Colloid attachment and detachment behaviors concern a wide range of environmental contexts but have typically been mechanistically predicted exclusive of one another despite their obvious coupling. Furthermore, previous mechanistic prediction often addressed packed column contexts, wherein specific forces and torques on the colloid could not be well-constrained, preventing robust predictions. These weaknesses were addressed through direct observation experiments under conditions where perfect sink assumptions fail and allow calibration of the contact between the colloid and collector. Attachment and flow perturbation experiments in the presence of colloid-collector attraction (favorable conditions) permitted calibration of contact parameters without the complexity that comes with colloid-collector repulsion (unfavorable conditions). Combining calibrated contact parameters with discrete representative nanoscale heterogeneity, developed to predict unfavorable attachment, provided an independent means to predict unfavorable detachment. The result was mechanistic prediction of colloid attachment and detachment that quantitatively agreed with experimental observation for both ionic strength and flow perturbation results, improving significantly upon previous qualitative prediction.