Bladder cancer is a commonly diagnosed cancer, especially in men, and 70% of new diagnoses are considered non-muscle invasive bladder cancer (NMIBC). Bladder cancer is prone to high rates of recurrence, and this risk is greatest in high risk NMIBC. Intravesical bacillus Calmette-Guerin (BCG) is standard of care for reducing rates of recurrence for high risk NMIBC. Despite its favorable efficacy a significant proportion of patients do not have durable prolonged response to BCG and some patients progress to muscle invasive bladder cancer worsening prognosis. Predictive tools are needed in clinical practice to identify patients who are not likely to respond to BCG and need alternative treatments. The European Organization for Research and Treatment of Cancer (EORTC) and Club Urologico Espanol de Tratamiento Oncologico (CUETO) have proposed outcome prediction tables for NMIBC patients, providing risk stratification and recurrence and progression probability scores. While valuable in clinical practice, these tables have limitations and overestimate recurrence and progression for high risk NMIBC. Several efforts have attempted to refine our ability to better understand which patients derive the greatest benefit from BCG. T1 pathologic substaging, tumor budding, and artificial intelligence techniques studying pathologic features of the tumor and the microenvironment have been applied to high risk NMIBC as a means of identifying patients less likely to respond to BCG. Molecular markers, genomic alterations and transcriptomic signatures are promising and hold the potential to aid in forecasting tumor progression and response to therapy. However, their application is still in its initial phases and necessitates additional validation through further studies. This review describes both clinical and molecular risk factors that could prove beneficial in anticipating the response to BCG, with a particular focus on high-risk T1 bladder cancers.