On a treasury corridor rewrite, after a custodian near-miss

The case is a single-family office with assets spread across  ,   and  . The office ran three custodians, two advisory relationships, and a treasury that was, in the principal's own words, a folder of one-off decisions the desk had been carrying for years. None of that was unusual. Most of the family offices we see in this season are running on institutional plumbing that was never written down. What was unusual was the near-miss.

In late spring 2026 the incumbent custodian in   signalled a transition to a successor entity as part of a regional consolidation. The transition was well flagged, well documented, and procedurally correct. It was also the first time the office had been asked, in writing, to confirm who had authority to move   across the corridor, and at what threshold, on what signature, against which reference rate. The office answered the letter three weeks late, with two names from memory and a fax number the principal had not used in a decade.

No money moved that should not have. The custodian, to its credit, did not treat the late reply as authority. The principal treated the late reply as a board-meeting moment. The desk was engaged the following Monday.

The work was a corridor rewrite, not a treasury rebuild. The distinction matters. A rebuild starts from the balance sheet and works outward. A corridor rewrite starts from the written policy and works inward, then back out again. The exercise is to read every existing document the office has signed in the last five years and ask one question: does this still hold against the corridor the principal wants in place for the next five?

The corridor map that came out of the twelve-week engagement is a single bound document. It runs to roughly   pages, of which the cover, the table of contents, the opening section, and the corridor grid are shown to the principal and the trustee on every quarterly cycle. The remaining sections sit in a sealed envelope in the principal's safe and are reviewed annually by  , the office's external auditor.

The substantive content is in four cells. Two axes, four cells, written rules on top. The two axes are household treasury and office treasury on one side, and reporting currency and operating currency on the other. The four cells are: household reporting, household operating, office reporting, office operating. Each cell carries a named corridor, a named hedge ratio, a named counterparty limit, and a named approver. The same template has been the desk's working draft for two years; what was new on this mandate was the discipline of getting every cell signed off before the office signed anything else.

Three numbers anchor the policy. None of them are the office's; they are the market's. The corridor is written against the   yield curve at the relevant tenors, the broad trade-weighted dollar index as the cross-currency reference, and a benchmark short-dated rate in   for the office's base-currency cash. The numbers move; the policy does not. The principal signs the policy once. The treasurer re-reads it quarterly and flags any cell that has drifted outside its corridor.

A corridor policy written in 2026 sits on top of a curve that has done more moving in two years than in the previous decade. The 1-month par yield closed June 2024 at 5.47 percent, fell to 4.28 percent a year later, dropped to 3.72 percent by end-January 2026, and ticked back to 3.78 percent by end-July 2026 (US Treasury, Daily Treasury Par Yield Curve). The short end has travelled roughly 170 basis points inside the window in which the principal signed the new policy.

The point of writing the corridor against the curve, rather than against an absolute number, is that the policy stops needing to be re-signed every time the Federal Reserve moves. The hedge ratio at the front end of the curve is set in basis points of carry, not in dollars of yield. The trigger threshold at the long end is set in standard deviations from the trailing two-year band, not in a target yield. Both rules can hold through a 100 basis point move without a re-signature.

The 10-year tells the other half of the story. From 4.36 percent at end-June 2024, to 4.24 percent a year later, to 4.26 percent at end-January 2026, and back to 4.75 percent by end-July 2026 (US Treasury). The long end held above 4.2 percent through the entire window while the short end fell 170 basis points. A corridor policy written against the absolute 10-year level would have looked responsible in mid-2024 and quietly wrong by mid-2026.

That asymmetry is what most offices miss when they sign a treasury policy as a one-page summary of intent. The summary reads well. The summary does not survive the rate cycle. The corridor map survives the rate cycle because it is written against the shape of the curve, not the level.

The corridor map is the framework the desk reaches for when an office has multi-currency cash, a custodian footprint across more than one jurisdiction, and a principal who wants the policy written once and re-read quarterly. The map has two axes and four cells. The axes are stable. The corridors drawn on top of the cells move with the curve, the FX reference, and the cross-border funding regime.

The work on this mandate was unusual only in that it was triggered by an event. Most corridor rewrites on retained mandates happen because the existing policy no longer matches the office the principal now runs, not because a custodian wrote a letter. The mechanics are the same. The desk reads the existing policy against the existing footprint, writes the gap, signs the new policy, and re-reads it on a standing cadence. The trigger does not change the work; it changes the urgency.

The currency half of the corridor is written against the broad trade-weighted dollar. The Federal Reserve's nominal broad index held in a 117 to 129 band from January 2024 through July 2026 (FRED, DTWEXBGS). That band is wide enough to wipe out a hedge ratio set in absolute terms, narrow enough to keep a corridor set in standard deviations honest.

For an office running cash in three jurisdictions, the practical question is which currency bucket is matched to which underlying liability. The match is rarely perfect; the policy is to name the gap, name the hedge, name the trigger, and re-sign only if the gap moves outside the corridor for more than two consecutive quarters. The discipline is the point. The principal does not need to take a view on the dollar. The principal needs the policy to take a view on the dollar on the principal's behalf.

The quarterly read is a one-page summary against a written corridor grid. The summary says which cells are in corridor, which are not, what action was taken, and what sign-off is required to move any cell outside its band. The principal signs the summary, not the underlying trade tickets. The treasurer signs the tickets. The trustee sees the summary once a year, in the context of the broader audit.

The measure of the rewrite, twelve months on, is not a number on a return line. It is whether the principal can answer the next custodian letter inside a week. The answer on this mandate is yes. The letter lands on a Monday. The desk has already replied by Friday.