Small-Sample Capability Estimation for Recruitment Speed: A DMAIC Case Study on an Indonesian Outsourcing Platform

Abstract

Workforce sourcing, the process by which outsourcing and gig-economy intermediaries recruit job applicants for client firms, lacks standardized capability metrics that are comparable to those used in manufacturing quality control. This technical case study reports the design, implementation, and critical evaluation of a Define-Measure-Analyze-Improve-Control (DMAIC) procedure used to establish provisional sourcing-speed benchmarks (pessimistic, average, optimistic) for an Indonesian human-resource outsourcing platform, stratified by province and job category. Because most province–job-type strata contained fewer than ten observations (range 1–19), a Chebyshev-inequality-based trimming rule was adopted instead of parametric three-sigma control limits, automated through Excel VBA for scalability. Results show that benchmark reliability is highly stratum-dependent: only the highest-volume stratum (n = 19) yielded values interpretable with reasonable confidence, while low-volume strata produced control bands too wide for operational use. This paper discusses why a fixed 2-standard-deviation trim is a weaker foundation than sample-size-adjusted alternatives recently proposed in the outlier-detection literature, and outlines the validation and data-aggregation work still required before such benchmarks can be deployed as genuine capability limits.