MassMove report on transportation priorities in Massachusetts does not have sound methodology and should not be used as the basis for decision making. It’s a wonderful example of how to concoct a research program to deliver the “findings” the sponsor wants. Shame on the Mass State Senate.
Meaningful survey results require a valid questionnaire and an unbiased administration of the survey. The CTE brain trauma study of NFL players used biased survey samples, which clouds conclusions from the study.
In the survey research world, so-called non-response bias is an ever growing challenge for solid, viable research, even if many people simply ignore the threat. Election seasons present a great opportunity to explain the concept.
Why? Because non-response bias is a threat to credible survey research, but it’s the goal of every politician! Politicians want a specific result and creating a non-response bias is part of the process of winning an election, even if it’s a bit unsavory to say so.
Survey accuracy numbers tell us how much we should believe the survey results as an indicator of how group of interest feels. This article describes how to interpret accuracy numbers and the the factors that affect accuracy.
We offer a free Excel-based survey sample size calculator. Here we discuss how to use it and provide a form to request the calculator.
Summary: What survey mode is best to use? A recent Cint survey seemed to answer that question. 91% (sic) of respondents said they prefer web-based, smartphone, or SMS surveys. However, this survey suffers from a severe sample bias, given that the survey was likely conducted electronically. Just as response rates cannot be generalized across surveys done in different industries in different countries, so too the best survey mode or method cannot be generalized. Know your group of interest and what’s the best survey mode for your group.
~ ~ ~
Ever seen survey results that are too good to believe? Well, as the old adage goes, if it’s too good to believe, then you shouldn’t believe it. I call these — with all due deference to Homer Simpson — “D’oh! Results.” Even Homer in hindsight knew he should have seen that coming. Yet, it seems that many people accept survey results without critically thinking about them. I have some upcoming articles on the blind acceptance of the Net Promoter Score® and the Customer Effort Score, but I recently came across an example that displays a survey administration bias that should be obvious to even a casual reader.
Cint recently published the results of a market intelligence study of theirs. (Note: Cint removed the press release, otherwise I would link directly to it.) The study had several interesting findings, among them a discussion of survey methodology preferences by the public.
In terms of methodology, new technology was by far the most popular means of undertaking market research, with over 91% stating their preference as “smartphone”, “web” and “SMS”. Only 4% of those surveyed would choose to undertake market research by mail as their first choice, and just 1% would like to be surveyed over the telephone.
Wow, 91%!! Nowhere in the article is the methodology of their survey presented, but the first sentence of their online article give a pretty good hint.
17 January 2012: A survey out today from global provider of smart solutions for gaining market insight, Cint…
I sent an inquiry to Cint asking about their methodology, but to date I have received no response. (I actually find that fact alone very interesting for a firm engaged in customer engagement.) Anyone want to make a bet on the methodology that Cint used in their study? I am pretty confident that they surveyed using a “smart technology,” that is, a web survey.
If so, isn’t it obvious that people responding to web-based survey would say they preferred to be surveyed via the web? D’oh! Isn’t 91% an awfully high percentage? That number jumped out at me as so large that it didn’t pass a face validity test. It immediately made me skeptical of the results and want to learn more.
Sample bias such as this is not unusual. Years ago, I recall seeing a survey conducted at a conference on CRM technologies asking attendees about their senior management team’s involvement in CRM initiatives. Isn’t it senior management that approved of someone going to the conference? The administration bias is obvious. D’oh!
In fact, I am guilty of a similar bias. My Survey Workshop series spans three days. I wonder whether I should deliver the classes on Monday to Wednesday, Tuesday to Thursday, or Wednesday to Friday. I have a personal preference, but I really want to do what’s best for my market. So, on the feedback survey at the end of my class, I decided to ask which set of days is preferable. Guess what I find from this convenience sample, which is a common type of sample bias? If the class is a Monday-to-Wednesday class, that’s the preference. If it’s a Tuesday-to-Thursday class, that’s the preference. Etc. I have an obvious sample bias. D’oh! But at least I know that and can interpret my results in that light.
The Cint folks make no mention of the shortcomings of this aspect of their study’s findings. And it seems that others miss this critical point.
An American Express Open Forum article discussed the Cint study and included this synopsis of the impact of the survey methodology findings:
Get online. Consumers overwhelmingly prefer tech when it comes to surveys; 91 percent cite “smartphone”, “Web”, or “SMS” as their preferred methods. Just 4 percent like mail in surveys, and a mere 1 percent want to be surveyed by phone.
A Business Wings review of the article included:
If, after reading this, your business is now more likely to consider undertaking market research, you may wish to understand the preferred methodology. The clearest lesson is not to call people, as only 1% of those surveyed said they liked being contacted by telephone, compared to 91% who prefer to offer opinions via smartphone apps, SMS or online surveys.
Neither writer critically looked at how Cint’s survey methodology would impact these findings or seemed to question the astonishing 91% as unrealistic. D’oh!
As a university professor of 25 years, I can tell you that academics for years have stressed the need to build “critical thinking skills” among our students. We still have much work to do. D’oh!
Just as response rates cannot be generalized across surveys done in different industries in different countries, so too the best survey mode or method cannot be generalized. Know your group of interest or “population” in surveyor lingo. What’s the best survey mode for your group?
So how could you form a study that would properly understand survey methodology preferences? Here’s an example
Snap Surveys and Silver Dialogue conducted a study on multi-mode surveying methods. [Unfortunately the study is not available online anymore.] They queried visitors at National Trust properties in the United Kingdom. (These are historic properties many of which I have visited upon trips to Britain. I highly recommend them!) Visitors were asked by an interviewer upon entering the property if they would be willing to take a survey after their visit, and how they would prefer to be surveyed.
Here are their findings, quoted verbatim from their study:
- The most preferred questionnaire modes were paper/post 30% and web 26%.
- Point of visit methods, e.g. paper and kiosk, had far higher response rates (91%) than post visit, e.g. web and smartphone.
- Using a multi-mode approach increased the overall visitor survey response rate to 61% from around 45% achieved in previous paper based visitor surveys.
- A multi-mode approach reached a wider audience than previous surveys.
- Asking people to “pledge” to complete a survey after their visit helped to boost response rates.
Notice how these results conflict with Cint. Web garnered a 26% vote. The approach used here determined the best survey mode works for this transactional survey setting, but it may not work for all settings. I certainly would not apply their findings to other survey programs, and they didn’t claim you could. Now how confident do you feel in following Cint’s prescription?
Further, note that Cint queried about what survey mode people said they “preferred”. They didn’t conduct empirical research to measure the impact upon response rates of meeting someone’s preferred survey mode, as Snap did. The impact of survey mode upon response rates is what we, as surveyors, really want to know, along with the impact of survey mode upon the scores respondents provide. (I am currently co-authoring a paper on this latter topic. Stay tuned. It’s pretty dramatic.) Self-reported preference for survey mode isn’t the outcome of interest; the survey mode that leads to highest response rate is the outcome of interest — along with the most valid data.
Here’s another example of the distinction between survey mode preference and response rates. If students at universities were queried about how they would like to complete the end-of-semester course evaluations, I would suspect that they would want a web survey as opposed to the traditional scanable paper surveys administered in class.
Two places where I’ve taught, Harvard Certificate in Management Program, and Babson College, moved to a web-based survey method from the paper surveys administered in class. Hey, ya gotta use new these new smart technologies, right? Response rates fell in half. And they were surprised. And these are academics who are schooled in research design! D’oh!
Perhaps the topic most inquired in the field of surveys, especially customer satisfaction surveys, is response rates. What’s a good response rate? What’s the statistical accuracy with a certain response rate? But what is sometimes missed is that the response rate should not be the only concern. Bias in the response group should be an equally important concern. Lots of responses will give a good statistical accuracy, but it’s a false sense of security if there’s a sample bias. In particular, we’ll discuss survey fatigue and non-response bias.
Surveys can be conducted using many different technologies. Phone surveys, mail surveys and web surveys are the best known means. A more recent addition to this list is IVR surveys, also known as automated phone surveys. IVR stands for “Interactive Voice Response.” When you call a company for service, your phone call — and all those annoying phone menus — is managed with an IVR, and IVR technology provides the capability to conduct surveys. In this article I will examine the strengths and weaknesses of the IVR survey approach, and I will use a recent experience with a Sears IVR to measure customer satisfaction to illustrate some points.
What is an IVR survey?
An IVR survey shares many characteristics with a telephone survey since the survey is conducted over the telephone. But rather than have a live interviewer conduct the survey, the survey is delivered using a prerecorded script. The respondent is asked to enter his responses using the keypad of the telephone and by verbalizing free-form comments. (IVR survey programs can even analyze the tone of the responses.) Since there’s no interviewer to clarify questions, the script must be clean and anticipate all the possible responses someone being surveyed may have.
IVR surveys are most commonly used to measure a customer’s reactions to service just delivered through a call center environment, known as a transactional survey. You may have had this experience. When you call some company before being connected to an agent, you may be asked if you would be willing to take a short survey at the end of the call. When the service is complete and the agent disconnects from the caller, the caller is automatically transferred into the survey program. It’s important that the agent not know who is going to be surveyed to keep the measurements clean.
IVR surveys need not be limited to this situation. You could invite a respondent group to take an IVR survey by calling them or by emailing them with a toll free number to call to take a survey.
My recent experience with a Sears IVR survey shows the IVR method’s strengths and weaknesses.
What are the Strengths of the IVR Survey Method?
Speed of Survey Data Collection. The most striking advantage of the IVR survey method is how quickly a company can collect feedback from an individual. Within 5 minutes after a service interaction is finished, the company can get feedback from the customer about her experiences. No faster means of collecting feedback exists.
Service Recovery Flag. That completed survey from the respondent gets analyzed immediately by the IVR survey program. Management can establish criteria to flag a specific survey response for review. For example, if someone provides a very low rating for overall service quality, information about that service transaction and the subsequent survey can be forwarded to a designated manager. The customer could get a call immediately to address the issue. The immediacy helps ensure that the customer is by the phone to take the call.
Lower Survey Response Bias. Response bias results from a respondent’s predispositions when they enter into a surveying routine. Common ones are acquiescence where the respondent gives the answer he feels the surveyor wants and auspices where the respondent gives an answer to please the interviewer. Since the respondent is interacting with a machine, the likelihood of any response bias is much reduced. We may not understand the script, but few of us will be intimidated by the process. However, the flip side may prove true should a low score result in an immediate follow-up phone call by a designated manager. That call could prove intimidating to some — and make them leery to take future IVR surveys.
Lower Cost for Larger Survey Programs. IVR surveys have a cost model that has a high fixed cost, but lower variable cost. The script must be developed and recorded, which can be costly. But the cost to deliver each individual survey is relatively small.
What are the Weaknesses of the IVR Survey Method?
Lack of Anonymity. Like a telephone survey, no anonymity exists. We may feel confident that our responses will be treated confidentially, but heck, they have our phone number. (If you call them on a toll free number, they get your phone number, even if you block caller ID.) They know who we are. Thus, any survey research purpose that relies heavily on anonymity, such as an employee survey, is best done through some other method.
Keep it Very Simple. Again, like telephone surveys, the questionnaire or script must be kept simple. Complex survey questions that can need to be administered with the visual aid of a printed page or web page cannot use an IVR survey approach. For the most part, IVR survey designers are limited to interval-rating scale questions, that is, questions that ask for a response on a 1-to-5 scale or something similar. (The much touted Net Promoter Score® survey method cannot be done by IVR since it uses an 11-point, 0-to-10 scale.) Check list (or categorical) questions are difficult to administer since all the options must be read to the respondent.
Narrow Scope. When an IVR survey is conducted as a transactional survey at the end of an interaction with a contact center, e.g., a phone call, the only things that can really be evaluated in the survey is what happened during the immediate interaction. Yet, that interaction is likely one part of a broader transaction, which is composed of a series of interactions.
In the case of the Sears customer survey, they used the IVR survey not at the end of my call with their 800-4MyHome call center, rather they used it to measure my feelings about the technician’s visit to my home. This shows that the IVR survey method can be used to measure the broader transaction, although Sears through the stupidity of its functional silos, chose to measure just the transaction involving the technician.
Telephone Turn-off. Some people just dislike telephone interaction, including telephone surveys. We can thank telemarketers, especially those who use surveys as a pretense for a sales call. Some people also don’t like interacting with a machine. And those of us who have grown to hate IVR phone trees — and I count myself fully in that group — may also hate anything that involves an IVR.
In summary, IVR surveys can be a very useful tool to get very quick feedback for transactional surveys, and they are flexible enough to be used in other situations. However, care must be taken in the survey questionnaire design, which the Sears IVR survey definitely did not do.
When deciding on an online web survey tool what are the key features to consider and how should you go about the selection process? This article reviews important features of online survey tools and the decision process for selecting survey software, such as SurveyMonkey or QuestionPro.
Internet survey software tools make it easy for anyone to do a survey. SurveyMonkey, the best known of these web survey tools, states its goal is “to enable anyone to create professional online surveys quickly and easily.”
That’s true. Survey Monkey, Zoomerang ™, QuestionPro, and all the rest of the online survey tools DO make it simple to create a survey and launch it. They can be a boon to capturing customer feedback as well as from employees, and other stakeholder groups, quickly and simply.
However, using these survey software tools can be like driving without a rear view mirror and without a map — and perhaps with a distorted windshield. You may think you’ve reached your intended destination using the best route. But in fact, you may not have taken the best route. Worse, you may not even be at your intended destination, and you won’t know it. And another destination may have been a better one to meet your needs.
In this article, I’ll define survey software tools, and then discuss their benefits, the two types of web survey tools, their key features, and — to conclude the thought above — the dangers of “driving” with these online survey tools. (Hint: they can’t tell you how to create a good survey.)
What are Survey Software Tools?
Simply put, survey software tools provide the capability to create a survey where the painstaking survey administration tasks are automated. Typically, we think of these automation tools as online survey tools that generate an HTML form to post to the internet, allowing your group of interest to submit their responses via the web. That’s the focus of the remainder of this article. However, there are other survey automation tools, such as
Computer Aided Telephone Interviewing (CATI) software that manages surveys delivered by telephone interviews.
Interactive Voice Response (IVR) surveys, which are like telephone surveys, except that the scripts are recorded and the respondent enters answers using the keypad of their telephone. These are frequently used in call centers at the conclusion of customer service calls.
Optical Mark Recogniztion (OMR) or Optical Character Recognition (OCR) that allow you to scan a paper-based survey to capture the responses. (Think back to your high school SAT exams…)
Benefits of Web Survey Tools
The key benefits that can be derived from internet survey software tools are:
Simplicity. You can create a survey quite quickly — once you learn how to use the tool. Survey invitations are typically sent out electronically through an email. That email will contain a link to an HTML form that is the survey questionnaire. Responses are then captured electronically, either through incoming emails or by direct loading into a data base, depending on the type of survey tool used. This data collection function is by far the key benefit of online survey tools. If you’ve ever processed a hardcopy survey by postal mail and keyed in the data, you’ll see the benefit from survey automation tools very quickly!
Low Cost. The out-of-pocket cost of these tools is quite low. You can buy a PC-based tool for $500 or less or you can buy it as a service in a hosted application, typically paying a monthly or annual subscription fee. The fees can be as low as $15 to $20 per month for a basic tool. If you need more sophisticated features, the monthly cost could be much higher. As noted above, the major cost is not the acquisition cost of the survey capabilities, but rather, the time to learn how to use the survey software tool.
Scalability. Unlike telephone or postal mail surveys, you can survey 10,000 people as easily as 10 people. This scalability is part of what makes this a low cost method.
Quick Data Capture Turnaround. Since you may be soliciting feedback information about organizational performance, such as a customer satisfaction survey, how quickly you get the data back is important, especially if this is a transactional or event survey. Quick data capture allows addressing a customer issue promptly, which is a key element in successful service recovery practices.
With email most of us practice a “do or delete” approach. People will either take the survey or delete the invitation. So, you’ll find with a web form survey that most of your responses will come back in one to two days. Responses will quickly tail off to nothing. Only IVR surveys can be faster since they are administered at the close of the transaction.
Response Rate. When web-form surveys were a novelty, response rates were quite high, perhaps above 50%. Based upon my own survey project work and feedback from those in my Survey Design Workshop, response rates typically are in the 20% to 40% range. However, the relationship you have with your group of interest is the single most important factor in driving the response rate, not the administration method.
Questionnaire Complexity. Internet web form surveys are fairly flexible. (We can argue forever about which is more flexible: a printed page or a web form.) You can create a fairly complex survey questionnaire with these tools, using extensive branching and other available survey features. Do note that the survey tool you chose will set constraints on the questionnaire you design! It’s better to determine what functionality you need and then select the tool, rather than the other way around.
Dangers in Blind Reliance Using Online Survey Tools
Yes, Dangers. I opened this article with some seemingly twisted logic about survey software taking you blindly to a wrong destination without you realizing you were blinded. Let me explain…
Survey software tools make it easy to write a good survey. True. They also make it easy to write a lousy survey! And you’re probably better off doing no survey than doing a lousy survey. Why? Because a poorly designed survey can generate misleading data that provides you with delusions of knowledge. Researchers refer to this as “instrument validity“, which simply means that the survey instrument measures what you intend to measure. That may sound like a no-brainer. It’s not! Examples of invalid surveys abound, many done by professional market research organizations.
These online survey tools cannot tell you how to create a good survey. They also can’t tell you that you have a bad survey! Only knowledge of good survey design practices will help prevent you from designing a bad survey. (Obviously, this is why I recommend you buy my Customer Survey Guidebook or attend one of my Survey Design Workshops!)
The Internet web form survey approach exacerbates this blind spot. With a paper-based survey, respondents will give you feedback on the quality of your survey questionnaire. You’ll see big question marks and comments in the margin. (I’ve been known to do this on occasion…) With telephone surveys, you can listen to respondents formulate questions or talk with the interviewers to find out where the survey instrument has shortcomings.
How does the respondent provide you this feedback about an online, web survey? Perhaps you’ll see something in a comment field. But I doubt it. This is an incredible danger zone for this surveying method — and one that the survey software tool vendors never mention.
Technical knowledge of how to use the survey software too does not equate to knowledge of how to design a valid survey questionnaire!!
A twelve year old can push a car’s gas pedal and turn the steering wheel. Would you want them to drive a car? Yet, today many well-intentioned people are using SurveyMonkey and the like, creating survey questions that lack validity and then make business decisions on that bogus data.