4.2 KPI dictionary (OEE definitions, loss taxonomy, calculation rules)

A KPI (Key Performance Indicator) serves as a navigation instrument, rather than a punitive “report card.” If the dashboard displays “Green” but the production line is stopped, the dashboard is misrepresenting the truth. To manage operations efficiently via digital systems, we must strip away “Vanity Metrics” and enforce consistent, rigorous definitions for every number displayed on the floor.

The gold standard: OEE (overall equipment effectiveness)

OEE represents the objective measure of value-added time. It accounts for every second the machine is not producing good parts at its maximum theoretical speed, highlighting areas where we are losing valuable capacity.

The Formula: OEE = Availability × Performance × Quality

Availability (a)

Definition: The percentage of scheduled time the machine was actually running.
Formula: Run_Time / Planned_Production_Time
The Trap: Avoid deducting “unplanned breaks” from the denominator in an attempt to make the number look better.
Rule: When a shift is 8 hours and experiences 1 hour of breakdowns, the Availability is fundamentally 87.5%, not 100%.

Performance (p)

Definition: The speed at which the machine ran compared to its theoretical maximum.
Formula: (Total_Count / Run_Time) / Ideal_Run_Rate
The Trap: It is best practice not to use a “Budgeted Rate” or “Average Rate.” Instead, use the “Nameplate Rating” (the physical limit of the machine).
Rule: When the machine can achieve 100 UPH (Units Per Hour) but is scheduled for 80 UPH, running at 80 UPH results in 80% Performance, not 100%.

Quality (q)

Definition: The percentage of units that were good the first time (First Pass Yield).
Formula: (Total_Count - Defect_Count) / Total_Count
The Trap: Avoid counting “Reworked Units” as Good in this primary metric. Rework hides the true cost of failure.

Canonical KPI dictionary

Adopt these standardized definitions across the facility. When Finance calculates “Efficiency” using different parameters, they are calculating a financial variance, rather than a true engineering metric of machine performance.

Metric	Formula / Logic	Source	Refresh	Owner
OEE	A × P × Q	MES	Real-time	Ops Mgr
TEEP	OEE × (Planned_Time / 24 Hours)	MES	Shiftly	Plant Mgr
MTBF	Total_Run_Time / Count_of_Failures	CMMS/MES	Weekly	Maint Mgr
MTTR	Total_Downtime / Count_of_Failures	CMMS/MES	Weekly	Maint Mgr
FPY	(Input - Fails) / Input (First Pass Only)	MES (Test)	Real-time	Quality
Utilization	Run_Time / 24 Hours	Machine State	Real-time	Planner
Cycle Time	Timestamp_Out - Timestamp_In	MES (Tracking)	Real-time	Process Eng

Loss taxonomy (the reason tree)

We cannot effectively fix an abstract “Efficiency Loss.” We can, however, address a tangible “Feeder Jam on Slot 4.” We must adhere to a standardized hierarchical tree for downtime categorization, often aligning with the ISA-95 standard.

Level 1: the major state

Running: Producing units.
Idle: Experiencing no demand (Starved or Blocked).
Unplanned Down: Experiencing Failure or Error.
Planned Down: Assigned to Maintenance, Setup, or Break.

Level 2: the root cause (mandatory)

When a machine stops, the MES should prompt the operator (or read the PLC error code directly) to assign a Level 2 Reason.

Availability Losses (Stops):
- Equipment Failure: Motor, Sensor, Belt, Software Crash.
- Setup/Adjust: Changeover, Calibration, Warm-up.
- Material: Starvation (No Input), Blockage (Buffer Full), Material Empty.
Performance Losses (Slowdowns):
- Micro-stops: < 2 Minutes (Jams, Mis-picks). Auto-categorize these.
- Speed Loss: Running at reduced rate due to quality risk.
Quality Losses (Defects):
- Scrap: Material destroyed.
- Rework: Process repeated.

Calculation rules & governance

Calculation algorithms must be standardized and consistent across all production lines to ensure comparable data.

The “ideal cycle time” constant

Definition: The absolute fastest time the machine can theoretically process one unit (e.g. 10.5 seconds).
Governance: Stored centrally in the MES Master Data (Product-Resource relation).
Lock: Typically, only Process Engineering can authorize updates to this value.
Rule: When Performance exceeds 105%, it strongly indicates the Standard Cycle Time is incorrect. Engineering should investigate and adjust the Master Data.

Handling micro-stops

Operators cannot realistically explain every 30-second stop, as doing so would negatively impact their workflow and productivity.

Rule: When downtime is less than 2 minutes, the system should auto-tag it as a “Micro-stop” or “Minor Stoppage.”
Analysis: Engineering reviews the aggregate “Micro-stop” bucket weekly to identify chronic, underlying issues.

Starvation vs. blockage (the conjoined twins)

Starvation: The Machine is ready, but the upstream process has sent no parts. (The fault lies Upstream).
Blockage: The Machine is ready, but the downstream conveyor is full. (The fault lies Downstream).
Logic: The MES should read the sensors at the in-feed and out-feed to auto-classify these states, rather than asking the operator to guess.

Changeover time (SMED)

Start Trigger: Last Good Piece of Old Product.
End Trigger: First Good Piece of New Product.
Measurement: This includes run-down, setup, material loading, and first-article validation.

Final Checkout: KPI dictionary (OEE definitions, loss taxonomy, calculation rules)

Metric	Metric / Control	Threshold / Rule
Math	OEE Availability	Ensure Denominator is Planned Production Time, not Total Calendar Time.
Speed	Ideal Rate	Base this on “Nameplate” (Physics limits), not “Budget” (Finance targets).
Taxonomy	Hierarchy	Map all scheduled downtime to the standard Tree (Avail/Perf/Qual).
Logic	Micro-stops	System auto-codes stops < 2 mins. No user input is required here.
Quality	FPY	Treat reworked units as 0% Quality for that specific pass.
Governance	Master Data	Engineering formally manages Cycle Times.
Drift	Performance Cap	When Performance consistently > 105%, flag a potential Master Data Error.