Simple SummaryThe EU-funded Welfare Quality® project generated species-specific assessment protocols to evaluate the welfare of pigs, poultry, and cattle. With the implementation to be used for certification purposes, it is important that the protocols show consistency over time, which describes the extent to which equal results are achieved if the assessment is performed repetitively. The protocols should not be sensitive to slight changes in the on-farm situation but mirror the long-term welfare situation on-farm. The present study aimed at testing this consistency over time of the indicators included in the ‘Welfare Quality® animal welfare assessment protocol for sows and piglets’. Thereby, the study focused on the indicators to assess the welfare principle ‘appropriate behavior’. As a result, the indicators applied to assess the animals’ ‘appropriate behavior’ did not represent consistency over time. Thus, further investigation is needed before implementation on-farm. Conclusively, the present study contributes to the development of generally accepted and objective assessment protocols for animal welfare and thereby to the improvement of farm animals’ welfare overall.The present study’s aim was to assess the test−retest reliability (TRR) of the ‘Welfare Quality® animal welfare assessment protocol for sows and piglets’ focusing on the welfare principle ‘appropriate behavior’. TRR was calculated using Spearman’s rank correlation coefficient (RS), intraclass correlation coefficient (ICC), smallest detectable change (SDC), and limits of agreement (LoA). Principal component analysis (PCA) was used for deeper analysis of the Qualitative Behavior Assessment (QBA). The study was conducted on thirteen farms in Northern Germany, which were visited five times by the same observer. Farm visits 1 (F1; day 0) were compared to farm visits 2 to 5 (F2–F5). The QBA indicated no TRR when applying the statistical parameters introduced above (e.g., ‘playful‘ (F1–F4) RS 0.08 ICC 0.00 SDC 0.50 LoA [−0.62, 0.38]). The PCA detected contradictory TRR. Acceptable TRR could be found for parts of the instantaneous scan sampling (e.g., negative social behavior (F1–F3) RS 0.45 ICC 0.37 SDC 0.02 LoA [−0.03, 0.02]). The human−animal relationship test solely achieved poor TRR, whereas scans for stereotypies showed sufficient TRR (e.g., floor licking (F1–F4) RS 0.63 ICC 0.52 SDC 0.05 LoA [−0.08, 0.04]). Concluding, the principle ‘appropriate behavior’ does not represent TRR and further investigation is needed before implementation on-farm.