panel was sourced, how it is maintained, and how they track against rogue completes and quality measures in place for data protection.

Often modern panels pool participants which can give varying levels of quality. If you find a good panel ask them to say how much of the sample is their panel versus the pool.

Useful checks are to ask for a spread of responses across different times and days. This might slow down the fieldwork, but normally you would not expect a working person to be available to complete a survey during working hours. Similarly, a flood of responses at 4am would seem suspicious.

The pooled panels often compete to fill slots which can mean a dump of poor quality responses if you go too fast. The purpose of research is to represent a real market, so it should reflect real people's use of time.

Email lists - good practice

Relying on email lists, means being aware of legislation on data protection and consent (eg GDPR). Rules for contacting European individuals are strict, and may restrict your own use of your own lists.

A key factor in email lists is the response rate. The number of invites required is the (sample_size/response_rate). Businesses with close one-to-one relationships can get response rates of 20% or more. More distant consumer facing firms might get less than 0.5%. If the expected response rate is 5%, then to get 100 completed interviews 100/5%=2000 invites would need to be sent out. If the response rate is 0.5%, then 20,000 invites would need to be set out to get to 100 completes.

Online market research communities

A hybrid between email and panels, is for businesses to use their own online market research communities. Consumers or customers who have signed up specifically to take part in market research, and possibly with access to a research portal as a private panel.

Where a panel or a direct email list is not available, an open survey method might be used where people are contacted (by phone or email) and asked to take part or to spread a link in a 'snowball' sample where friends are encouraged to ask other friends to join in. This can be particularly useful for special interest groups like sports associations, or using social networking sites.

Synthetic data and AI personas

The arrival of AI means that tools like ChatGPT can act like real research participants, even completing questionnaires. This is known as synthetic data.

AI can generate answers and provides a useful first check on new ideas and to validate content to ensure it doesn't have obvious glitches. However, for surveys AI is like asking an actor to pretend to be someone. You get data, but it isn't clear if that data reflects reality. Tests suggest AI generated data might reflect average positions but it less useful for new ideas or new products, or more specialist areas where AI doesn't have the training to understand real customer needs.

---

Smarter online questionnaires

The sample is just one part of online surveys. You should also consider the questionnaire itself.

Research tools like Google Forms allow simple surveys to be carried out but can look cheap. A survey should reflect the communication and marketing standards of your business, as you are looking for customers or potential customers to take part.

Skips and calculations

A major strengths of full online survey tools is the additional control on factors like skip patterns, routing, validating answers, or using previous answers to generate new questions or prompts for tasks such as pricing research. The extra control increases the quality of the data.

Coping with languages

Similarly, for international research, survey tools make it easier to manage multiple languages and currencies, and can offering techniques such as predictive text and more interactive approaches to filling in a questionnaire - such as bot-style interfaces.

Collecting audio and video

In addition Internet research opens up new possibilities for research. Audio and video are obvious possibilities, but also location and other data might be collected with user consent. New question methods are also possible - such as menu-driven questionnaires and non-linear survey techniques, or mock-up virtual environments. Our Cxoice Survey Technologies and series on Rethinking Questionnaires shows some new techniques.

Thinking about iterative survey design

In addition, with larger sample sizes available at lower cost, the process of interviewing can become much faster and more iterative. Design-test-design process can be used to iterate and explore ideas with the same customers, or with different customer groups.

Despite the newer forms of questionnaire, it's still the case that an online survey needs to be sharp and to the point. Grid-type questions tend to grate on respondents if used excessively (try sliders for a more interactive approach, but still limit their use).

The only area this is not the case is in text-based or chat questions, where typically respondents are willing to give much longer and thoughtful answers than in any other form of market research. For this reason, don't just rely on automated text-analytics. Actually read the answers. In a sample of 1000 customers there will be at least 3-4 stand-out suggestions.

Include a final wrap up question to catch anything unsaid

Respondents also expect to be able to say what they think, so always give some space at the end of the questionnaire for extra comments. We think this should be compulsory for any survey you field

And note that even with good quality samples, people taking online surveys can be less engaged than when talking to an interviewer, and might be distracted by other tasks they are doing at the same time (eg TV or other conversations). Consequently, take some time to check data quality and consistency.

---

DIY online surveys

For regular standard market research, there are plenty of off-the-shelf packages for carrying out web-surveys - students often just use Google Forms - and there are sites such as Surveymonkey, which allow you to field simple questionnaires quickly and cheaply, sometimes even providing you with sample too. Many have followed our lead on the use of questionnaire wizards to get you started with pre-prepared surveys for specific topic areas.

However, the fact that web-surveys open up a range of different questioning and investigation techniques that makes Internet research such a rich area. Increasingly, surveys can be tailored to fit very specific research and data needs, instead of trying to force the research to fit something standard.

The overall design of the questionnaire is also important - respondents are potential customers. Poor design leads to drop-out rates and can itself affect how customers feel. This is not just about the design of the questions, but also making sure that the look of the survey is also properly designed and will work on the wide range of browsers that are in-use.

Automating analysis and reporting

Once data are collected, analysis and reports can be produced automatically. Our Cxoice Survey Technologies provide live-data presentations where charts are created automatically (under the survey designers control) where the data is still accessible for drill-down on-the-fly, and not just a set of static charts.

This enables the reporting and charts themselves to be filtered for different groups (eg regions), and can be used to automatically filter to different groups, and produce automated summaries. Static Powerpoint and pdf type charts can also be generated automatically directly from the live-data presentation itself.

---

Qualitative research

Qualitative research is now much more common to be done online after the Covid pandemic, however, for qual, person-to-person contact is still very important, so face-to-face is still very common.

The wide availability of video conferencing and webinar platforms makes it relative straightforward to run in-person online groups and most of the inline meeting tools already come with good controls for presenting information, running polls and managing contacts and chat behind the scenes.

AI will generate transcriptions or in-the-flow subtitles, and even live translations for international projects.

Qualitative preparation still matters

However, preparation is still required. The core elements of a discussion guide and prepared elements like votes, demonstration or show material and even basics like ensuring the meet and greet and information over consent and data use are presented. Test runs internally can make the full customer contact version much smoother from the start.

As with online surveys, online respondents can be distracted by other elements in their environment. Ask people to turn off background distractions like TVs and to find a quiet space for the discussion. Use people's names to invite comments and keep people attentive, and mute out individuals not talking to reduce background noise.

Discussion forums and groups

Discussion forum and chat-based groups or communities are also options without needing full face-to-face and where discussion run over a period of time, allowing thoughts, tasks and debate to develop over a period of time without needing to have everyone all online at the same time. Engagement is key to keep people responding and taking an interest.

Qual can also be used in accompanying tasks via mobile devices. For instance, walking through a shopping experience live with a customer on a phone. Be cautious about collecting incidental data from contacts with other people.

Market research communities

Online conversations can develop into community research forums as a type of social networking space for commenting, forum and messaging between users plus polls and information updating to test reactions. These Market Research Online Communities (or MROCs) have become a major method for companies to build their own internal panel of customer for research over the past few years.

More advanced qualitative techniques include things like delphi research where experts are invited first to complete a survey and then after everyone has completed the survey, they are shown the results and invited to change their response or justify why their answers might be different from the consensus.

---

Social listening and text-analysis

An extension of qualitative and quantitative research takes the form of listening in to the natural conversations that people are having about products and services online, and using text analysis to drill down and capture insights.

Listening to online chatter has been around since the mid-1990s, but the volume now available means that huge quantities of online data and information are now available on 'natural' online conversations that people are having about products and services. The amount of data means automated tools like text-analytics and sentiment analysis are now available to be able to capture, track and analyse this information.

A subtle danger is that the loudest or most eloquent opinions can drown out the views of the wider customer base, so some caution is needed as online opinions may not fully reflect cross-market views. However, as systems of feedback, these can be very fast and helpful methods for capturing market insight, with the possibility of researchers dropping in questions or testing the waters to uncover insights about new ideas.

---

For help and advice on carrying out on-line market research contact info@dobney.com