Schedule Override Value Index (SOVI)

GMI-RE-029Renewable EnergyIntegration-RequiredPattern O-5: Override Value / Trust

Definition

SOVI measures the net impact of planner modifications to the AI-generated maintenance schedule by comparing outcomes of planner-modified tasks versus tasks executed as the AI originally scheduled them. This metric determines whether planner overrides are adding value through local knowledge and experience or degrading outcomes through habit-based scheduling that undermines AI optimization.

The Measurement Gap

The Measurement Gap

When AI scheduling systems generate maintenance schedules, planners review and modify them before execution. Some modifications reflect genuine local knowledge: the planner knows a technician will be unavailable, knows a road is closed, knows that two assets should be serviced together because they share a crane mobilization. Other modifications reflect scheduling habits from the pre-AI era: the planner always schedules site visits on Tuesdays, always groups work by geography rather than priority, or always defers work the planner considers low-value regardless of the AI's assessment.

Without SOVI, organizations cannot distinguish value-adding overrides from value-destroying overrides. Both look the same in the scheduling audit trail: the planner changed the AI's recommendation. The organization either trusts all overrides (giving planners unconstrained authority to modify AI schedules) or restricts all overrides (removing the local knowledge that planners contribute). Neither approach is optimal.

SOVI provides the evidence to make this distinction. By comparing outcomes (first-visit fix rate, repeat failure rate, on-time completion) between AI-original and planner-modified tasks, the organization can determine whether overrides are net positive or net negative, and whether the answer varies by override type.

The Decision This Metric Supports

SOVI answers whether planner overrides of the AI schedule are improving maintenance outcomes or degrading them. When SOVI is consistently positive across a significant sample, planners are adding real value and their override authority should be preserved. When SOVI is near zero or negative, overrides are neutral or counterproductive, and the organization should provide feedback to planners about which override types improve outcomes and which do not.

Formula

Primary Formula:

SOVI = Outcome_Score_Override minus Outcome_Score_AI_Original

Where Outcome_Score is a composite:

Outcome_Score = w1 x First_Visit_Fix_Rate + w2 x (1 minus Repeat_Failure_Rate_180d) + w3 x (1 minus Overdue_Rate)

Variables:

Variable Definition
First_Visit_Fix_Rate percentage of tasks resolved on the first visit, computed per group (override vs. AI-original)
Repeat_Failure_Rate_180d percentage of tasks where the same failure mode recurred within 180 days, computed per group
Overdue_Rate percentage of tasks that exceeded their due date, computed per group - w1, w2, w3 = weighting factors. Recommended starting weights: 0.4, 0.4, 0.2 (first-visit fix and repeat failure weighted equal...

Measurement frequency: Quarterly, with a minimum of 50 override events and 50 AI-original events per quarter for statistical significance. Smaller samples produce noisy SOVI values that should not drive policy decisions.

Segmentation requirements: Analyze SOVI by override type (date changes, priority changes, bundling changes, sequence changes). Different override types may have different value signatures. Date changes motivated by local knowledge (weather, availability) may show positive SOVI while priority changes motivated by planner preference may show negative SOVI.

Edge cases requiring special handling:

  • Overrides motivated by information the AI did not have (emergency at another site, technician illness, parts delay): these are legitimate overrides that do not reflect AI quality. Tag override reasons and analyze separately.
  • Confounding from selection bias: planners may override tasks they believe are harder, producing an apparent negative SOVI that reflects task difficulty rather than override quality. Control for task complexity if sufficient data supports the analysis.

Proposed formula, pending validation pilots.

At a Glance

Unit
Score
Trend Direction
Higher; overrides improve outcomes
Reporting Frequency
Quarterly
Category
Maintenance Scheduling Intelligence
← PM Starvation Index (PMSI)Task Bundling Compression Ratio (TBCR) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment