When the whole field is wiped out and replaced

Research question

When the buyer adds a requirement that the whole named field cannot satisfy, does the recommendation survive as a plural set, and if not, what replaces it?

Why it mattered

A fourth kind of category for the survivability runs: a duopoly, led by QuickBooks and Xero. A duopoly might be expected to protect the field, since two strong incumbents share it. This run tested whether that structure survives a requirement that neither incumbent can meet natively.

Method

Evaluator: Google AI Mode, single surface, logged out. Three draws, each a fresh conversation. The run began from an unqualified category recommendation, then introduced an e-commerce company profile, then native multi-currency and multi-entity consolidation rather than third-party add-ons, then native Shopify and Amazon integration with built-in inventory. Instrument details and scoring are retained internally.

Direct observations

• Both category leaders, QuickBooks and Xero, were eliminated at the multi-entity requirement in all three draws.
• Once the leaders were removed, the recommendation relocated to a higher-tier category, entry-level ERP, and one product, Oracle NetSuite, was the surviving anchor in all three draws.
• None of the five vendors that opened the field survived to the end. The final vendors were introduced only after the field had migrated to ERP.
• The final field collapsed to a single vendor in two of the three draws, and to two in one draw.
• The category leader survived the buyer profile but was eliminated at the second requirement, not the first, a delayed version of the pattern seen in other categories.
• Pricing and integrations were volunteered without being asked in every draw.

Interpretation (not established)

In this category and on this surface, the recommendation did not survive as a plural set. It was replaced. The duopoly structure did not protect the field: one requirement removed both leaders at once and pushed the recommendation up a market tier to a product that was never in the opening list. This is close to the collapse seen in another category but mechanically different, because the survivor is a new entrant rather than a surviving incumbent. Whether this is a distinct movement class or a form of collapse is not established here.

Outcome and remaining uncertainty

Not Supported: the survival prediction, that two or more vendors remain at the end in a majority of draws, did not hold. The majority collapsed to a single new-tier vendor. This is a real result, not a failure of the run: the wipeout and migration pattern replicated cleanly across all three draws, and the secondary predictions about the leader, the requirement that removes it, and the volunteered pricing all held. The scope is narrow. One evaluator, one surface, three draws, no removal test, nothing causal. How to code the movement class is a pending founder decision, surfaced as a candidate observation.