Blumen Research

Derived importance in the Importance Satisfaction Model

In the post covering the value of the importance satisfaction importance model I mentioned three areas that have to be taken into account before drawing results from the model. The notion that respondents will correctly identify importance levels for each product or service attribute will now be up for debate.

From the last post:

It is up for debate if customers can correctly identify importance levels for each attribute. When asked an importance level for several attributes what client would not want to indicate that each one is important. You have my business now focus on everything. While not all clients take this approach it is still a situation that occur. On the flip side of having clients indicate importance levels you can use various regression models to put together derived importance levels.

While it true that each respondent would want each and every issue taken care of in a timely and effective manor, they can also distinguish which ones are extremely important and which ones are important. The fewer questions located on the questionnaire the easier it will be for the respondents to identify important aspects. It would take a bit of testing to establish that number, but anywhere around 10 - 15 questions would be fare to nail importance. Over that number and you might be entering into the derived importance territory.

Derived importance

The first step is carefully designing the actual survey. Survey length even if you are going over the 15 questions should still be taken into account so the respondent does not opt out at a certain point or suffer from survey fatigue which will affect satisfaction and derived importance results.

While deciding on the number of questions to include you can conduct qualitative research with a subset of the total target to refine the list of attributes to be included in the survey. The shorter the survey, the more respondents you are bound to have which increases the likelihood of establishing statistically valid results for within segments, just in case it is still difficult to establish the attributes to importance based on the model.

Calculated derived importance is done by correlating the overall satisfaction rating with the satisfaction level for each attribute or by conducting a multiple linear regression and standardizing the coefficients.

Improvement Actions and the Importance Satisfaction Model

In order to raise the value of customer satisfaction where through a customer satisfaction index or in percentage you need to focus on the areas that requires improvement and maintains the areas of excellence. In an ideal situation with unlimited budgets each company would be able to focus entirely on every aspect of their business at each point in time to increase customer satisfaction. Since we never deal with ideal situations it would be worth while to tackle the headaches and maintain high satisfaction levels in areas that your client base is already ecstatic with. To correctly establish business operations to focus improvements you can utilize an importance satisfaction model.

Importance satisfaction models are divided up into four quadrants representing four areas of your business. On the x-axis is the importance level as established by the client base and on the y-axis is the level of satisfaction for each attribute. To divide up the model into the four quadrants, the customer satisfaction median is used as the central point dividing the y-axis while the median of the importance levels is used as the central point dividing the x-axis.

Four quadrants can be defined as follows:

Excellent

Attributes located in this area are ranked high in importance and high in satisfaction. It is the responsible of the company to maintain these levels of satisfaction to keep the client coming back and hopefully refer new a couple of new clients in.

Improvement

The work begins now to get this area improved. This quadrant contains all attributes with high importance but low satisfaction levels which in turn affects total customer satisfaction results and will impede a company’s growth.

Surplus

Clients are highly satisfied with these attributes, but don’t view them as being important. The performance of the company within this area will exceed the client’s expectation and should of course be maintained.

Fix, but don’t fret

These are attributes with low satisfaction and low importance. Companies don’t have to pay to much attention to these attributes as client’s don’t view them as being important, however client’s will also be pleasantly surprised by improvements in this area.
Our four quadrants make up the model and is a useful tool to guide management to make decisions on the business operations to focus effort and resources. During times where cost reductions are required management can turn to the surplus area.

As with most models there are considerations that have to be taken into account. I will outline them here and focus on each one in later posts.

1.  It is up for debate if customers can correctly identify importance levels for each attribute. When asked an importance level for several attributes what client would not want to indicate that each one is important. You have my business now focus on everything. While not all clients take this approach it is still a situation that occur. On the flip side of having clients indicate importance levels you can use various regression models to put together derived importance levels.

2.  Since the survey is conducted using ordinal scales (Very satisfied, Not satisfied etc.) it is also debatable whether you can represent the median for satisfaction or importance results as a definitive number.

An area that I will examine will be utilizing sliding scales within surveys, which may fix this issue.

3.  After the model is complete and the four quadrants are established will it then be wise to focus efforts on attributes with the To Be Improved section with higher importance levels or lower satisfaction levels?

Within the subsequent pots I will focus on these areas within the importance satisfaction model, but at the same time it is ready to be utilized by management.

Analyzing your competitor’s recommendation levels

Running customer satisfaction surveys on your current client base while not a completely straightforward task with respect to building out the survey and the analysis is at least straight forward with respect to building out the client database. But after completing your survey and establishing the relevant metrics leading to high recommendation levels and economic growth, every company should come back and consider the recommendation levels of their top competitors.

Within every industry there will be scores of competitors with ranging market share. Understanding the core issues leading to high recommendation levels and underlying reasons leading to lower customer satisfaction and recommendation levels of your top competitors will provide your company with the know how to bite into at least a small portion of their market share and possibly focus on a business operation within your own company that otherwise might have gone unnoticed.

Now the main question is how do we target our competitor’s customers to analyze their satisfaction and recommendation levels?

To gather scores for your top competitors there are essentially two methods.

The easiest approach is for businesses that compete for the consumer’s or client’s “share of wallet”, meaning that the buyer may be using several vendors to purchase their products or service or will purchase one category from you and another from a separate vendor.

In this case you can analyze your current share of wallet of your existing client base and within the same survey ask your clients’ which other vendors they use, using a radio button and how likely it is that they will recommend the vendor to a friend or business contact.

The second option which will apply to most companies is a supplement to your customer satisfaction analysis, a short top competitor analysis. For this case you will want to keep the survey short and sweet as the respondents currently have no vested interest in assisting your company. To gather the required data you need to survey a representative sample of the entire customer base, which you should already know since you target the same customer base. Also, make sure to have an accurate and representative sample based on geographic location.

While targeting the customer base for your market you should have a max of four questions on the questionnaire.

1. Using a radio button style question, ask the respondent who their top three vendors are from a list of about the top ten vendors in the industry.

2. In your opinion who is the top vendor in the market?

3. How likely is it that you would recommend [vendor] to a friend or business contact?

4. What was the reason for providing your recommendation score?

After compiling all your responses and ensuring results are statistically significant you will have a birds eye view of the competitive landscape and the recommendation levels for your top competitors as well as why the respondent provided the score.

Taking a look at the entire customer base is more difficult and time consuming than analyzing your own client customer satisfaction which is why the survey itself should be short enough to gain a quick view and a high response rate.

Wal-Mart thrives during the recession

Wal-Mart in its last quarterly call reported an increase in market share and posted unchanged profits compared to the same quarter in 2008. The key areas that companies large and small can take from Wal-Mart’s success are: overseas sales, prosperous customers, operational improvements, including reduced inventory and reducing the number of the brands it carries.

Relentless focus on reducing costs in house and focusing on target segments that generate the greatest amount of revenue has lead Wal-Mart to the profitable quarter in a time of recession.

Based on future projections and the need to remain a top the retail business Wal-Mart is investing in technological advancements, including human resource data systems, and a store remodeling plan that should lead to even further operating efficiencies.

If executives in every industry are not following the initiatives being undertaken at Wal-Mart they be wise to start.

Market research advice for CMOs

A recent article in BusinessWeek entitled “Using Market Research in a Recession” by Harvard's John Quelch is absolutely bashed by Tom Anderson. Anderson makes some bold statements including, “The article leads me to ask, what do professors know about the business of market research? Almost NONE of them have any real world experience. There’s only so much knowledge you can accumulate from explicit sources.”

My main concern with the article was the high level of the advice for CMOs where a proper analytical description alongside the theory should have been provided. Theory alongside practice would have been a compelling piece. However, such an article may be in the works in the Harvard Business Review and not BusinessWeek.

Qulech goes on to discuss uncertainty in the consumer market and how today’s CMOs can manage that uncertainty. According to Quelch big consumer marketers are looking to cut 10 to 20% off of research budgets and at the same time must secure highly quality data. He then goes on to list seven steps that CMOs can take in order to minimize the impact of the reduced spending.

From the seven, which you can take a look at within the article, what I view as added value, even at a higher end are the following four:

1. Stay focused: Conduct research on existing core customers and gain an understanding of loyal and profitable customers.

2. Enlist trusted partners: Explore synergies across partnerships where data and insights can be shared.

3. Seize opportunities overseas: Conduct market research in emerging economies.

Seizing opportunities overseas could possibly be the most important issue after staying focused, as it provides the company with a scope and direction for the future. Insights from overseas research enables corporations to find areas to focus R&D efforts, new product lines that can be developed for new markets and the core differences between the company’s current target markets and emerging markets.

Anderson states “As for business opportunities overseas, it depends on the product/market, but in any case, the US market is likely to recover before other markets globally.” I would have to say that he may be being a little to short sited with this response and the US recovering first should not be correlated to conducted research in emerging markets and possibly expanding a company’s reach.

4. Keep an eye on the new consumer: Not only should marketers understand future consumer consumption patterns, but also the underlying principles of why the consumer is buying and the consumer’s new research tools, such as Twitter.

All in all the article is nothing revolutionary for CMOs, but a starting point non the less to get the ball rolling. Now all they have to do is understand how to conduct the analysis on their core customers, emerging markets and the new consumer and they will be off to the races as the economy turns around.

Word of Mouth Model

David Armano contains a great visualization of the impact of word of mouth on his blog Logic + Emotion. On average promoters, actively satisfied customers, tell three people while detractors, actively dissatisfied customers tell ten. The problem is prevalent within marketing and market research has to capture the reason why customers become detractors and how to improve the situation while marketers need to reach out to the detractors through various touch points to decrease the damage and hopefully at least move them into the passive category.

Sampling Methods for B2B Surveys

While conducting a business to business customer satisfaction survey even though you will have access to the entire population or clients, it is important to use a proper sample of the client base. There are two types of sampling each with their own methods: probability and non probability sampling. For our purpose it is highly advised and encouraged to use probability sampling.

Probability or random sampling provides each client with the same chance of being included in the sample. Having one client selected does not impact the chance of a second client being included in the research.

Two statistical techniques for B2B random sampling:

Simple random sampling: Based on the total client base, each client has the same probability of being selected, and therefore no bias is present within the research. This method should be carried out if there are no segments or prevalent differences between clients.

Within Excel you can use the randomizing function (RAND ()) beside each client and then order it from greatest to least. Once ordered, simply take the first number of clients you want to send the questionnaire to and remember to base the total number of surveys sent on the expected response rate.

Stratified sampling: This method will reduce the sampling error as it is based on known subsets of the population such as region or revenue class. First you should establish the subsets or sub-populations (strata) to base the sampling on and then use the same random sampling technique for each subset to achieve a sufficient number that ensures the sampled subset is statistically representative of the total subset. After the surveys have been complete it is then vital to establish if the response is representative of the population and if not, weighting of the results should be applied.

Stratified sampling generally focuses on the most important subset, which will be different from project to project and reduces experimental bias.

Optimal Sample Size

Establishing the sample size and sampling method begins the market research process and can make or break the research. If the research sample does not accurately measure the actual population then the research does not represent the views of the target population. The sample size will indicate the level of accuracy for the research and the sampling method will ensure that the sample represents the actual population.

We will first take a look at the sample size and tackle the sampling issue in a separate post.

Sample Size: Population not known

“A sample is a subset of a population. Typically, the population is very large, making a census or a complete enumeration of all the values in the population impractical or impossible. The sample represents a subset of manageable size.” Wikipedia

To calculate the optimal sample size we need to first decide how accurate we want the results to be and how confident we want to be in the results derived from the research. A balance between the research budget and the level of accuracy required should be made to decide on the total sample size.

A well accepted confidence level is 95% and we should be prepared to accept an error of 3%. Without going into the actual formula (s = z/e)2 for the sample size calculation with those accuracy levels we would require a sample size of 1,067.

Sample Size: Population known (B2B Clients)

For B2B surveys where the size of the population is known the actual sample size can be slightly different. For the same case explained above, if the total number of clients was 1000 then the sample size would be 516. However, in most cases we will need to take demographics into account, such as region. At the highest level without drilling down to far into the client relationship management data, companies should establish the sample size based on the customer regions which will also be used during the sampling method. Samples from each region should be directly proportional to the actual population from each region.

The bottom line

Establish the accuracy required for the research (95%) and the acceptable error for the results, the confidence level (+/- 3). Based on those two factors and the actual population size the sample size can be derived.