Enhancement of planning poker process using factors quantification for software effort estimation / Pantjawati Sudarmaningtyas

Faculty of Computing

Thesis (Doctor of Philosophy) -- Universiti Malaysia Pahang - 2026

Includes bibliographical references

Effort estimation remains a persistent challenge in Agile Software Development (ASD), despite extensive prior research. Widely adopted collaborative techniques, particularly Planning Poker (PP), continue to exhibit limitations, including high estimation dispersion, slow convergence, and inconsistent accuracy, largely due to unstructured consensus mechanisms and a strong reliance on individual judgment. To address these challenges, this study proposes Enhanced Planning Poker with Factor Quantification (EPPwFQ), a human-centric estimation approach that systematically integrates empirically validated factors into the Planning Poker process while preserving core Agile collaboration principles. The incorporated factors consist of project-related dimensions, namely requirement complexity and design–implementation characteristics, and peoplerelated dimensions, namely experience, domain knowledge, and technical ability. Grounded in the Design Science Research (DSR) paradigm, the study contributes by identifying and operationalizing key influencing factors, designing the E-PPwFQ artefact through structured factor quantification, and empirically evaluating its performance against established Agile estimation techniques, including Affinity Estimation, Delphi, Three-Point Estimation, and conventional Planning Poker. Controlled experiments were conducted using multiple Agile teams and projects, with productivity, efficiency, consistency, and accuracy as evaluation metrics. The results indicate that E-PPwFQ improves the speed of estimation convergence compared to conventional Planning Poker, as evidenced by a lower em/et ratio (3.66 versus 9.06), while also producing more consistent estimates, as reflected in a reduced coefficient of variation (57.74%), comparable to Three-Point estimation. Distributional analysis shows better balance and fewer extreme estimation values, indicating that agreement is forming more steadily. In terms of accuracy, E-PPwFQ demonstrates performance comparable to conventional Planning Poker (MRE = 1.0576 versus 1.0577), suggesting that its primary contribution lies in enhancing process efficiency and estimation stability rather than minimizing absolute estimation error. Overall, the proposed approach provides a structured and scalable enhancement to Agile effort estimation by improving convergence efficiency and consistency while maintaining a balanced level of accuracy suitable for real-world Agile project environments.