Text analysis is the next great MR tool

I guess I shouldn't say "next" great tool in my headline. It's is already used globally in the marketing research industry. It is also used in many other areas as well. I recently talked with a representative from Eaagle. I was surprised how helpful this software is for the analysis of open-ended responses. It does much more than count repeated words. Eaagle's text mining software graphically provides what the marketing research analyst needs to find meaning from a large number of textual responses. You can find specific statements were groups of key words are found. It's like doing cross-tabs of descriptive words and then showing the results graphically. The program is much to sophisticated to describe in this short post. I know there are other software packages available, but I haven't explored all of the options yet. I'll eventually get around to it.

I can see a great application for text analysis of this type. I'm sure many of you have built your surveys by writing down the questions you want to ask first. Then, using your own logic and intuition writing categorical answers for those questions to use as multiple choice responses. Sometimes marketing researchers find far to many respondents answer, "other" when they complete a question. Even with open-end text explanation, it is upsetting when you missed an important closed-end response. Also, text analysis can help if you don't want to spend all the time and expense to manually code open-ended questions.

Why not create a survey that uses primarily open-ended questions first. After text analysis you can "close up" the responses. You will even find many more issues to explore. Use the results of the open-ended survey to create a second survey that is practically written by the respondent. In that way you find out what is important to the response base and not just thrown on the table by management or clients.

Delivering such surveys online is essential. Telephone surveys depend on interviewers to write what they think they hear. With online surveys you receive exactly what respondents type.

Since online surveys are often less expensive than other methods, you might even be able to conduct the two surveys for the price of one using a different method. Also, using the "two survey" method you will have both qualitative and quantitative data to use when trying to explore complex issues.

I'm considering this technique for a syndicated study in the Kansas market. My concept is not developed enough to discuss here. But when it is, I'll let you know.

SNS Site Sampling

Peanut Labs and LinkedIn Research
May 22, 2009

Innovations in Social Networking Sites (SNS) are leading to new marketing research opportunities. Peanut Labs is offering Sample 3.0. The system reach ages 13-64 including Gen-x, Gen-y and Boomers. Sample 3.0 uses social networks to reach 37 million members with a profiled audience size of 3.7 million. Within in one 60 day period they reached 853,000 research respondents. Their average response rate is 26%. Over, 81% of social networks are in Sample 3.0.

Surveys that occur within social communities are considered to be “fun” activities. Responses are gathered in hours not days or weeks. Peanut Labs incent respondents in real time with virtual currency “Peanuts” unique to each social network. You can now reach 90% of the entire Internet universe through social network environments. Facebook members use “Muny” to feed their pet or with other Facebook applications as survey incentives.

Peanut Lab’s Optimus proprietary technology uses digital fingerprints to make sure respondents are who they say they are and are not professional survey takers. They eliminate duplicates, speeding, spoofing, and survey bots. Algorithms are used to eliminate suspect survey respondents.

More than 30 of the Honomichl Top 50 research firms are using Peanut Labs to find hard to reach survey demographics. Revenues at Peanut Labs grew 300% in 2008.

It is often difficult to use a panel to reach qualified online survey participants in medium and small size markets. Is it possible that the products offered by Peanut Labs and LinkedIn Research will help this situation? Peanut Labs is focused on consumer sampling while LinkedIn Research targets B2B sampling.

LinkedIn is the world’s largest professional social network. It helps researchers conduct B2B research using 28 million professionals. Now we can reach large populations of quality B2B professionals. Besides selecting potential respondents by demographics, they can also be selected by keyword. The network is opt-in only. They cover about 150 industry segments. LinkedIn conducts member authenticity screening, survey programming and incentive delivery.

A study by Anderson Analytics (November, 2008) found that there are four types of LinkedIn users:

Savvy Networkers – likely to be early adopters and be tech savvy. They are more likely to use more SNS services. They have an average of 61 connections. They use LinkedIn for a wide variety of reasons other than job searching.

Senior Executives – Less tech savvy. They have the highest personal income of LinkedIn users. Use to explore corporate networks.

Late Adopters – Had invitations from friends and coworkers when they were not yet members. They are less technically savvy, connect to close friends and colleagues and have the fewest number of connections. (Average 23)

Exploring Options – They are working but maybe looking for other jobs. They are more likely to be using Careerbuilder.com. They are tech savvy.

I’m a “Savvy Networker.” That’s the first time I’ve been called “savvy.” Hopefully I’ll be given the opportunity to use one of these services to conduct an online survey. I’ve been concerned about using SNS sampling in the past. It seems logical to assume that people using social network sites like Facebook and Twitter might be a little more narcissistic than others. Even though you can match the demographics of a population using one of these methods, I wonder if unknown psychographic variables might bias results. Peanut Labs uses a combination of so many different social networks that this bias might be minimal. When you think about it, basic marketing research techniques don’t change that much, but where we find respondents will always depend upon how potential respondents communicate with each other.

Research from Facebook?

The Buzz is a blog that shares and discusses the best in alternative marketing campaigns like: viral, blog, influential, worm, buzz marketing, and branded content.

A recent article entitled, FACEBOOK’S INSTANT MARKET RESEARCH TOOL they stated that a Facebook team presented a new tool at the World Economic Forum held in Davos, Switzerland that would allow Facebook to poll users and get instant feedback. The system would allow companies to target people based on their profiles. The question is, will Facebook users stand for it. Much may depend on potential compensation for respondents. Facebook could be the next super panel. I also think a big PR campaign prior to launch of such a campaign would be appropriate. Otherwise Facebook users could boycott. Who knows?

Anywhere on the web where big numbers of users gather seems to be a target of marketing researachers. User groups of product users are often a great way to recruit research participants for both online surveys and focus groups. But make sure that group members have an intense interest in the product and that you ask for permission from the group moderator. It is best that the moderater distributes URLs for surveys and URLs for focus group screeners.

Can you believe it..1 billion Internet users

There was an article in the Wichita Business Journal on Friday that stated that in December comScore reported that the total global Internet audience surpassed 1 billion users. China has the largest percentage of Internet users at 18%, followed by the United States at 16.2%. Japan, Germany and the United Kingdom follow the US. As you might expect, Goggle was the most popular Internet "property" in December followed by Microsoft sites and Yahoo sites.

It seems most of my friends and family are turning to Facebook, and they are not alone. According to comScore, Facebook has grown by 127% this year. wow!

The impact of the Internet on communication just keeps accelerating. Online marketing research is likely to also expand as the number of Internet users continues to grow.

Many companies cut their advertising budgets during recessions. But often they use the break to redesign their marketing strategies to be ready when the business environment improves. It is also the perfect time to pay attention to product development. Online marketing research techniques can be efficiently used during the product development process with great success.

Good luck in 2009.

You can't have positive results without keeping a positive attitude. (I just made that up)

Those Boring Online Surveys

When you look at typical online surveys from the respondents’ point of view, I have to admit that most online surveys are not very interesting. It is especially the case if the respondent is not a member of the intended audience and has no interest in the subject matter.

In the November 2009 publication of Quirk’s, Deborah Sleep from Engage Research, discussed some research they conducted about respondent behavior when taking online surveys. They looked at how survey length and question type impact dropout rate and many other issues. Solutions for problems with respondent boredom were discussed in the second part of the article in the December issue. They suggest making online surveys more interactive, like the rest of the internet. Basically, we should look at how our surveys communicate with the respondent. The study showed how using more graphic content makes surveys more interesting. It also suggests questions using more Flash elements. Like dragging photos to positions on scales rather than using check boxes. It seems that to keep people interested more “movement” in online surveys is suggested. The article supported my view that sometimes we use questions that are too “formal” and much more “wordy” than they need to be.

I agree that marketing researchers need to make surveys more interesting to respondents. If people continue to experience boring surveys the industry runs the risk of lower and lower response rates, even when incentives are used.

However, how far should we go? If graphic elements are not applied consistently wouldn’t the validity of responses be in question? How about the cost/benefit of creating more interactive online surveys? Most survey packages used in the marketing research world are not very friendly to flash elements. If they are, employing a programmer to create the flash programming can be expensive. One of the major benefits of the online survey methodology its economical advantage. Then we add a custom built panel to insure data quality and now we add interactivity. Whew, maybe we should go back to mail surveys even with high postal, printing and data input costs.

I suggest that if you go the interactive direction that you have question styles programmed that you can reuse for many clients. I have one survey package that is much more like a layout program than most survey packages so interactive elements can be added. I’m sure the software originator could come up with some templates for more interactive questions. Also there is a firm that provides a system only to marketing research firms that includes a number of interactive question styles.

There are more limitations. What if you’re conducting discrete choice conjoint or max diff? I suppose you could add some simple illustrations. They would have to be similar to each other to avoid having an impact on individual responses. The survey software you use may limit how much cool stuff you can include.

What does this all come down to? Maybe we should think about making our online surveys more attractive, engaging and interesting. Just writing a good questionnaire isn’t enough. We should put more effort into thinking about how surveys actually communicate with respondents. But we shouldn’t forget to think about cost effectiveness. I think if we do a good job sitting in the respondents chair, we can keep their interest until they hit the submit button.

Fast and furious leads to lower quality data

Today’s nervous companies want information to be available in real-time. If it has to be generated using a primary data gathering technique, speed is an important consideration when choosing marketing research vendor. When we are pressured to provide information under-the-gun, researchers often skip some steps, or rush through them. Now that so much research is done on the internet, it becomes convenient to test research instruments insufficiently.

In the past, when mail surveys were the only game in town, the testing phase was taken very seriously. A substantial number of surveys were sent to members of the sample and responses were studied very carefully. Then research instruments were edited to make sure questions were understood, skip patterns functioned, and information was provided in the correct format. Often surveys that needed a lot of work were retested until perfected. Testing surveys took a lot of time. The objective was to obtain the highest quality of raw information from the respondent as possible to minimize editing and database cleaning prior to tabulation.

When using telephone surveys, vendor supervisors listen to their interviewers during survey testing to discover problems with the survey. Supervisors also ask phone interviewers directly about stumbling blocks and questions that are being misunderstood by respondents.

My suggestion is to add two or three days to the research process when conducting online research. Since response is usually so fast, it doesn’t take that long to adequately test an online survey instrument. I have often suggested that clients have co-workers look over and take online surveys prior to the survey launch. I also personally complete the survey many times to look at survey in several browsers and make sure the output is in a format I can use for analysis. However, I’m now suggesting that online surveys be tested by a small portion of the survey sample, or a group that is very similar to the sample to be used for the study. Look for questions that are not being answered and responses in open ended questions that don’t make sense. I have written a checklist of over 100 items to be considered prior to launching a survey into the field. Items are added to that list as I discover additional problems that can occur. A survey that is well designed and tested will provide a higher quality of data. Higher quality data leads to better decision making. As researchers, we desire users of our research to have confidence that our research is dependable. Adequate survey testing is still an important part of the research process. Just calm down and take things a little bit slower and you’ll be happy with the results.

surveydepartment.com hosts first podcast

Hi everyone, I posted an episode to my podcast, Online Marketing Research Podcast.

Click this link to check it out:
Podcast #1: Start out with observation.

- Fritz

Physical Observation: Describe This Family

While looking through Zikmunc and Babins textbook "Essentials of Marketing Research," I found an interesting section about observation. It seems people have actually searched trash to conduct marketing research. Apparently Charles Coolidge Parlin counted garbage cans at the turn of the twentieth century. Parlin wanted to demonstrate that the target for Campbell's soup wasn't just for affluent people. Campbell's didn't want to advertise in the Saturday Evening Post because they thought that the Post was read by people who wanted to make soup from scratch. So Parlin designed a sample of garbage routes in Philadelphia. They dumped the garbage on a floor of a National Guard Armory and counted the number of Campbell's soup cans. There were a large number of cans that came from blue-collar areas. So Campbell's advertised in the Saturday Evening Post. What a stinky project? It would have been a lot easier to send them email purchasing journal surveys.

Can you describe the family's that generated the trash shown in the picture above? I bet you can come pretty close if you try? Use the comment section to profile the mystery family. Have fun!

Panels Are Not Easy To Build

Over the past two weeks I tried to see what would happen if I built my own panel to use for online marketing research. Usually, I purchase services from panel companies whey I'm doing work for clients that require target lists. After figuring out how to create a double-opt in system which would stand up, I placed advertisements using Google. Over two weeks I spent about $350 in click-through charges and generated about 8 panel members. That makes me think that panel companies must spend an enormous amount to solicit panel members when most advertise that they have 100's of thousands of members. I think I'll stick with purchasing opt-in panel members from reputable panel companies.

Respondents Police Online Survey Panels

Over the past few years so many online survey panels have sprung up that it's impossible to know which to use. Some claim to have 100's of thousands of participants, even millions. It is often the only way to obtain a focused list of survey participants. Many of the large online survey companies have their own panels and spend a great deal of money and care marinating them. It's also interesting that their is a website that panel members use to rate panels. This creates a good list of panels as well. You can look for yourself at www.surveypolice.com.