Experiment

The starting observation: an IBM FlashSystem search surfaced three glossary pages tied to Krisada's work, on three different domains with three different implementations: https://www.as400ibmsystems.com/glossary/ibm-flashsystem/, https://flashsystem.as400system.com/glossary/flashsystem/, and https://www.midlandinfosys.com/about/glossary-of-ibm-iseries-hardware-terms/ibm-flashsystems.

The as400ibmsystems.com entry had recently been substantially built out. The flashsystem.as400system.com entry was extremely thin at the time it was observed, two short blurbs and a CTA, sitting on a FlashSystem-named subdomain of an AS400-focused domain. The Midland Information Systems entry came from an older, unmaintained glossary and directory architecture, with content that also isn't unusually deep and that nobody has actively optimized.

Because content depth varied wildly across all three while all three still co-occurred in the same SERP, the glossary-schema explanation weakened, and entity-level context, subdomain naming, domain topical focus, ecosystem interconnection, became the stronger working explanation.

Phase two, Entity Boot Camp, tests that explanation directly: the thin flashsystem.as400system.com page was rebuilt with substantially more content, semantic depth, structured data, and internal links, while the Midland property is being left untouched as a natural, unintervened control.

Control contamination note: the thin page was improved immediately after being noticed, before a baseline screenshot could be captured. Historical search and warehouse data are being used as the closest available proxy baseline instead.

The hypothesis is falsifiable and worth tracking systematically: monitor the same query set and the same three domains at set checkpoints, and watch whether the rebuilt page's position changes relative to the pre-rebuild baseline and relative to the untouched Midland control.

If the rebuilt page holds or improves while Midland stays flat, that supports entity-level context as the driver and strengthens the case for combining it with real content depth. If the rebuilt page's position doesn't move, or Midland moves just as much with no intervention at all, entity-level context was not doing the work the hypothesis assumed, and the pattern needs a different explanation.

More broadly, if this pattern holds across more query classes, entity architecture distributed across a portfolio may be more durable, and cheaper to build, than fighting for single-domain rank against entrenched entities on any one property.