Multi-digit number processing was long considered to be based on an analog holistic representation of magnitude. In 2001, however, the unit-decade compatibility effect was published, which suggested decomposed processing and facilitated an international research agenda on place-value representations and processing. The unit-decade compatibility effect describes that responses to incompatible pairs (47_62; decades and unit comparisons lead to different decisions) take longer and are more error-prone than those to compatible pairs (42_57; decades and unit comparisons lead to the same decision). Now, after 25 years, we conducted a quantitative meta-analysis to examine the strength and homogeneity of the unit-decade compatibility effect. We consulted two major databases, which lead to 154 articles. After removal of duplicates and having excluded all studies that were irretrievable or failed to meet our inclusion criteria, we included 63 studies and 68 experiments in our final analyses. The average effect size from 56 experiments, where effect size was provided or could be derived, was η2 = .37. However, individual effect sizes varied strongly depending on factors related to study design and on sample characteristics leaving space for moderator variables or other biases. Overall, this meta-analysis suggests that the unit-decade compatibility effect has a large and significant effect size, however, it is modulated by other relevant variables. Consequently, the unit-decade compatibility effect is well-suited to investigate place-value processing alongside its potential moderators.