Three cases where the exact figure can be shown, because the sites are mine or I operate them. Each one carries its measurement window, its source, and the months that went badly, which are the ones almost nobody publishes.
The finding: between April and July the number of URLs receiving impressions fell from 192 to 154 while clicks multiplied by six. The metric most reports open with would have shown a 20% decline during the best quarter this project had. Zero link building.
The finding: a drop on a site of mine, with the test that tells lost rankings apart from a changed query mix. Google's own guide is the best result for this question, and this starts where that guide stops.
The finding: two pairs of sites I operate, same subject in English and Spanish. The Spanish half converts about three times better at comparable positions, and all four candidate explanations are inside, including the one that would weaken it.
The four recipe properties in this portfolio are client work: they are not named and their figures are rounded. Sites I own appear by name with exact numbers. That difference is deliberate and each case explains it. The rest of the work, without public figures, is on the work page.
That is exactly what the diagnostic delivers.