Service Capacity Optimization Data
Description
İşte 3000 karakter sınırına uygun, özlü bir İngilizce açıklama (~2,900 karakter): This dataset supports the manuscript "From Cost Assumptions to Customer Thresholds: A Satisfaction-Based Queueing Framework for Service Capacity Optimization." It enables empirical testing of a framework that replaces the unmeasurable "waiting cost" parameter of classical M/M/s queueing models with empirically grounded customer satisfaction thresholds. Data collection. Observations and a concurrent customer survey were conducted at a campus restaurant during peak lunch hours (12:00–13:00). For each of 173 respondents, trained observers recorded the exact clock times of order placement and meal delivery, yielding an objective, individually measured actual waiting time. At the same time, each respondent completed a survey measuring perceived waiting time, satisfaction with waiting, overall service satisfaction (four items adapted from Saad Andaleeb & Conway, 2006), visit frequency, and demographics on 7-point Likert scales. This concurrent design — objective and perceptual measures captured from the same individual at the same encounter — is uncommon in the queueing literature and rules out recall bias as an explanation for perceptual effects. Hypotheses tested. (1) Waiting time satisfaction mediates the link between waiting time and overall satisfaction; (2) perceived waiting time exerts a stronger effect on satisfaction than actual waiting time, even under precise objective measurement; (3) capacity expansion reduces waiting time predictably under M/M/s; (4) empirical satisfaction thresholds can replace assumed cost parameters in capacity optimization. Key findings. Mean actual waiting time is 20.10 minutes. Actual and perceived waiting time correlate at r = .58. Both relate negatively to satisfaction, but perceived (r = –.49) more strongly than actual (r = –.36). Mediation analysis (PROCESS Model 4) shows waiting time satisfaction fully mediates both pathways; the indirect effect of perceived waiting time (–.295) is nearly five times that of actual waiting time (–.061). Variables. anketno (ID); ordertime/servicetime (observed timestamps); duration/durmin (actual waiting time in seconds/minutes); perception_duration (1=very short, 7=very long); time_satisfactiongeneral (1–7); satisfaction1–4 and time_satisfaction1–4 (1–7 item scales); frequency, gender (1=woman, 2=man), age, educationlevel, profession; filter_$ (long/short-wait subgroup flag). Interpretation and reuse. durmin is system time, not pure queue time. The data support M/M/s, M/G/s (Whitt 1993), and batch-arrival re-analyses; arrivals are overdispersed (IoD = 1.993) and service times sub-exponential (CV = 0.382). The dataset is suitable for replicating the perceived-vs-actual effect, deriving alternative satisfaction thresholds, testing moderators (gender, frequency), and cross-context meta-analysis when combined with other service-setting data.