When the Model and Property System Disagree
How to produce a point-in-time portfolio answer without averaging away the conflict.
By Nolan Grout
A portfolio answer can be individually accurate in two systems and still be wrong for the decision.
The model may use a forecast. The property platform may hold posted actuals. A lease abstract may reflect an executed amendment that has not reached a reporting export. Before choosing a value, the system must reconcile meaning, entity, period, and authority.
Match the identity
The same asset, tenant, lease, or loan may have different identifiers across systems. Entity resolution should be explicit and reviewable. A confident fuzzy match is not enough for a material answer.
Match the definition
“Occupancy,” “NOI,” and “maturity” can mean different things across teams. The system should preserve the definition attached to each value instead of treating a shared label as proof of equivalence.
Match the time
A point-in-time answer needs an as-of date. Record both when a value was effective and when the system learned it. That distinction matters when a correction arrives after a report was published.
Match the authority
Different sources can be authoritative for different fields. The executed lease may govern contractual dates. The ledger may govern posted actuals. The approved model may govern forecast assumptions.
Authority should be configured by field and use case, not inferred from whichever API responded last.
Put unresolved differences in a queue
A useful conflict record includes:
- the competing values;
- source and effective date for each;
- the decision affected;
- materiality;
- assigned owner;
- resolution and approver.
Answer with the conflict attached
Sometimes the correct executive answer is not one number. It is: “The ledger shows X, the operating report shows Y, and the difference is assigned for review.”
That answer may be less polished, but it is more trustworthy than false precision.