What is Rim Weighting?

Of all the techniques used in quantitative market research, weighting can be one of the most useful…and one the trickiest to apply properly. Among the various weighting approaches available, rim weighting is an especially valuable addition to the researcher’s toolbox.

The “rim” in rim weighting comes from the acronym for Random Iterative Method. The name may sound complex, but like any kind of weighting, it’s a solution to a fairly straightforward problem – the need to adjust a sample so that it is representative of the target population. This need arises frequently in market research cases where low response to a survey among certain segments leads to a dataset that is not representative of known population characteristics. For example, if a researcher knows that her target population is split evenly among gender lines, yet 65% of the survey responses are from women, she may need to use weighting during the analysis to allow for the skewed response pattern.

Of course, only having to worry about the proportionality of just one characteristic is easy. Unfortunately, we often have to ensure that our data matches the population in a variety of ways – not just gender, but also age, income and any number of other traits.  That’s where rim weighting comes in. The technique allows the analyst to adjust multiple characteristics in a dataset all at the same time in a way that it ultimately keeps the different characteristics proportionate as a whole.

To use a very simple example, let’s say that we know that a target population of college students is divided as follows in terms of gender and age distribution:

Male	40%
Female	60%

 

18-24	70%
25-34	25%
35 or older	5%

Now, let’s say that we conduct a student survey, and the demographics of our respondents look like this:

Male	30%
Female	70%

18-24	65%
25-34	20%
35 or older	15%

When we analyze the results, we know that we want to use weighting based on the known distributions of the student population. Rim weighting allows us to weight both characteristics at the same time. It does this by using an algorithm that distorts each variable as little as possible. The ultimate result is a weighted data that closely matches the target population across all the pre-defined dimensions.

Rim weighting is useful when you know some characteristics of your target population, but you aren’t sure about the relationship between them. In the case of the example above, we knew that 40% of our population were males and 5% were 35 or older, but we didn’t know what percentage of the population were males who were also 35 or older.  By making adjustments to multiple characteristics at the same time, rim weighting infers that information for us.

That last point is important, because it shows one of the limitations of rim weighting. If there is a strong relationship between two characteristics, for example household size and marital status, rim weighting based on those characteristics would probably produce an inaccurate result.  Rim weighting is also most effective when the actual values don’t differ a great deal from the target values. (That is true of weighting in general, as we have written about here.)  In the example of our student survey, if the respondent pool was 5% male and 95% female, rim weighting would not produce as accurate of an analysis.

Keeping those limitations in mind, rim weighting is an extremely valuable analysis tool to market researchers in the right situation.

Research & Marketing Strategies (RMS) is a market research firm located in Syracuse, NY. If you are interested in learning how we can help you turn data into actionable insights, contact our Business Development Director Sandy Baker at SandyB@RMSresults.com or by calling 1-866-567-5422.

Read Full Post »

Is Big Data Overwhelming You? Use Small Data

First the big hype was all about “Big Data” but now there is some buzz generating for “Small Data”.  The concept of big data has caused strain at many companies as they realize their databases lack both the size and sophistication necessary to analyze large amounts of data and produce actionable next steps.  Ultimately, utilizing big data is not feasible for many companies, and likely will never be.  Companies need to work with what is available and focus on deriving insights from the data it has on hand.

As a market research and marketing consulting company, we’ve been familiar with the concept of small data for a while. However, we simply refer to it as just plain old “data”.  We’re in an age where it is becoming easier than ever for small to medium size businesses to collect data on their customers and the usage of products of services.

Here are 4 basic types of small data to review that does not require a sophisticated database, tracking tool, or staff hours:

• Loyalty Data – Collecting loyalty data is all about uniquely identifying your customers or clients (through loyalty cards, customer IDs, usernames, etc.). Once customers are identified, you can collect longitudinal data recording purchasing habits, as well as comparing purchasing data to demographic data to create niche buyer segments.

•  Satisfaction Data – Many companies collect satisfaction data, but ultimately don’t do anything to aggregate and analyze it. This is a great place to start for understanding customer satisfaction. This can be collected through a simple customer comment box, survey posted on your website, or through a Quick Pulse telephone survey.

• Lead Data – Tracking the source of leads can be hugely beneficial for a company. Asking all new customers, “How did you hear about us?”, and analyzing those results can be quite eye-opening, yet many businesses fail to collect this information. Tracking the source of new customers will help with optimizing your marketing budget by letting you know what is working, and what isn’t.

• Product and Service Usage Data – Tracking which products and services are being purchased and how frequently they are being purchased can provide lots of insights. While this ultimately might lead to more questions (like why a product is or isn’t being utilized?), it is certainly a step in the right direction to understanding what your customers want from your company.

While every company is at a different stage with data management, companies need to become experts on garnering actionable insights from the data it has and look for ways to acquire data it needs. Technology is making it easier than ever for companies to get started with data management, whether it be big data, small data, or any data. CRM systems (Customer Relation Management) and POS (Point-of-Sale) systems are two great examples of systems that allow you to track, report, and analyze customer and transaction data.  Additionally, businesses that have a web presence have a variety of different website and social analytic tools at its disposal. 

Utilizing and making sense of your company’s available data can provide the insights necessary to successfully market, operate, and grow a business.  Are you a company looking to analyze data, or trying to bring of all your data and findings together in a productive manner?  Feel free to call our Business Development Director, Sandy Baker, at 1-866-567-5422 or email her at SandyB@RMSresults.com.

Read Full Post »

6 Survey Coding Tips | Survey Research in Central NY

Including open-end responses in a survey is always a great way to get information; types of information that weren’t touched upon in the script or answers that wouldn’t have previously been listed for a question.  It can be a great source for qualitative data within a quantitative survey.  Survey coding is the process of taking the open-end responses and categorizing them into groups.  Once coded, they can be analyzed in the same way multiple response questions can be.  The survey coding process can be very tedious in order to ensure reliable results.  After reading through 500 comments of your dissatisfied customers, you may feel like you need counseling, but the results will be invaluable to your company.  So ask your market research vendor to send you all of the open-end responses in an excel file at the conclusion of the fieldwork.

The survey coding process for open-end questions must be undertaken carefully, as responses can be open to judgment and interpretation of the individual.  The results can vary from person to person depending on what code you use for the open-ended comment.  The process can be very subjective, even though one may think market research analysis is meant to be objective.

Here are a few tips for survey coding:

1. Read through all open-ended responses ahead of time.  This helps the analyst get a feel for the themes that are recurring in the data set.  It will also help the analyst understand how the population is responding to the specific survey question.  It can be surprising how many people answer an open-ended question in a similar fashion to one another without any guidance. 
2. Start by creating a lot of categories before narrowing the field.  Once all of the categories are laid out and initially coded, begin to further combine the data to limit the analysis to about 8-12 codes.
3. Make sure everyone’s comment counts.  A reporting standard for RMS is to create 8-12 all-encompassing survey coding categories, even if there are a few outliers that are lumped into the category of “Other,” they must be mentioned and/or footnoted.
4. Create accurate and unambiguous codes, which cover the responses they apply to.  If someone were to look at the code title after reviewing a question, it should be clear what type of comments fell under it.  It may help to create a pop-out box in the report with further detail on the code with some explanation. For example, stay away from creating a category like “Service Related” – does that mean Customer Service? Billing Service? Service in relation to usage?
5. Feel free to use more than one code.  Many times respondents offer multiple comments on a question spanning from topic A to topic Z.  If a response was limited to only one code, the respondent’s other areas of concern would be understated.
6. Consider coding % of respondents rather than % of responses.  This will give an equal value to all responses.  This prevents those who express concern in multiple areas from overpowering those who had a single code response.

The responses to open-ended questions are the most raw and unaffected parts of survey analysis.  The questions are completely unaided, and respondents can say or write anything that comes to mind.  They are not limited to the selecting choices or guided in their response.  With open-ended questions you get a true sense for how the respondents feel.  At the same time, the survey coding can be compared and analyzed for more quantitative/ statistical analysis.  The process of reviewing all open-ends and survey coding has become an important standard in reports delivered by Research & Marketing Strategies (RMS) and the results are always very rewarding for our clients.

Read Full Post »