"Viability fell to 70%" is meaningless until you answer: 70% of what? That reference point is the baseline, and choosing it is the most consequential decision in the workflow.
The naive baseline — the first test — under-reads systematically: dormancy isn't fully broken yet, protocols improve, and germination often peaks a year or two into storage. So the pipeline uses an early-window empirical baseline: the highest germination observed in the accession's first five tested years. It reads the accession at its demonstrated best, which makes the threshold measure genuine decline.
Every knob here is an assumption you can test. Widen the window and the baseline can only rise (thresholds get stricter, intervals shorter); narrow it and vice versa. Rerunning at 3, 5, and 7 years and flagging the taxa whose answer changes turns an arbitrary choice into a measured one — in the dataset behind the catalog, 5 years is stable for roughly 96% of taxa.