Judgmental Forecasting in Crisis (generic process)

Judgmental Forecasting in Crisis (generic process) is the generic process card, GP-503, for the demand estimate a workforce function produces when a severity 1 or 2 event has made its statistical forecast invalid. The method is on Judgmental Forecasting and is not restated here. What this card adds is the operating process the crisis demands of that method: who may override at which horizon, what grade of evidence an override must carry, how long it stands before the baseline returns, and how it is scored once the event is over. It supplies step 5 of Crisis Trigger and Re-Plan (generic process), where business continuity is the partner function that owns activation and the workforce function owns the re-plan. It is part of Process Standardization and sits under the continuity-interface section of the function catalog in Anatomy of a ROC Standard.

The discipline rests on a replicated finding: judgmental adjustments are as likely to worsen a forecast as improve it unless they are large, documented and made on information the model could not have. A crisis is the case where the model certainly lacks the information, which is why the override is permitted, and where pressure to adjust is highest, which is why it is governed.

Process profile

The profile extends the L0 identity fields of Process Decomposition (L0–L3). Local values are held by the deploying function and marked SET LOCALLY.

Field	Generic value
Card	GP-503 · local card id assigned by the standards committee at deployment
Class	both: the override runs on the real-time clock; the scoring on the strategic clock
Trigger	a demand estimate requested by GP-501; a containment collapse under Platform Degradation Runbook and Buffer Staffing (generic process); or the forecasting lead's declaration that the model is invalid on the entry criterion (actuals outside the model's interval for a run of consecutive intervals, SET LOCALLY)
Cadence	event-driven; re-evaluated at the severity row's update cadence; scored at the post-event review and again in the quarterly review of Forecast Accuracy Governance (generic process)
Scope in	the model-invalid declaration; freezing the baseline; the evidence and its grades; the override tier by horizon and size; the expiry; the scenario set; the scoring; the adjustment of override authority
Scope out	the method (the concept page); the routine override outside a crisis; the re-plan that consumes the estimate (GP-501); the intraday reforecast in normal operation (Intraday Reforecast (generic process))
Inputs	the stored baseline; the incident record and scope; signals with a source grade; the scenario catalog's event class; the collision calendar
Outputs	the override record (baseline, adjustment, grade, tier, expiry, reason); the scenario set with the override as base; the expiry notice; the paired score per override, adjuster and event class
Upstream	GP-501 (the request); the real-time desk (signals); the continuity command (event information through the liaison)
Downstream	GP-501 step 7 (the requirement); scheduling; GP-106; Forecast Value Added in Workforce Management
Resources	the forecast system of record with versioning; the override log; the signal intake; the scoring workbook. Platforms are named at L3, never here
Decision points	is the model invalid; is the evidence graded well enough for the tier; renew, revise or expire; adjust authority
Authority tier	tier 1, the analyst seat inside guardrails (the real-time analyst for the intraday horizon inside a band, SET LOCALLY); tier 3, the operations manager or ROC duty lead, for the day-to-week horizon or outside the band; tier 4, the function head, for an override that changes the plan of record or a client commitment; the executive sponsor by exception
Risk and compliance flags	an override made under stakeholder pressure rather than on evidence; an override carried past its event; a client-contracted forecast lock; an [A] size carrying a plan, which Human Gates and Number Grades forbids
Variant-prone steps	the authorization step where a client contracts the forecast; the baseline step for a channel with no statistical model; the tier table for predecessor-organization forecasters
Automation candidates	five L1 steps (listed below)

L1 and L2

The method at L1 and L2 is on Judgmental Forecasting: baseline first, document every adjustment, constrain by size, track accuracy separately, template recurring events, demote poor adjusters; the paired-evaluation arithmetic is on Forecast Value Added in Workforce Management. A deploying function copies both through Wiki:Packs/Process Decomposition. What follows is the crisis flow that wraps the method; its sub-steps are the method's, so no separate L2 table is carried.

1. Is the model invalid? Test the entry criterion, or accept the GP-501 request as the declaration. Owner: the forecasting lead.
2. Freeze and store the unadjusted baseline, versioned, before any adjustment is discussed. Owner: the forecast analyst.
3. Name the tier the override will need, from its horizon and its size relative to the baseline. Owner: the forecasting lead.
4. Collect the evidence: each signal with its source grade in words, the queues it names, a direction and a size with a range. A signal missing any of the three stays proposed. Owner: the forecast analyst.
5. Is the evidence graded well enough for the tier? An [A] size with no range may carry an intraday override at tier 1 and nothing larger; a re-plan requirement needs an [E] with a range. Owner: the forecasting lead.
6. Build the scenario set with the override as the base case and bounds above and below it, per Scenario Planning and Contingency Staffing. Owner: the forecast analyst.
7. Set the override's expiry: hours or days by event class (SET LOCALLY), after which the baseline returns unless renewed. Owner: the forecasting lead.
8. Authorize at the tier named in step 3; the signer writes a reason, not a click. Owner: the seat at that tier.
9. Issue the estimate, graded, to GP-501 and to the real-time desk. Owner: the forecast analyst.
10. At each cadence: does the evidence still hold? Renew with new evidence, revise, or let it expire. Owner: the forecasting lead.
11. At expiry: the baseline returns, or the event enters the collision calendar as a known event with its own adjustment. Owner: the capacity planner, who confirms events on the calendar.
12. Score each override against its stored baseline by the paired method, per override, adjuster and event class, and pass the scores to GP-106. Owner: the forecast analyst.
13. Adjust override authority on the scores: retain, retrain or remove. Owner: the function head.

Variants

Recorded per Standard and Variant in Process Documentation:

• Statutory. None known; the class takes no row.
• Contractual. A client whose contract fixes a forecast lock, or supplies the forecast, inserts a client-notification step between steps 8 and 9 under a variant row owned by the account seat, reviewed at renewal; the lock itself is on Forecast Lock Process.
• Structural. A channel with no statistical model, typically an asynchronous backlog, has no baseline to freeze at step 2; the override is the forecast, and the step is varied with a review date until a model exists.
• Transitional. Forecasters from a predecessor organization keep their former override tier under a variant row until certified on the shared tier table, with an expiry the standards committee dates.

L3 register

Job aid	Anchored to	Form
The entry criterion sheet: the model-invalid test and its run length	step 1	work instruction
The baseline freeze procedure in the forecast system	step 2	work instruction
The tier table by horizon and size band	steps 3 and 8	work instruction
The override request form and the evidence-grade table	steps 4 to 5	work instruction
The scenario-set template per event class	step 6	playbook
The expiry defaults by event class and the override log	steps 7, 10 and 11	work instruction
The paired-scoring workbook	step 12	work instruction; method on Forecast Value Added in Workforce Management
The authority review agenda	step 13	playbook

Automation candidates

Released per class of action on a published catch rate per Human Gates and Number Grades. The card carries no L2, so candidates are named at L1; the five are ranked by how easy a catch rate is to publish for each.

1. The baseline freeze (step 2): a versioned save, and an override made before the baseline is stored cannot be scored.
2. The entry-criterion detector (step 1): actuals against the model's interval over a run length; after #1 because the run length is set locally and a false trip is only known in hindsight.
3. The override log and expiry alarm (step 7): a register with dates; after #2 because a missed expiry is detected only when a stale adjustment is found in a later forecast.
4. The scenario-set computation given an override (step 6): arithmetic on the base case; after #3 because its inputs are [E] figures whose ranges a person supplied.
5. The paired scoring (step 12): the comparison is computed, the authority decision is not; last because a score is only publishable once the event has closed and the actuals are in.

The override itself (step 8) is never a candidate. An automated override is a model, and the method's premise is that the adjuster holds information the model does not; the textbook treatment makes the same point, that judgment earns its place only where it brings information rather than reaction. The authority adjustment (step 13) is a human act for the same reason the scoring must precede it: a forecaster's track record under pressure is the only evidence that distinguishes information from bias.

Maturity Model Position

Four scales on this wiki use the word level; the launch page states which is which. Judgmental Forecasting places the undocumented override at Level 2 on the WFM Labs Maturity Model™ and the stored, documented, scored override at Level 3; this card is its crisis form. A function whose crisis overrides carry a tier, a grade and an expiry, and are scored afterwards, is at Level 3; one that adjusts override authority on those scores and absorbs recurring event classes into the model as templates is at Level 4.

See Also

• Process Standardization — the hub and the inventory this card belongs to
• Judgmental Forecasting — the method: model, then override, then score
• Crisis Trigger and Re-Plan (generic process) — the re-plan this estimate feeds
• Forecast Value Added in Workforce Management — the paired scoring
• Forecast Accuracy Governance (generic process) — where the scores are reviewed on cadence
• Human Gates and Number Grades — the number and source grades an override carries
• Scenario Planning and Contingency Staffing — the scenario set around the override
• Standard and Variant in Process Documentation — how the variants above are recorded