QNT 375 Week 2 Research Design and Data Collection Example

Records, Surveys and a Natural Experiment: Designing the Data Collection for a Car Wash Retention Study

[Student Name]

University of Phoenix

QNT/375: Business Data Analytics

Week 2 Assignment

[Instructor Name]

[Date]

Summit Shine Car Wash, its locations, members and figures are composites written for a model paper.

Week 1 framed a problem for Summit Shine Car Wash, a composite chain of 22 express washes in the Denver area: identify which members are likely to cancel and why, so the company can spend a $600,000 retention budget well. It posed four research questions about cancellation patterns, wash frequency, wait times and competitor proximity, the effect of the price increase and the competitor's openings and the ability to predict cancellations. This paper designs the data collection.

Research Design

Cooper and Schindler (2014) classify business research designs by purpose, from exploratory to descriptive to causal, and by method of data collection, from monitoring existing records to communicating with respondents. Summit Shine's study needs more than one design. RQ1 and RQ2 call for descriptive and correlational analysis of existing records. RQ3 asks a causal question, whether the price increase and the competitor's openings raised cancellations, which records alone cannot settle but which a natural comparison can address. RQ4 calls for a predictive model built on historical records. A member survey will add reasons that records cannot reveal.

Secondary Data

The company already holds rich data:

Membership system: 61,000 current and about 19,000 members who canceled in the past 12 months, with plan, price, start date, cancellation date, home location and, for about a third of cancellations, a reason chosen from a list.

Wash logs: each wash recorded by license plate reader, with member, location, date and time.

Queue cameras: estimated wait times by location and hour.

Weather: daily snowfall and precipitation from public records.

Competitor openings: dates and addresses of the competitor's five new sites.

Secondary data are inexpensive and complete for the whole membership, but they were collected for operations, not research, so definitions must be checked. For example, the cancellation reason list was changed in the spring, and wait times are estimates, as Week 3 will examine.

Variables

Outcome: canceled within the study period, yes or no; and months of membership before cancellation.

Predictors: plan type and price; tenure; average monthly washes in the prior three months and change from the previous three; home location; average weekend wait at the home location; distance from home location to nearest competitor site; month and snowfall; whether the price increase applied.

A Natural Comparison

The competitor opened sites near seven Summit Shine locations, at different times, and none near the other fifteen. Comparing changes in cancellation rates before and after each opening at nearby locations with changes at distant locations over the same months helps separate the competitor's effect from seasonal and chain-wide trends. A similar comparison can be made for the price increase, which applied to members on older plans at renewal.

The competitor did not open everywhere at once, and that unevenness is what lets the data speak.

The Survey

Records show what members did, not why. A short survey will ask recently canceled members their main reasons and current members their satisfaction and intentions. Population: members active or canceled in the past 12 months. Sample: a stratified random sample of 1,200 canceled and 1,200 current members, drawn proportionally from each location. Method: email with a mobile-friendly link, a reminder after four days and a $5 wash credit for completion. Dillman et al. (2014) recommend multiple contacts, short questionnaires, clear wording and small incentives to raise response rates, all of which the plan uses. Expected response: about 20 percent of canceled members and 30 percent of current members, roughly 600 completed surveys.

Groves (2006) reviewed dozens of surveys and found that a low response rate, by itself, did not reliably signal a biased result, but that bias arises when the reasons for not responding are related to the variables being studied. Canceled members who left angry may be less likely to respond; to check, the analysis will compare respondents' recorded characteristics, such as tenure and plan, with those of nonrespondents.

Sample Questions

"What was the main reason you canceled your membership?" with options including price, wait times, a closer car wash, washing less often, moving and other.

"How often did you wait more than 10 minutes in the month before you canceled?" never, once or twice, three or more times.

"How likely are you to rejoin in the next six months?" on a five-point scale.

How Many Responses Are Enough

With about 600 completed surveys, the study can estimate the share of canceled members giving a particular reason with a margin of error near five points, at the usual 95 percent confidence level, enough to rank the main reasons. Comparing reasons between locations near and far from the competitor will be less precise, since each group will have about 300 respondents. Larger samples would cost more in incentives and time; the design accepts this limit because the main evidence comes from records covering all 80,000 members.

Timeline and Cost

Data extraction from the membership, wash log and camera systems will take two weeks with the IT manager's help. The survey will run for ten days. Incentive costs are about $3,000 in wash credits, which cost the company little since members redeem them at existing locations. Analyst time is the main cost, about 120 hours across the project.

Pilot Testing the Survey

Before sending the survey to 2,400 members, the team will test it with 20 staff members who are also members and 10 recently canceled members reached by phone. The test will check whether questions are understood as intended, whether answer options cover the common reasons and how long the survey takes. A survey that takes more than five minutes will be shortened.

Ethics and Privacy

Combining license plate logs with membership records involves personal data. Analysts will work with records stripped of names and plate numbers, linked by an internal identifier. Survey responses will be confidential and reported only in aggregate. The privacy notice in the membership agreement covers operational analysis; the survey will explain its purpose and that participation is voluntary.

Conclusion

Summit Shine's retention study will draw mainly on records it already has, strengthened by a natural comparison across locations and a short member survey. Defining variables, sampling carefully, planning for nonresponse and protecting privacy set up the data quality review in Week 3.

References

Cooper, D. R., & Schindler, P. S. (2014). Business research methods (12th ed.). McGraw-Hill Education.

Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, phone, mail, and mixed-mode surveys: The tailored design method (4th ed.). Wiley.

Groves, R. M. (2006). Nonresponse rates and nonresponse bias in household surveys. Public Opinion Quarterly, 70(5), 646-675. https://doi.org/10.1093/poq/nfl033