An increase in the number of studied Type Ia supernovae (SNe Ia) has demonstrated that this class of explosions has a greater diversity in its observables than was previously assumed. The reasons (e.g. the explosion mechanism, progenitor system) for such a diversity remain unknown. Here, we analyse a sample of r-band light curves of SNe Ia, focusing on their behaviour ∼ 2-4 weeks after maximum light, i.e. the second maximum. We characterise the second maximum by its timing (t r 2 ) and the integrated flux (F r 2 ). We find that t r 2 correlates with the "colour-stretch" parameter s BV , which can be used as a proxy for 56 Ni mass, and F r 2 correlates with the transparency timescale, t 0 . Using F r 2 for a sample of 199 SNe from the Palomar Transient Factory and intermediate Palomar Transient Factory, we evaluate a distribution on t 0 for a sample of SNe Ia found in an "untargeted" survey. Comparing this distribution to the predictions of t 0 ranges from models we find that the largest overlap in t 0 values between models and observations is for the sub-Chandrasekhar double detonation models. We also compare our relations between t 0 and F r 2 with that from 1-D explosion models of Goldstein & Kasen (2018) and confirm that F r 2 can be used as a diagnostic of the total ejecta mass.