Union Negotiation Based on Objective Data

A time-distribution study can inform negotiation about organization, workloads or improvement. It does not by itself determine fair pay, a person's effort or required staffing. Technical information must be considered alongside safety, working conditions and representation rights.

The purpose is not to prove management or the works council “wrong.” A direct-work percentage and perceived workload can describe different things: support tasks may impose substantial physical or mental demands. No universal wrench-time threshold separates efficient and inefficient plants.

Agree the question before measuring

Specify the decision and the conclusions the study cannot support. “Identify material waits in maintenance during four weeks” is a verifiable scope. “Certify how hard the workforce works” is not an inference from state readings.

A shared protocol should include:

Population, resources, shifts and included/excluded periods.
Categories, observable examples and ambiguous-case resolution.
Selection of moments and resources, weights and analysis units.
Precision, independence assumptions and missing readings.
Information to affected people, access, retention and safeguards.
Reporting of uncertainty, disagreements and technical review.
Intended use and procedures for process changes.

Participation makes results easier to examine and understand; it does not make them incontestable. Errors, bias and undercoverage can still be identified after collection.

Categories that neither blame nor hide necessary work

A draft dictionary might separate direct execution, preparation, travel, waiting with a checked cause, training/meetings, included breaks and undetermined state. These are operational states, not legal findings about conduct.

An observer cannot label an instantaneous absence “unjustified” without the relevant information. Training, coordination, cleaning and prevention are not automatically unproductive. Breaks and safety measures are not automatic reduction targets.

Two observers can independently classify pilot situations and examine disagreement. A total of 100% does not prove all contexts are covered. A residual category identifies situations needing definition; hiding it only improves appearances.

Sample size and effort: a calculable example

Under an independent binomial approximation:

n=ceil(z²p₀(1−p₀)/e²).

A planning example with z=1.96, p₀=0.35 and e=0.03 gives n=ceil(971.0711)=972. The margin is in absolute percentage points. The calculation does not guarantee coverage when shifts are excluded, particular states are missing or readings are dependent.

As a logistical assumption, four valid readings per analyst-hour implies 972/4=243 analyst-hours. Two analysts each allocating 20 hours per week would require 243/40=6.075 weeks at that allocation. This is not an industry-standard duration; routes, readings per visit and shift coverage change the budget.

Distinguish sampling visits from participations within visits. Observing several people in one round does not automatically multiply independent information. Subgroup precision needs its own counts and design-appropriate analysis.

Interpret time shares without equating them with effort

Consider a simulated teaching case with 1000 independent readings:

State	Readings	Proportion
Direct execution	350	35%
Preparation	200	20%
Travel	150	15%
Verified material waiting	150	15%
Coordination and training	100	10%
Included breaks	50	5%
Total	1000	100%

For direct execution, the approximate normal 95% margin is 1.96√(0.35×0.65/1000)≈2.96 percentage points. The other 65% is not necessarily waste or recoverable time. Examine necessary support, changeable waiting causes and intervention effects.

A useful statement is “material waiting was estimated at 15% in the defined period, with uncertainty and causes to investigate.” It is not “the workforce wastes 15% of its shift.” Neither individual effort nor misconduct is established.

Move from observation to improvement proposals

When testing a supply change, preserve definitions, compare product mix and load, and measure safety, quality and production effects. Before/after differences do not independently establish causality. Include incidents and findings that fail to support the initial hypothesis.

Equipment OEE needs A=run time/planned production time, P=ideal cycle×total count/run time and Q=good count/total count. State distributions contextualize stops but do not alone calculate performance and quality. Suitable records may be manual; sampling and sensors need not be competing approaches.

Information, consultation and data protection in Spain

Article 64.5.f of the Workers' Statute provides for a prior works-council report on work organization/control systems, time studies and incentives;64.6 addresses information and consultation. Assess application with the collective agreement and actual measures. It is not sound to state generally that a council cannot challenge or prevent any study.

GDPR applies when data permit direct or indirect identification. Removing names is insufficient if workstation and time identify someone. Establish legal basis, purpose, minimization, information, access and retention; take care with small-group aggregates. The AEPD employment guide helps assess processing but does not automatically approve software.

Retention should be justified rather than “delete everything immediately after analysis” or retain indefinitely. Reuse for a different purpose requires a lawfulness assessment. Before linking findings to pay, employment-condition changes or individual decisions, review employment and data-protection requirements with the responsible specialists.

What an agreed annex could contain

A working draft can define a process metric, population, attached protocol, shared technical review, tolerable uncertainty, challenge mechanism and limits on use. Specify what happens when precision is not achieved, product mix changes or incidents arise. A 95% level and ±3 percentage points may be an agreed choice, not a mandatory standard for all negotiations.

Do not automatically link a bonus to one point estimate without considering perverse incentives. Relabeling states, omitting necessary work or reducing breaks to meet an indicator would undermine both measurement and work.

Tools and frequently asked questions

WorkSamp supports visit preparation and recording. It neither removes bias nor certifies agreements. Cronometras supports duration analysis; Induly records working time, not effort. The ASETEMYT directory lists tools and specialists.

Is work sampling legal? Assess the actual design and use, applicable information/consultation and personal-data processing. An anonymous report does not settle every obligation.

How long does a study take? Precision, resources, period coverage and logistics determine duration. Four or eight weeks are not compulsory.

Is a small study worthless? It may guide investigation or provide a less precise estimate. Do not claim precision it has not achieved.

Does it replace time study? No. State distributions and task durations answer complementary questions.