With the increasing complexity of customer demand, quality improvement faces significant challenges, as personalized and customized production patterns are frequently adopted and large-scale factors are involved in complex production. Aiming at the challenges posed by large-scale factors and small sample sizes in complex and customized production, this paper proposes a bootstrap-based factor screening method, which serves as the first stage of quality improvement. Firstly, a one-order polynomial model is adopted to fit the input-output relationship without assuming a specific distribution for the random term, accommodating the situation of small sample size. Secondly, the sequential bifurcation (SB) procedure is modified according to the distribution-free response model. Thirdly, three significance testing methods for factor effects are proposed, based on Students’ t test and enhanced through bootstrapping and bias correction procedures. Finally, Monte Carlo simulations are employed to compare the proposed significance testing methods with the classic Student’s t test under small sample sizes, and to verify the effectiveness and robustness of the proposed screening methods in both small-scale and large-scale factor scenarios.