Stewardship Escalation Path

Finance says a refunded order is not a sale. Growth says it is, because the acquisition happened. Both are right for their purpose and both are reporting to the board. The steward for the orders domain writes a careful note explaining the two positions, circulates it, and nothing happens for five months, because the steward can describe a disagreement and cannot end one.

An escalation path is the part of stewardship that converts a description into a decision. It names who decides when the steward cannot, states how long that takes, and says what happens by default if the deadline passes. Without those three elements a stewardship programme accumulates open questions at a rate set by how much the business changes, and the accumulation is invisible because nothing is failing.

The practice is unglamorous and it is the difference between a stewardship role and a title in a spreadsheet.

Why it matters

Unresolved definitional disputes do not stay unresolved. They resolve locally and silently: each team implements its own reading, ships it, and within two quarters there are two production metrics with the same name. The cost of the eventual reconciliation is roughly quadratic in how long the divergence ran, because every downstream artefact built on either reading has to be found and re-pointed.

There is a second effect that programmes underestimate. A steward who cannot escalate learns not to raise things. After two or three disputes that went nowhere, the steward stops surfacing the fourth, so the programme's own measure of contested terms falls and leadership reads it as maturity.

Implementation patterns

• Name a single decider per domain, not a committee. A committee's decision date is the date of its next quorate meeting, which is the failure mode dressed as governance.
• Put a clock on it. 10 days from the steward's escalation to a recorded ruling is a workable default. The clock matters more than its length, because an unbounded path is indistinguishable from no path.
• State the default outcome if the clock expires. Two options work: the incumbent definition stands and the entry is marked contested in the catalogue, or the steward's proposal takes effect. Choose one and write it down; an expiry with no default restores the original problem.
• Record the ruling where the definition lives, with the date, the decider and the reasoning in two sentences. The reasoning is what prevents the same argument in 18 months with new people.
• Publish the queue. Open escalations, their age, and the decider each is waiting on. A backlog of 14 items all waiting on one executive is a resourcing fact, not a stewardship failure.
• Allow both definitions when both are legitimate. The ruling is then about which one carries the unqualified name and which must be qualified everywhere it appears.

Industry example

Characteristic rather than attributed: a marketplace operating in two regulated markets found 31 open definitional disputes after a year of stewardship, with a median age of 4 months and no record of anyone being asked to decide. Introducing a named decider per domain with a 10-day clock cleared 23 of them in a quarter, and roughly a third were resolved by the decider simply declaring that both definitions were valid and assigning the unqualified name to one of them. The remaining eight were genuine conflicts that needed an executive, which is the population the path exists to surface.

Failure scenarios

• The path exists on paper and routes to a committee, so the effective service level is the meeting cadence and escalation is slower than not escalating.
• The decider has no authority over one of the disputing functions. The ruling is then advisory and the losing function ignores it, which teaches everyone that escalation is theatre.
• Rulings are recorded in the governance tool rather than in the catalogue entry, so the people who consume the definition never see them and the argument recurs.
• Everything is escalated. A steward who escalates routine questions burns the decider's patience, and the genuinely contested case arrives at an exhausted executive.
• No default on expiry, so a decider can hold a decision open indefinitely at no cost.

Trade-offs

The visible bottleneck is the honest part of the trade. A queue of open escalations attributed to a named person is uncomfortable and it is the only artefact that makes the cost of indecision legible.

When not to use it

With fewer than about three functions that can legitimately disagree, this is overhead. The disputes fit in one weekly meeting and formalising the route adds latency to conversations that were already happening. It is also the wrong instrument for data quality problems: a broken pipeline needs an incident process, not an arbitration path, and routing defects through a definitional route is how stewardship acquires a reputation for slowness it did not earn.

Interview question

Q: Your stewardship programme has stewards, a catalogue and a monthly forum, and after a year almost nothing contested has been decided. The stewards are engaged and the descriptions are current. What is missing, and what exactly would you add?

What a strong answer covers: the distinction between describing a disagreement and ending one; a named decider rather than a forum; a clock with a stated default on expiry; recording the ruling in the catalogue where consumers read it; publishing the open-escalation queue so indecision has a cost; and recognising that a falling count of contested terms can mean stewards have stopped raising them.

Quick check

Quiz: Why does an escalation path need a stated default when the clock expires? Without one, a decider can hold a decision open at no cost, which is indistinguishable from having no path; the default forces either the incumbent definition or the steward's proposal to take effect.

Flashcard: A steward can describe a dispute between finance and growth but not end it. What do you add? A named decider per domain, a 10-day clock, a written default if the clock expires, and the ruling recorded on the catalogue entry itself with its reasoning.