A foreign-issuer market cap, hidden inside a registry's own redundancy. How a 10× understatement was concealed in plain sight, and the single load-bearing decision that surfaced it. Implementation details are sealed; the doctrinal lesson is public.
On a routine spot-check, the terminal's market cap for a well-known foreign-domiciled issuer returned a number that did not survive a single glance — roughly an order of magnitude below the figure any analyst would expect. The price was right. The deposit ratio was right. The arithmetic was right. Only the share count was off, by a factor that does not occur by accident.
▸ The Concealment
The regulatory registry reports an issuer's share count under more than one concept name. For domestic filers the values agree — same number, same period, multiple paths to the same answer. For foreign-issuer ADRs the values diverge. One concept reports the count in the post-deposit US-listed receipt units. Another reports the count in the underlying ordinary (foreign-listed) units. The two values differ by the deposit ratio.
Both values are simultaneously correct. They describe different things. The registry does not warn you. The schema does not flag the divergence. The only way to know is to know.
▸ The Tie-Break
When both concepts report on the same period with identical acceptance timestamps, the picker is forced to choose. The naive implementation iterates the list of candidate concepts in declaration order and accepts the first value that arrives. For domestic filers this is harmless — every iteration order gives the same answer. For foreign-issuer ADRs it is catastrophic. The list order silently decides whether the market cap is correct, or wrong by the deposit ratio.
The fix is conceptual, not algorithmic. The precedence between the two concepts must be named. It cannot be inherited from whichever concept happened to appear first in the source file. The picker now consults an explicit precedence table; the ordinary-units concept wins the tie; the post-deposit concept loses; the secondary fall-back (a third concept that includes treasury holdings) is ranked lowest.
▸ The Surface Verification
The fix lands. The misreported issuer's market cap moves from a tenth of its true value to its true value. Domestic regressions are nil — the precedence table is invariant on companies that agree across concepts. A spot-check on three other foreign issuers (each with a different deposit ratio) confirms the correction propagates as expected.
Implementation specifics are sealed — the function names, the source paths, and the exact code are not public. The doctrinal lesson is. See Rule 26 of the 33: registry concept tie-breakers must be explicit, not implicit.
▸ The Lesson
Iteration order is a load-bearing decision. When two authoritative sources disagree by an order of magnitude and both are simultaneously correct, the question is not which is right. The question is which one your code believes first. The answer cannot be implicit. It has to be explicit, named, and justified in a comment next to the rank table. Anything else is a market cap waiting to be off by a factor of N.