Internal linking, measured
Internal linking was treated as housekeeping for years — technically correct, commercially uninteresting. Shepard's studies, correlating link counts, anchor-text variation and placement against ranking movement across large site samples, moved it into the category of things teams actively engineer.
Title tags and other unglamorous variables
He has published similar work on Google's title-tag rewriting behaviour, canonical handling, anchor-text diversity and other variables that mostly get treated as folklore. The throughline is testing widely-held beliefs against actual data rather than repeating them at conferences.