LIST OF ACRONYMS
7 CHAPTER : RESEARCH DESIGN AND METHODOLOGY
7.11 PART 1 - CUSTOMER EXPERIENCE SURVEY (QUANTITATIVE)
7.11.3 Sampling selection and size – customer
The aim of a survey is to collect information about a population (Fielding, Lee &
Blank, 2017:277). However, collecting information from the entire population is not always practical and possible, as per Table 7.2 above. Alternatively, the researcher may select a sample from the population. A sample is described as the selection of a smaller group from a larger group of the population with the aim to generalise information about the total population (Saunders et al., 2012:210; Neumann, 2014:246).
There are two categories of sampling techniques available to be used, namely probability and non-probability sampling. For the purpose of this phase of the dissertation a probability sampling technique was used. Probability sampling refers to each case having equal chance to be selected (Fricker, 2017:274). More specifically, stratified random sampling was found to be most suitable for use within the research of this dissertation. Saunders et al. (2012:228) assert that it entails dividing a
population, and grouping it according to common and important characteristics or attributes.
The motivation for selecting this technique can be found in the definition of total customer experience. Berry et al. (2002:2) argues that total customer experience include the product, service functionality and emotional experience. In order to truly understand the level of customer experience from customers in this sample, it is important to consider some product-specific attributes and their influence on total customer experience. Customers were segmented, based on their product status, healthy heart score and number of active days. The combination of a customer’s product status (bronze, silver, gold, platinum and private club), together with their healthy heart score and the number of active days in a month, determine the size of a customer’s rewards or discounts. For this reason, customers were grouped according to the attributes described above, after which a unique number for each was calculated in Microsoft Excel and allocated to each record from which the sample frame was determined.
The sample for the customer experience survey consisted of 23 685 identified and contactable participants as per Table 7.3 below. A number of clients, however, were removed from the initial population list due to reasons listed in Table 7.3 below.
Table 7.3: Customer experience sample
Total number of customers 26 450
Less: WRP clients less 1 month 1 758 Less: Clients with no e-mail addresses 667 Less: Clients with Company X e-mails 233 Less: Client & broker e-mails same 107
Total sample 23 685
The survey was therefore biased towards former WRP members who had a valid e-mail address. Given the population of 23 685 former customers, the calculated sample size, using an alpha level of 0.05 and sampling error of 5 percent and applying these values to Cochran’s formula, was calculated as 379. However, Saunders, et al. (2012:219) stress the importance of a high response rate; with Fielding, Lee and Blank (2017:277) suggesting that a higher response infers a lesser chance of non-response bias.
7.11.3.1 Response rate – customer
Regardless of the method of sampling, the challenge for any research to overcome is low response rates (Mellahi & Harris, 2016:426). The response rate according to Fielding, Lee and Blank (2017:277) refer to the number of returned questionnaires divided by the total number sampled. This ratio, according to Whelan (2015:1) has been used as a standard, albeit an imperfect one. However, others argue undeliverable questionnaires should be excluded and subtracted (Fincham, 2008:2).
This view is supported by Saunders et al. (2012:220), who argue a perfect response rate is highly improbable and hence always results in some form of non-response.
Non-responses are defined as not being able to obtain information from a person for whatever reason i.e. death, unavailability or blatant refusal to reply (OECD, 2008).
The authors argue it is more appropriate to differentiate between non-achievement, non-availability and non-response. Non-achievement or non-availability is described as the failure to make contact with a potential participant, for whatever reason, referring to non-response as instances where participants were successfully contacted but refused to participate.
Mellahi and Harris (2016:426) examined 1 093 research papers between 2009 and 2013, employing survey methods to determine an acceptable response rate. The typical response rate for International Business was determined to be approximately 35 percent; with Human Resources and General Management typically achieving 50 percent (Mellahi & Harris, 2016:435).
However, Neuman (2014:342) notes that, due to survey popularity, response rates have declined over the years; with (Monroe & Adams, 2012:2) reporting online survey response rates dropping as low as 2 percent. In practice, online platforms such as Fluid Surveys (2014) and SurveyGizmo (2015) suggest an average response rate ranging between about 10-24 percent, based on their platform use.
Everlytic, a South African marketing and communication organisation analysed over 2 Billion e-mails (Everlytic, 2017). The data was compiled from emails sent on behalf of Everlytic’s, South African clients during the period January 2016 to December 2016. The median e-mail campaign size consisted of 2 553 recipients whereas the mean consisted of 13 436 recipients. In addition to measuring sent emails, Everlytics (2017) also tracked and analysed the open rates, click-through and click-to-open rates. The open rate is described as the percentage emails opened by customers compared to emails send. However, only focussing on the open doesn’t tell the whole
story. Everlytics (2017) suggest, blending open rates with other metrics in order to provide a more holistic picture of the distribution success. The click-through rate is described as the percentage of recipients that clicked on a link inside the email send, counting only unique clicks per contact, providing better insights into content and call to action effectiveness and whether recipients find the content valuable. The click-to-open rate refers to the ratio of unique clicks compared to the total number of unique emails opened. Everlytics (2017) argue the click-to-open rate indicates the level of engagement with the mail content, and does not only look at the number of open e-mails, but also the amount of engagement. In the Benchmark 2017 results, (Everlytics, 2017), the open rates revealed a mean of 27.60 percent and a median of 27.79 percent for customers of financial institutions surveyed. The click-through rates revealed a mean of just 3.33 percent and a median of only 1.35 percent. Click-to-open rates revealed a mean of a better 14.32 percent and a median of 7.20 percent.
This view is supported by Saunders et al. (2012:220) who state that a perfect response rate is highly improbable and as such suggests using a larger sample sizes to ensure an adequate number of responses.
In an effort, to bolster the expected response rate, face-to-face and telephone interviews were considered. However, although both have many advantages to reduce non-response bias, one major disadvantage is the influence of interviewer bias (Szolnoki & Hoffman, 2013:58). In addition, in an attempt to further improve the expected response rate, the researcher contemplated offering incentives in the form of a couple of sports and fitness watches. Contrasting views exist whether enticing participants with incentives will actually lead to higher response rates (Mellahi
&Harris, 2016:428). The authors observed, in some cases scholars warned incentives had the opposite affect than was intended and other suggesting it intrinsically alters motivation and reason for participating in the first place. In other cases, they observed, higher response rates without offering incentives. This together with the limited timeframe to collect the data informed the survey design to use online surveys as the only primary data collection method.
For the purposes of this dissertation a conservative response rate of approximately 20 per cent was assumed. This was, considering the participants sampled, were former and not current customers with very little motivation or incentive to take part in the survey (Monroe & Adams, 2012:2). The re-calculated actual sample size was
required responses. The results of which will be discussed in more detail in (par. 8.6, p. 193).