Properties of light are difficult for elementary students because the underlying processes can only be observed through their effects, and flipbook-based e-modules are often promoted as a way to make these processes visible. Most evidence for such e-modules, however, comes from development studies and single-group pre-test–post-test comparisons. This study tested whether an interactive flipbook-based e-module produced higher learning outcomes on the properties of light than regular instruction once pre-existing achievement was taken into account. A quasi-experimental pre-test–post-test control-group design was used with two intact fifth-grade classes at a public elementary school in Indonesia (N = 22; experimental n = 12, control n = 10). Both classes took the same 0–100 achievement test before and after instruction. The experimental class started 11.92 points higher (Hedges g = 0.73), so the primary analysis was an analysis of covariance (ANCOVA) with the pre-test as covariate, supported by permutation, bootstrap, rank-based, and robust sensitivity analyses. Both classes improved substantially, by about 33 and 35 points. After baseline adjustment, the e-module class did not outperform the control class, F(1, 19) = 0.25, p = .626, partial η² = .013; adjusted difference = −3.38 points, 95% CI [−17.65, 10.90], and all sensitivity analyses agreed. The e-module was therefore not superior to regular instruction in this setting, although the small sample cannot exclude moderate effects in either direction. The study also shows that within-group gains and normalised gain can overstate the benefit of digital learning materials when baseline differences between intact classes are ignored.