Regression analysis for evaluating employee productivity in IT project management

Employee performance assessment is a key element of effective IT project management, especially in the context of the transition to a remote or hybrid work model, which has become dominant after the global changes caused by the COVID-19 pandemic, digital transformation and social shifts. In such conditions, traditional approaches to measuring performance, based on direct observation or formal reports, lose objectivity and do not take into account the real context of task performance. Most modern activity control tools are focused mainly on monitoring, rather than on a deep analytical analysis of the factors that shape employee performance. This article proposes the use of regression analysis methods as a formalized statistical tool for quantitative performance assessment. The regression model allows you to simultaneously take into account several influential factors – the number of tasks completed, the accuracy of estimates, the quality of work (the number of defects) and participation in team processes – and determine their contribution to the overall level of performance. The mathematical justification of linear regression in the context of project management is presented, formulas for calculating coefficients by the least squares method and indicators for assessing the quality of the model (R2, p-value, standard error) are given. The construction and interpretation of the model are demonstrated using the example of a realistic data set of ten developers. The results obtained showed a high level of explanatory power (R2 = 0.97), which confirms the close relationship between the independent variables and the productivity indicator. The greatest impact on efficiency was found in the work quality factor (β = -2.68), while participation in teamwork and the number of tasks have a moderate positive effect. The regression model can be used both for retrospective analysis (identifying patterns and deviations in productivity dynamics) and for predicting future results or evaluating new employees by their metrics. It provides transparency, reproducibility and the possibility of analytical interpretation without excessive use of control tools. The conclusions outline the advantages and limitations of the approach: high accuracy and adaptability of the model require highquality data and mathematical training of managers. A promising direction for further research is the integration of regression methods with machine learning and Bayesian models to identify the causes of performance fluctuations and build personalized forecasts. The proposed model is aimed at small and medium-sized IT teams that need a practical, statistically sound alternative to complex corporate performance evaluation systems.