How to create a survey

How to create an online survey

Benefits and limitations of an online survey

Online survey is a type of survey where respondents answer questions on their personal device — a computer, tablet, or smartphone. Survey invitations can be sent out to an existing contact list or to participants of a survey panel.

Advantages of an online survey

Low costs and low labor intensity required. Online survey is the most budget-friendly quantitative survey tool. Digital technologies help researchers save on the printing, work of the interviewer and the data entry operators. It also prevents human error in data entering.

Quick response. It takes respondents less time to complete a survey on their own than with an interviewer. Over 60% of responses come within the first two days after the invitation. In addition, online survey services process the results in real time.

Wide audience coverage. Online surveys allow researchers to collect responses from remote area residents in a short time.

No interviewer influence. Since respondents complete the survey independently on their own devices, their responses are not affected by an interviewer.

Display of visual materials. Modern survey builders enable you to demonstrate images, audio, and video files to respondents. This way you can test different designs, video ads, etc.

No time restraints to surveying. The respondent decides when and where to complete the survey.

Study in sensitive topics. Online survey guarantees anonymity to the respondents, which allows you to collect honest answers even on delicate topics.

The incoming data can be processed automatically. No need to enter each survey into a table yourself — online services calculate survey results automatically.

Limitations of an online survey

Low response rate. When potential respondents are asked to participate in a survey by an interviewer, they might be more hesitant to decline in order not to seem rude or indifferent. With online surveys, this is not a factor. Therefore, you should think about how to motivate respondents in advance: it may be a present, a discount or cash.

No random sampling available. First, respondents have to have an internet connection to be able to participate in an online survey. Thus, a part of general population is excluded. Second, the researcher has no control over who agrees to accept the invitation to the survey. Online surveys are commonly completed by people interested in certain goods, who like them, or have enough time to participate. As a result, we get accidental sampling rather than random sampling.

Low control over responses

As respondents complete a survey on their own, it is impossible to monitor how much attention they pay to questions. In order to verify the answers, the researchers commonly use control questions and trick questions.

Types of online survey

3 things to do before running a survey

Before creating a survey, you should define research goal and objectives. The goal is the main question asked by the research. It defines the results you want to see in the end. Objectives elaborate the goal and answer the question “How to achieve the result?”. When composing each survey question, you should check whether it works towards the defined goal.

Another important aspect is the choice of the target survey audience. It usually corresponds to the product’s target audience. Target audience can be described in terms of demography, geography, or target needs.

  • Demography has significant influence on the person’s preferences. Additionally, it is a convenient parameter to select respondents with a survey panel.
  • Geography is your target market. If you work on the international market, conduct separate surveys in different countries as the market situation may significantly differ from country to country.
  • The target need can be detected with a direct question like “Did you have a need for X in the last X months?”, or behavior questions like “Have you done X in the last X months?”.

Once the survey is composed, you should do pilot testing. It will help you check how well it works with collecting primary data, so that you can elaborate or adjust it.

First, the survey tools are tested with an in-depth interview. The researcher asks interviewees to complete the survey and then describe which parts caused difficulties in their own words. 15–30 interviews with target audience representatives will be enough to take see the survey from the respondents’ perspective and identify flawed questions.

10 tips on how to make a survey

Nowadays, target audiences are surveyed not only by sociologists but also by specialists without relevant education: product managers, business and UX analysts, PR and HR managers. At first sight, conducting a survey seems simple. However, mistakes often come to light when responses are already collected: part of questions was answered with “Other”, a part of respondents refrained from answering a question by choosing “Do not know / No answer” option for some reason, and some respondents gave answers not matching the question.

How do you create a survey to get the most information out of it? Here are 15 recommendations to avoid most common mistakes and make the survey an accurate data collection tool.

Think through all possible answer options

You should check that the suggested answer options cover the entire range of all possible situations. It shouldn’t happen that a respondent has no suitable answer option. If you doubt the answer list is comprehensive, add an “Other” field for entering the answer in free form for you to add options into the list of answers later.

Think through all possible answer scenarios

Think through how many answer options can respondents choose per each question to prevent a situation where a respondent can only give one answer while, in fact, several options apply. For instance, if you ask, “How do you get to work?”, keep in mind that a respondent may use different means of transportation on their way to work: they can walk first, then take a bus, and use subway.
Pro tip: PeakPoll survey builder keeps track of the order in which respondents select answer options.

Use different types of questions and research methods

Questions should be diverse and grab respondents’ attention. Alternate between closed-ended, partially closed-ended, and open-ended questions. Use the classic question sequence: what, where, when, who, how, and why. Ask questions about different matters: facts, emotions, desires, needs, problems, time, and means. Then it will be interesting for the respondent to answer.

Be careful with personal questions

You shouldn’t ask questions about marital status, income and political views right from the get go. Such questions in the very beginning of a survey may put the respondent on guard and deter them. It is better to ask about socio-demographic details in the end. Having answered most questions, the respondent will be more likely to complete the survey. When asking personal questions, try to be tactful in your wording and include the “Do not know / No answer” option as a sign of respect towards respondents.

The simpler, the better

The phrasing of the questions should be simple, clear, and unambiguous. Avoid using professional jargon, academic terms, abbreviations, acronyms and slang. For words like “subscription”, “loyalty program”, “turn-key”, it is better to put an explanation next to them in brackets.

One question = one question

Don’t try to combine two questions into one. The respondents are likely to answer only one of them: the one that they have an opinion on. It is better to divide such questions into several semantic parts to avoid confusing respondents.

Don’t include too many open-ended questions

A survey is a quantitative research tool that asks participants to select answer options from a closed-ended list. Open-ended questions require more efforts from the respondent, so don’t put too many such questions, unless necessary. In the end of a survey, it is reasonable to add one open-ended question for the respondents to share their thoughts on the research subject. This way you will be able to cover “gray” survey areas and identify unexplored topics.

Pay attention to the wording of hints

Hints should be homogeneous: of roughly the same length, with the same level of detail, and based on the same characteristic. Intervals of values in hints should not overlap. Make sure that each hint is unique and does not incorporate values of neighboring hints. In case of evaluative questions, hints should be well-balanced: two hints pro and two hints against. For example, “very good”, “good”, “neither good nor bad”, “bad”, “very bad”.

Do not encourage grammar Nazis

Spelling and punctuation in a survey reflect on your company’s reputation. A carelessly written survey may be perceived as a sign of the same attitude to clients. Before launching a survey, proofread it twice yourself and ask your colleagues to check it with a fresh pair of eyes.

Trust, but verify

If you want to check how well respondents pay attention to a survey, add several trick questions. For example, ask the respondents to rate non-existent goods. If a respondent give it a rating instead of selecting the “Do not know / No answer” option, that should cast doubt on the integrity of their previous answers.

10 tips on how to create an effective online survey using Peakpoll pro tools

Greet the respondents

Add a preamble on the welcome page before the start of the survey so that respondents better understand the context of the research. In a couple of sentences, describe the survey topic, and how the results will influence the customer experience. That can also help motivate respondents to participate, clarify terms and definitions used, and filter out non-relevant audience.

Think through the question order logic

The order of questions in a survey should be consistent and coherent. For example, if a respondent answers “Never” on the question “How frequently do you buy X?”, you shouldn’t ask them to rate these goods next. Instead, you should redirect them to the next question cluster.

People are extremely sensitive to this parameter. If they see that questions don’t follow a clear logic, they get a feeling that the survey is absurd, and their answers are irrelevant.

PeakPoll survey builder enables conditional branching to hide irrelevant questions. Use Logic rules and Display rules to specify the conditions under which a question or a question cluster will be shown to a respondent.

Use the available respondent details

You can also tweak question display logic based on additional parameters in the survey link. For example, you are mailing a survey to an existing contact list where you specify delivery method: in-store pick-up, pick-up point, or door delivery. You can ask the respondents who ordered door delivery more than 3 times in the last half-year to rate the quality of courier services.

Randomize answer option order

When listing hints, keep in mind that people tend to choose the first option more often than others. To prevent biased results, set random display of answer options. However, this method doesn’t fit all questions. For example, if you ask about purchase frequency, the answers should be ordered from the most frequent to the least frequent options. Such options as “Other” and “Do not know / No answer” shall always be at the bottom of the list.

Use different types of questions

Use different types of questions and research methods. For example, semantic differential method (Scales question type) allows you to quickly and subtly measure the perception of an object. To do this, ask respondents to evaluate an object using bipolar scales from -3 to +3 where the two opposite end points are represented by antonyms. Paired comparison method can be used to test design options of user interface elements (for example, buttons).

Customize survey design

In survey settings, you can customize the colors of key survey elements (background, text, buttons) or use preset color schemes, for instance, to design a survey in brand colors.

Add visual elements

In order to make a survey even more engaging, add images (company logo, icons as answer options), audio, or video files. With PeakPoll survey builder, you can replace scale points with a color gradient, stars, emojis or custom images.

Motivate respondents to participate

Thank respondents for participating in a survey. You can reward them with a discount or additional services when making a purchase. As long as the bonus is of interest to your target audience. You can motivate respondents to complete a survey by stating on the welcome page that they will get a gift if all questions are answered. Insert a general promo code or upload a list of personalized promo codes on the finish page. The displayed promo codes will be shown in the survey results for each respondent.

Test objects in subsamples

If you need to select the best option among others, in terms of research method it is better to evaluate each option in separate respondent subsamples. A/B or split testing can help you with this task. For this purpose, respondents are divided into equal groups in terms of number of objects. Each respondent tests only their option. It helps to reduce the strain on the respondent and smooth the priming effect (when the perception of object B is affected by the perception of the previously seen object A).

Display a random set of answer options

Perception and comparison of 10–15 options can be cognitively challenging. If you need the respondents to choose the best option from a long list of objects (for instance, jewelry designs), it’s best to enable the display of 3–5 random choices for each respondent. This way you will disperse the cognitive load among all survey participants and get better data.

Set automatic insertion of previously selected answers in the subsequent questions

Suppose you want to know what people in your neighborhood think about the offering and the quality of service in local coffee shops. You can first give respondents a list of places and ask them to choose the outlets they have visited at least 2–3 times over the past 3–6 months. Then, having a corresponding setting on, you can ask them to rate the answer options from the previous question on a scale from 1 to 10. This way you reduce the number of objects to be assessed and avoid dead-end situations where respondents have to rate a place they’ve never been to.