Abstract

The B2B cybersecurity marketplace is characterised by long sales cycles, high organisational risk, and technically sophisticated decision-makers, conditions under which conventional rule- based lead qualification systematically misallocates sales effort. This dissertation develops, evaluates, and interprets a predictive lead-scoring framework for this setting, grounded in the Cross-Industry Standard Process for Data Mining (CRISP-DM) and applied to a synthetic B2B cybersecurity CRM dataset of 5,000 opportunities. Five classification algorithms — Decision Tree, Logistic Regression, Linear Support Vector Machine, Neural Network, and XGBoost — are benchmarked on accuracy, AUC, precision, recall, F1-score, and probability calibration. Logistic Regression is selected as the final model on three converging criteria: the joint-highest AUC of 0.970, a recall of 0.911 that is operationally decisive in a context where missed wins are costlier than false alarms, and a Brier Score of 0.0139 indicating strong probability calibration suitable for direct pipeline prioritisation. Model interpretation combines global SHAP importance, SHAP dependence plots on a complementary Random Forest, Partial Dependence Plots, and individual SHAP waterfall explanations. First, predictive power in this domain is concentrated in three binary process-and-stakeholder signals — proof-of-concept initiation, CISO engagement, and security-team involvement — each associated with a four- to five-fold uplift in mean predicted win probability. Second, the relationship between firmographic scale and win probability is non-monotonic: mid-market organisations ($50M–$150M revenue) represent the highest-probability segment, while very large enterprises and high-IT-spend accounts receive negative SHAP contributions. Organisa- tional and individual adoption dynamics are addressed conceptually through the Predictive Sales Analytics Adoption (PSAA) framework rather than through primary qualitative data, which is identified as a priority for future mixed-methods work.

Library of Congress Subject Headings

Computer security--Marketing--Forecasting--Data processing; Industrial marketing--Forecasting--Data processing; Predictive analytics; Regression analysis

Publication Date

7-23-2026

Document Type

Thesis

Student Type

Graduate

Degree Name

Professional Studies (MS)

Department, Program, or Center

Graduate Programs & Research

Advisor

Hammou Messatfa

Comments

This thesis has been embargoed. The full-text will be available on or around 4/29/2027.

Campus

RIT Dubai

Plan Codes

PROFST-MS

Download

Share

COinS