Digital practice packages combine feedback, reattempts, and performance cues, yet little is known about how post-activity interest relates to later retention. Guided by self-determination theory, we compared a digital quiz package with a paper worksheet using the same 36-item pool in 30 fifth-grade classrooms from 24 primary schools (N = 437). Within classrooms, two algorithmically balanced groups were formed and randomly assigned to conditions. Students completed a pretest, practice activity, intrinsic-interest measure, and delayed posttest 7–10 days later. In the available-outcome analysis (N = 397), students assigned to the digital package showed higher delayed retention after adjustment for pretest; the classroom-clustered estimate was B = 2.62, 95% confidence interval (CI) [1.75, 3.50]. In complete-case mediation (N = 391), the digital package was associated with higher intrinsic interest (B = 0.95, 95% CI [0.74, 1.17]), and the estimated indirect effect was positive (B = 1.21, 95% CI [0.40, 2.09]). Because the mediator was not randomised, the pathway is associational, not causal.