What they did
SearchPilot took a client whose breadcrumb structured data was broken, showing raw URLs in the search snippet, and fixed the markup so proper breadcrumbs would display. They ran the change as a controlled split test, splitting pages into matched control and variant groups across three page types so seasonality and algorithm noise could be separated from the real effect.
What they found
The fix lost traffic on every page type. Product pages saw an estimated 12% loss in organic traffic, subcategory pages 6%, and high-level category pages 7%. The corrected markup made Google display a truncated breadcrumb, dropping the most specific subcategory, which gave a less compelling snippet and fewer click-throughs.
How much weight to give it
This is a genuine controlled experiment with control and variant groups, so the traffic drop is real for this site rather than a guess. It is a useful warning that a change which is technically more correct can still hurt performance if it changes how the snippet looks.
What it doesn't prove
It comes from a single e-commerce site and hinges on how Google rendered that site's breadcrumbs in the snippet. It does not prove breadcrumb markup is bad for SEO in general, only that this particular fix produced a worse snippet here.