One number, three answers: KPI governance at Vallora Logística

Three departments, three absenteeism rates for the same month. How 40 KPIs became 14, each with a definition sheet, an owner and a change rule.

Fictional case. Vallora Logística S.A., its departments, people and KPIs do not exist. The three absenteeism rates were calculated from the fictional dataset of the Absenteeism Radar; the inventory, the definition sheets and the rules were written for this case. The method is real.

Download the three monthly rates (CSV) · Download the KPI inventory (CSV)

The CSV files and the comments in the R code are in Portuguese, the case’s original language.

The story

At the July 2025 results meeting, Vallora heard three absenteeism rates for June. HR showed 4.38%, Operations 3.54% and Finance 2.15%, and the meeting spent half an hour arguing about which one was right.

Absenteeism rate calculated each department’s way, from January 2024 to December 2025. The vertical line marks June 2025, the month discussed at the July meeting.

Absenteeism rate calculated each department’s way, from January 2024 to December 2025. The vertical line marks June 2025, the month discussed at the July meeting.

None of them was wrong. Each department calculated what it needed for its own work and gave the result the same name. The problem was not arithmetic, it was governance: nobody had written down what “absenteeism rate” meant at the company, or who got to decide.

The scene repeated itself over the following months, and in December leadership asked for an inventory of every KPI that circulated in the monthly presentations. Forty turned up, from six departments, and only 13 were well defined. The work took five weeks, in four stages:

1. Interviews with the people who produce the numbers and the people who use them, in all six departments.
2. Inventory and diagnosis of each KPI.
3. Definition sheets for the KPIs that go to leadership.
4. Governance rules: who proposes, who approves, how a change is recorded.

The divergence: three calculations, one name

The three rates come from the same data and answer different questions. The decision was to keep all three, each with its own name, and reserve “Absenteeism rate” for HR’s definition, which is the same as the Radar’s.

	HR	Operations	Finance
Question it answers	How much of the contracted time was lost?	How many scheduled shifts went uncovered?	How much do we pay for hours not worked?
Numerator	Absence hours of every type, including long-term absences	Scheduled days with any absence, including late arrivals or early departures of more than 10 minutes	Absence hours paid by the company: doctor's notes, the first 15 days of sick or injury leave, and statutory leave
Denominator	Scheduled working hours	Scheduled days (people absent for more than 15 days come off the schedule)	Scheduled working hours
Unit	Hours	Days	Hours
2024 value	3.77%	3.55%	2.19%
2025 value	3.77%	3.48%	2.12%
Name after the project	Absenteeism rate (official, goes to leadership)	Schedule absence rate (managed by Operations)	Cost of paid hours not worked (managed by Finance)

HR

Question it answers

How much of the contracted time was lost?

Numerator

Absence hours of every type, including long-term absences

Denominator

Scheduled working hours

Unit

Hours

2024 value

3.77%

2025 value

3.77%

Name after the project

Absenteeism rate (official, goes to leadership)

Operations

Question it answers

How many scheduled shifts went uncovered?

Numerator

Scheduled days with any absence, including late arrivals or early departures of more than 10 minutes

Denominator

Scheduled days (people absent for more than 15 days come off the schedule)

Unit

Days

2024 value

3.55%

2025 value

3.48%

Name after the project

Schedule absence rate (managed by Operations)

Finance

Question it answers

How much do we pay for hours not worked?

Numerator

Absence hours paid by the company: doctor's notes, the first 15 days of sick or injury leave, and statutory leave

Denominator

Scheduled working hours

Unit

Hours

2024 value

2.19%

2025 value

2.12%

Name after the project

Cost of paid hours not worked (managed by Finance)

• Operations. Operations counts the whole day even when the absence is a late arrival, which pushes the rate up, but takes anyone absent for more than 15 days off the schedule, which pulls it down. On balance, it lands a little below HR. In four of the 24 months (January and February 2024, September and November 2025), the balance flipped: in months with few long-term absences, counting the whole day weighs more, and Operations comes out above HR.
• Finance. It leaves out unexcused absences and late arrivals that are docked from pay, and whatever INSS pays from day 16 of a leave. That is why it runs well below the other two. In Brazil, the employer pays the first 15 days of sick or injury leave; from day 16 on, the social security agency (INSS) pays.

# Tipos da base: 1 atestado médico, 2 consulta ou exame (parcial), 3 acidente
# de trabalho, 4 afastamento do INSS, 5 falta injustificada, 6 atraso ou saída
# antecipada, 7 licença legal (CLT art. 473), 8 acompanhamento de dependente,
# 9 licença maternidade/paternidade (dias já fora da jornada no Radar).
# RH: horas de ausência dos tipos que contam no Radar ÷ horas previstas.
rh <- ausencias |> filter(considera_taxa == 1)

# Operação: dias escalados com qualquer ausência (atraso ou saída antecipada só
# acima de 10 minutos) ÷ dias escalados. Quem está afastado há mais de 15 dias
# (do 16º dia de um evento longo em diante) sai da escala, no numerador e no
# denominador.
afastado <- ausencias |> filter(dias_corridos > 15, dia_do_evento > 15) |> distinct(id_funcionario, data)
escala <- jornada |> anti_join(afastado, c("id_funcionario", "data"))
dias_com_ausencia <- escala |>
  semi_join(filter(ausencias, id_tipo_ausencia != 6 | horas_ausencia * 60 > 10), c("id_funcionario", "data"))

# Financeiro: horas pagas pela empresa ÷ horas previstas. Entram atestados e
# acidentes de até 15 dias, consultas, licenças legais e os 15 primeiros dias
# dos afastamentos longos por doença ou acidente (tipos 1, 3 e 4), que a empresa
# paga por lei. Ficam fora faltas e atrasos (descontados), o que o INSS paga e
# o acompanhamento de dependente.
pagas <- ausencias |>
  filter(id_tipo_ausencia == 7 |
           (id_tipo_ausencia %in% c(1, 2, 3) & dias_corridos <= 15) |
           (id_tipo_ausencia %in% c(1, 3, 4) & dias_corridos > 15 & dia_do_evento <= 15))

Diagnosis: from 40 to 14

Of the 40 KPIs, 14 go to the leadership meeting, 14 stay with the departments and 12 leave the presentations, merged or discontinued. Each KPI was given a status and a destination, using the criteria below.

Status found	Criterion	Count
Well defined	Written formula, single source, an owner, and used in some decision	13
Inconsistent	Same name, different calculations across departments or sites	7
Duplicate	Different names for the same measure, or one is a slice of the other	7
Poorly designed	Volume with no reference point, a slice treated as a KPI, or the wrong incentive	6
Unused	Not part of any decision recorded in the minutes in the last six months	4
No owner	Nobody answers for the result or the calculation	3

Destination	What it means	Count
Leadership	Goes to the monthly meeting, with a full definition sheet, a target and an owner	14
Department level	The department tracks it with a short definition sheet; it goes to leadership only by exception	14
Merged	Becomes part of another KPI or a breakdown of it	6
Discontinued	Leaves the presentations; the data remain available	6

Where each KPI came from and where it went. The width of each band is the number of KPIs.

Where each KPI came from and where it went. The width of each band is the number of KPIs.

To go to leadership, a KPI has to pass three questions: it is tied to a company objective, someone in leadership makes a decision with it, and it moves enough from one month to the next to deserve the meeting.

The most instructive case is “Days without accidents”. It looks positive, but it rewards not reporting accidents; the Accident frequency rate measures the same thing without that incentive.

Inventory: the 40 KPIs

Only 13 of the 40 KPIs were well defined. The table shows the full inventory, with search and filters by department, status and destination.

All departmentsFinanceHROHSOperationsQualitySalesAll statusesWell definedInconsistentDuplicatePoorly designedUnusedNo ownerAll destinationsLeadershipDepartment levelMergedDiscontinued

40 of 40 KPIs

#	KPI (as it was called)	Department	Status	Destination	Note	Sheet
1	EBITDA margin	Finance	Well defined	Leadership		FIN-01
2	Logistics cost per order	Finance	Inconsistent	Leadership	Operations left freight out of the cost	FIN-02
3	Budget vs. actual expenses	Finance	Well defined	Leadership		FIN-03
4	Overdue receivables	Finance	Well defined	Leadership		FIN-04
5	Cost of paid hours not worked	Finance	Inconsistent	Department level	Was called absenteeism
6	Overtime cost	Finance	Duplicate	Merged	Goes into the Overtime sheet
7	Revenue per m²	Finance	No owner	Discontinued
8	Energy cost per pallet	Finance	Unused	Discontinued
9	OTIF by invoice	Sales	Inconsistent	Merged	Merged into OTIF by order
10	Customer NPS	Sales	Well defined	Leadership		CLI-02
11	Customer complaints	Sales	Poorly designed	Leadership	A count with no base; becomes complaints per thousand orders	CLI-03
12	Complaint response time	Sales	Well defined	Department level
13	Active customers	Sales	Poorly designed	Department level	Volume, no target
14	Average ticket per customer	Sales	Unused	Discontinued
15	OTIF by order	Operations	Inconsistent	Leadership	Becomes the official OTIF	CLI-01
16	Picking productivity	Operations	Well defined	Leadership		OPE-01
17	Inventory accuracy	Operations	Well defined	Leadership		OPE-02
18	Picking errors	Operations	Well defined	Leadership		OPE-03
19	Warehouse occupancy	Operations	Inconsistent	Leadership	Each DC measured it in positions or in m²	OPE-04
20	Damage rate	Operations	Duplicate	Department level	Absorbs the damage figures from Quality
21	Orders shipped per day	Operations	Poorly designed	Department level	Volume, no target
22	Average dock time	Operations	Well defined	Department level
23	Schedule absences	Operations	Inconsistent	Department level	Was called absenteeism
24	Operations overtime	Operations	Duplicate	Merged	Goes into the Overtime sheet
25	Orders per employee	Operations	Duplicate	Discontinued	Repeats picking productivity
26	Night team productivity	Operations	Poorly designed	Merged	A slice; becomes a shift filter
27	Customer-reported damage	Quality	Duplicate	Merged	Goes into the damage rate
28	Internal audits completed	Quality	Unused	Discontinued
29	Absenteeism rate	HR	Inconsistent	Leadership	Becomes the official definition	PES-01
30	Absenteeism from doctor's notes	HR	Duplicate	Merged	Becomes a breakdown of the official rate
31	Overtime percentage	HR	Duplicate	Department level	Single Overtime sheet
32	Turnover	HR	Well defined	Department level
33	90-day turnover	HR	No owner	Leadership	Calculated by hand, only now and then	PES-02
34	Headcount	HR	Poorly designed	Department level	Volume, no target
35	Training hours per person	HR	Well defined	Department level
36	Average time to hire	HR	No owner	Department level
37	Accident frequency rate	OHS	Well defined	Leadership		PES-03
38	Accident severity rate	OHS	Well defined	Department level
39	Days without accidents	OHS	Poorly designed	Discontinued	Rewards not reporting accidents
40	Safety inspections completed	OHS	Unused	Department level	Gets a target and stays with OHS

KPI tree

Fourteen KPIs, four objectives, one goal

2026 goal: grow with healthy margins without missing delivery dates

Finance

Grow with healthy margins

• FIN-01 EBITDA margin
• FIN-02 Logistics cost per order
• FIN-03 Budget vs. actual expenses
• FIN-04 Overdue receivables

Customers

Deliver on time

• CLI-01 OTIF
• CLI-02 Customer NPS
• CLI-03 Complaints per thousand orders

Operations

Accuracy and efficiency

• OPE-01 Picking productivity
• OPE-02 Inventory accuracy
• OPE-03 Picking errors
• OPE-04 Warehouse occupancy

People

Attendance and safety

• PES-01 Absenteeism rate
• PES-02 90-day turnover
• PES-03 Accident frequency rate

Foundation: definition sheets, KPI committee and version log

Each leadership KPI answers to one of the four objectives, and the code prefix (FIN, CLI, OPE, PES) tells you which. The codes keep their Portuguese initials, as in the original project.

Definition sheets

Each of the 14 leadership KPIs got a definition sheet with the same fields. The Absenteeism rate sheet is already open, as the model; the others open with a click.

Finance: grow with healthy margins

FIN-01 EBITDA marginEBITDA ÷ net revenue × 100

Code

FIN-01

Question it answers

How much of revenue is left after operating costs and expenses?

Linked objective

Finance: grow with healthy margins

Formula

EBITDA ÷ net revenue × 100

Numerator

EBITDA for the month (earnings before interest, taxes, depreciation and amortization)

Denominator

Net service revenue for the month

Excluded from both

Non-recurring revenue and expenses, recorded separately and explained in the month's note

Standard breakdowns

Site and service line (warehousing and transportation)

Source

ERP (accounting)

Frequency

Monthly, published by the 8th business day, after the accounting close

Polarity

Higher is better

Target and thresholds

2026 budget target; green at or above target, yellow up to 1 point below, red below that

Owner of the result

Finance and administration director

Calculated by

Financial control analyst

Decision it supports

Review prices and contracts with below-average margins

Related KPIs

Logistics cost per order, Budget vs. actual expenses

Version

1.0, in effect since January 2026

FIN-02 Logistics cost per order(Warehousing + freight) ÷ orders shipped

Code

FIN-02

Question it answers

How much does it cost, on average, to store and deliver an order?

Linked objective

Finance: grow with healthy margins

Formula

(Warehousing + freight) ÷ orders shipped

Numerator

Warehousing cost (staff, rent, energy, equipment) plus freight paid in the month

Denominator

Orders shipped in the month

Excluded from both

Orders canceled before picking and freight reimbursed by the customer

Standard breakdowns

Site, customer and order size band

Source

ERP and WMS

Frequency

Monthly

Polarity

Lower is better

Target and thresholds

Up to the budgeted cost per order; green up to budget, yellow up to 5% above, red above that

Owner of the result

Financial controller

Calculated by

Financial control analyst

Decision it supports

Where to renegotiate freight and where to review warehouse productivity

Related KPIs

EBITDA margin, Picking productivity, OTIF

Version

1.0, in effect since January 2026

FIN-03 Budget vs. actual expensesActual expenses ÷ budgeted expenses × 100

Code

FIN-03

Question it answers

Are expenses moving at the pace of the budget?

Linked objective

Finance: grow with healthy margins

Formula

Actual expenses ÷ budgeted expenses × 100

Numerator

Actual expenses for the month, on an accrual basis

Denominator

Budgeted expenses for the month

Excluded from both

Capital expenditure (capex), which is tracked separately

Standard breakdowns

Department, cost center and type of expense

Source

ERP

Frequency

Monthly

Polarity

Close to 100%

Target and thresholds

Between 97% and 103%; yellow up to 5 points outside the range, red beyond that

Owner of the result

Financial controller

Calculated by

Financial control analyst

Decision it supports

Which departments need to explain the variance and revise the full-year forecast

Related KPIs

EBITDA margin

Version

1.0, in effect since January 2026

FIN-04 Overdue receivablesMore than 30 days overdue ÷ total receivables × 100

Code

FIN-04

Question it answers

How much of what we are owed is more than 30 days overdue?

Linked objective

Finance: grow with healthy margins

Formula

More than 30 days overdue ÷ total receivables × 100

Numerator

Invoices more than 30 days overdue on the last day of the month

Denominator

Total receivables on the last day of the month

Excluded from both

Invoices in litigation, tracked separately

Standard breakdowns

Customer and days-overdue band (31 to 60, 61 to 90, more than 90)

Source

ERP (accounts receivable)

Frequency

Monthly

Polarity

Lower is better

Target and thresholds

Up to 2%; yellow up to 3%, red above

Owner of the result

Financial controller

Calculated by

Financial analyst

Decision it supports

When to put a customer on hold and renegotiate terms

Related KPIs

EBITDA margin

Version

1.0, in effect since January 2026

Customers: deliver on time

CLI-01 OTIFOrders delivered on time and in full ÷ orders shipped × 100

Code

CLI-01

Question it answers

How many orders reached the customer on time and in full?

Linked objective

Customers: deliver on time

Formula

Orders delivered on time and in full ÷ orders shipped × 100

Numerator

Orders delivered by the agreed date with no missing items

Denominator

Orders shipped in the month

Excluded from both

Orders canceled by the customer (since version 1.1)

Standard breakdowns

Site, customer, carrier and reason for the miss (timing or completeness)

Source

WMS and TMS

Frequency

Monthly

Polarity

Higher is better

Target and thresholds

95%; green from 95%, yellow from 92%, red below

Owner of the result

Operations manager

Calculated by

BI analyst

Decision it supports

Which customers and carriers need an action plan

Related KPIs

Complaints per thousand orders, Picking errors

Version

1.1, in effect since February 2026

CLI-02 Customer NPS% promoters minus % detractors

Code

CLI-02

Question it answers

Would customers recommend Vallora?

Linked objective

Customers: deliver on time

Formula

% promoters minus % detractors

Numerator

Promoters (scores of 9 and 10) minus detractors (scores of 0 to 6), in percentage points

Denominator

Valid responses to the quarter's survey

Excluded from both

Responses without a score and customers with less than three months under contract

Standard breakdowns

Customer segment and the site that serves it

Source

Customer survey

Frequency

Quarterly

Polarity

Higher is better

Target and thresholds

50 points; yellow from 40, red below

Owner of the result

Sales manager

Calculated by

Sales analyst

Decision it supports

Which accounts get a visit and an improvement plan

Related KPIs

OTIF, Complaints per thousand orders

Version

1.0, in effect since January 2026

CLI-03 Complaints per thousand ordersValid complaints ÷ orders shipped × 1,000

Code

CLI-03

Question it answers

How often does an order lead to a valid complaint?

Linked objective

Customers: deliver on time

Formula

Valid complaints ÷ orders shipped × 1,000

Numerator

Complaints logged in the CRM and judged valid in the month

Denominator

Orders shipped in the month

Excluded from both

Complaints judged invalid and requests for information

Standard breakdowns

Reason for the complaint, customer and site

Source

CRM and WMS

Frequency

Monthly

Polarity

Lower is better

Target and thresholds

Up to 3 per thousand orders; yellow up to 5, red above

Owner of the result

Sales manager

Calculated by

BI analyst

Decision it supports

Which reasons to tackle first, together with Operations

Related KPIs

OTIF, Picking errors, Customer NPS

Version

1.0, in effect since January 2026

Operations: accuracy and efficiency

OPE-01 Picking productivityLines picked ÷ hours worked in picking

Code

OPE-01

Question it answers

How many order lines does picking deliver per hour worked?

Linked objective

Operations: accuracy and efficiency

Formula

Lines picked ÷ hours worked in picking

Numerator

Order lines picked and checked in the month

Denominator

Hours actually worked by the picking team, including overtime

Excluded from both

Training hours and inventory-count hours

Standard breakdowns

Site, shift and order type (each picks or full cases)

Source

WMS and time-and-attendance

Frequency

Monthly

Polarity

Higher is better

Target and thresholds

Each site's standard, reviewed once a year; yellow up to 10% below, red below that

Owner of the result

Operations manager

Calculated by

BI analyst

Decision it supports

Size the schedule and prioritize layout improvements

Related KPIs

Picking errors, Logistics cost per order, Absenteeism rate

Version

1.0, in effect since January 2026

OPE-02 Inventory accuracyPositions with no discrepancy ÷ positions counted × 100

Code

OPE-02

Question it answers

Does the stock in the system match what is on the shelf?

Linked objective

Operations: accuracy and efficiency

Formula

Positions with no discrepancy ÷ positions counted × 100

Numerator

Positions counted in cycle counting with no item or quantity discrepancy

Denominator

Positions counted in the month

Excluded from both

Positions blocked for receiving or shipping on the day of the count

Standard breakdowns

Site, warehouse zone and ABC class

Source

WMS (cycle counting)

Frequency

Monthly

Polarity

Higher is better

Target and thresholds

99.5%; yellow from 99%, red below

Owner of the result

Operations manager

Calculated by

Inventory analyst

Decision it supports

Where to recount and where to review the slotting process

Related KPIs

Picking errors, OTIF

Version

1.0, in effect since January 2026

OPE-03 Picking errorsLines with errors ÷ lines picked, in parts per million

Code

OPE-03

Question it answers

How many lines go out with the wrong item or quantity?

Linked objective

Operations: accuracy and efficiency

Formula

Lines with errors ÷ lines picked, in parts per million

Numerator

Lines with errors caught at checking or reported by the customer

Denominator

Lines picked in the month

Excluded from both

Discrepancies caused by the customer's product data

Standard breakdowns

Site, shift and where the error was caught (checking or customer)

Source

WMS and complaints

Frequency

Monthly

Polarity

Lower is better

Target and thresholds

Up to 500 parts per million; yellow up to 800, red above

Owner of the result

Operations manager

Calculated by

BI analyst

Decision it supports

Where to reinforce training and checking

Related KPIs

Picking productivity, Complaints per thousand orders

Version

1.0, in effect since January 2026

OPE-04 Warehouse occupancyPallet positions occupied ÷ pallet positions available × 100

Code

OPE-04

Question it answers

How much of the storage capacity is in use?

Linked objective

Operations: accuracy and efficiency

Formula

Pallet positions occupied ÷ pallet positions available × 100

Numerator

Pallet positions occupied on the last day of the month

Denominator

Pallet positions available, excluding those blocked for maintenance

Excluded from both

Floor-stacking areas, measured in m² and tracked separately

Standard breakdowns

Site and position type (pallet racking or picking)

Source

WMS

Frequency

Monthly

Polarity

Within a range

Target and thresholds

Between 80% and 88%; yellow between 75% and 80% or between 88% and 92%, red outside that

Owner of the result

Operations manager

Calculated by

BI analyst

Decision it supports

When to look for extra space or take on a new customer

Related KPIs

Picking productivity, Logistics cost per order

Version

1.0, in effect since January 2026

People: attendance and safety

PES-01 Absenteeism rateAbsence hours ÷ scheduled hours × 100

Code

PES-01

Question it answers

How much of the contracted time did the company lose to absences?

Linked objective

People: attendance and safety

Formula

Absence hours ÷ scheduled hours × 100

Numerator

Absence hours of every type: unexcused absences, doctor's notes, long-term absences, late arrivals and early departures, based on the day's scheduled hours

Denominator

Scheduled hours of everyone who was active that day

Excluded from both

Vacation, days off, holidays, days before hiring and after termination

Standard breakdowns

Short-term (up to 15 days) and long-term; absence type; site, shift, department and tenure (up to 90 days)

Source

Time-and-attendance system (schedules and clock-ins) and payroll (long-term absences)

Frequency

Monthly, published by the 5th business day

Polarity

Lower is better

Target and thresholds

3.5% target for 2026; green up to 3.5%, yellow up to 4.0%, red above

Owner of the result

HR manager; each DC manager answers for their site

Calculated by

BI analyst

Decision it supports

Where to investigate first and when to open an action plan for a team

Related KPIs

Schedule absence rate, Cost of paid hours not worked, Overtime

Version

1.1, in effect since March 2026 (see the version log)

PES-02 90-day turnoverNew hires who left before 90 days ÷ new hires in the month × 100

Code

PES-02

Question it answers

How many new hires leave before completing 90 days?

Linked objective

People: attendance and safety

Formula

New hires who left before 90 days ÷ new hires in the month × 100

Numerator

People hired in the reference month who left before completing 90 days

Denominator

People hired in the reference month

Excluded from both

Temporary contracts and apprentices, tracked separately

Standard breakdowns

Site, role, shift and reason for leaving (resignation or dismissal)

Source

Payroll

Frequency

Monthly, with a 3-month lag

Polarity

Lower is better

Target and thresholds

Up to 15%; yellow up to 20%, red above

Owner of the result

HR manager

Calculated by

HR analyst

Decision it supports

Review recruiting and onboarding at the sites and roles most affected

Related KPIs

Absenteeism rate, Accident frequency rate

Version

1.0, in effect since January 2026

PES-03 Accident frequency rateLost-time accidents × 1,000,000 ÷ hours worked

Code

PES-03

Question it answers

How often does someone get hurt and need time off?

Linked objective

People: attendance and safety

Formula

Lost-time accidents × 1,000,000 ÷ hours worked

Numerator

Work accidents with at least one day of absence, reported on the official accident form (CAT)

Denominator

Hours actually worked in the month, including overtime

Excluded from both

Commuting accidents, tracked separately

Standard breakdowns

Site, shift, department and tenure

Source

OHS and time-and-attendance

Frequency

Monthly

Polarity

Lower is better

Target and thresholds

20% reduction from the 2025 average; yellow up to the 2025 average, red above

Owner of the result

OHS coordinator

Calculated by

Safety technician

Decision it supports

Where to focus safety inspections and training

Related KPIs

Absenteeism rate, 90-day turnover

Version

1.0, in effect since January 2026

Two details that show the method: Warehouse occupancy has no “higher is better”, because a warehouse that is too full jams the operation; and 90-day turnover only closes three months after hiring, so the sheet says so, and nobody asks for the number before it exists.

Governance: who decides what each number means

No definition changes without a request, a test and an approval on record. The decision belongs to a KPI committee: the financial controller chairs it, the owners of the 14 KPIs take part, and the BI analyst keeps the minutes. It meets once a month, for 30 minutes, only when there are requests in the queue.

Roles (RACI matrix)

R does the work, A approves and answers for it, C is consulted, I is informed.

Activity	KPI owner	BI analyst	Committee	Leadership
Propose a new KPI or a change of definition	R	C	A	I
Measure the impact of the change	C	R	A
Approve or reject the change	C	C	R and A	I
Update the definition sheet and the version log	I	R	A
Calculate and publish every month	C	R and A		I
Explain the result and the action plan	R and A	C		I
Review the whole set once a year	C	R	A	C
Discontinue a KPI	C	R	A	I

Propose a new KPI or a change of definition

Owner R · BI C · Committee A · Leadership I

Measure the impact of the change

Owner C · BI R · Committee A

Approve or reject the change

Owner C · BI C · Committee R and A · Leadership I

Update the definition sheet and the version log

Owner I · BI R · Committee A

Calculate and publish every month

Owner C · BI R and A · Leadership I

Explain the result and the action plan

Owner R and A · BI C · Leadership I

Review the whole set once a year

Owner C · BI R · Committee A · Leadership C

Discontinue a KPI

Owner C · BI R · Committee A · Leadership I

How a definition changes

1. The owner opens a request: what changes, why, and from when.
2. BI recalculates the last 12 months under the new rule and shows the difference.
3. The committee approves, rejects or asks for changes at its next meeting.
4. Once approved, BI updates the definition sheet, bumps the version and logs the change.
5. The history is recalculated; when that is not possible, the break is marked on the chart.

Version log (example)

Date	KPI	Version	Change	Reason
March 2026	PES-01 Absenteeism rate	1.1	New breakdown by tenure (up to 90 days)	The Radar showed that new hires are absent more often
February 2026	CLI-01 OTIF	1.1	Orders canceled by the customer leave the denominator	They counted as an Operations failure
January 2026	All 14 leadership KPIs	1.0	First version of the definition sheets	Governance project

March 2026 · version 1.1

PES-01 Absenteeism rate

New breakdown by tenure (up to 90 days)

Reason: The Radar showed that new hires are absent more often

February 2026 · version 1.1

CLI-01 OTIF

Orders canceled by the customer leave the denominator

Reason: They counted as an Operations failure

January 2026 · version 1.0

All 14 leadership KPIs

First version of the definition sheets

Reason: Governance project

Results

The monthly meeting traded six presentations and 40 numbers for a single page with 14 KPIs, each with its owner in the room to explain it. When Operations talks about its rate, it now goes by another name, and the July argument does not happen again. The Absenteeism Radar now uses the official PES-01 definition.

What this case demonstrates

• Separating the math from the governance. A divergence is rarely a calculation error; it is almost always the lack of a written definition and of someone who decides.
• Cutting without deleting. Twenty-six KPIs left the leadership meeting, and no department lost information it needs.
• Definition sheets that answer the questions that start arguments: what goes in, what stays out, whether higher is better, when the number closes.
• Changes on record, so that today’s number stays comparable with last year’s.
• A bridge to BI. The definition sheets become the DAX measures and the dashboard’s data dictionary, and the dashboard stops hiding rules.

Does your company argue about which number is right before discussing what to do about it?

Tell me how your KPIs are calculated today and I will tell you where to start.

Get in touch