Union Negotiation Based on Objective Data

Why do union negotiations fail due to lack of evidence?

Collective bargaining in Spain has a deep-seated problem that few organisations dare diagnose honestly: the parties arrive at the table with opposing narratives and without a shared set of empirical data.

When management says "there is room for improvement" and the works council responds "we are at the limit", both statements can be equally subjective. And a negotiation built on subjectivity does not produce sustainable agreements.

The paradox of subjective narratives in Spain

In our experience supporting industrial plants, the pattern repeats itself with alarming frequency. Management estimates that direct productive time is around 70–78%. The works council maintains that staff work at full capacity and that any increase in targets is unworkable.

The reality, when measured with statistical rigour, usually lies at an intermediate point that neither party had anticipated.

This disconnect is not resolved with more meetings or better labour relations. It is resolved with objective data, measured with a methodology that both parties accept as valid from the design stage.

The real cost of 470,000 days lost in labour disputes

According to data from the Ministry of Labour and Social Economy, in 2024 more than 470,000 working days were lost due to labour disputes in Spain, an increase of 12% over the previous year.

Not all of those disputes originate in productivity. But a significant proportion do. When there is no objective basis for discussing workloads, production rates or staffing needs, conflict inevitably escalates.

Each day lost to strike or mediation has a direct cost for both parties. And an even greater indirect cost in the deterioration of the working climate.

The perception gap between company and union representation

In a longitudinal study of 14 Spanish industrial plants — metallurgical, agri-food and automotive sectors — the average Wrench Time (proportion of direct productive time over total working day) was 31.4% with a standard deviation of 7.2%.

However, at the negotiation tables associated with those plants, subjective estimates ranged from 55% (union position) to 78% (company position).

That gap of 23 to 47 percentage points between estimation and reality is the space where misunderstandings, unworkable proposals and avoidable conflicts are born.

What is Work Sampling and how does it transform the negotiation table?

Work Sampling is an industrial engineering technique that allows estimating the proportion of time dedicated to different categories of activity through random observations over a representative period.

It is not time study. It is not surveillance. It does not measure people.

It measures processes and activity distribution on a statistical basis that allows the results to be generalised to the entire observed population.

Historical origin: Tippett's technique and Snap Reading

The technique was developed by L.H.C. Tippett in the 1930s in the British textile industry. His idea was revolutionary at the time and remains valid today:

Instead of continuously timing each operation, it is enough to take instantaneous random observations (Snap Readings) and count the frequency with which each activity category appears.

Tippett demonstrated that, with a sufficient number of random observations, the observed proportion converges towards the real proportion with a calculable confidence level.

Almost a century later, this methodology remains the standard for productivity diagnostics in industrial environments. Digital tools such as Cronometras have greatly simplified the capture and analysis of this data, but the underlying statistical principle remains intact.

Principle of statistical inference applied to work

Work Sampling is based on the central limit theorem and the binomial-normal approximation. The procedure is as follows:

A taxonomy of work states (activity categories) is defined.
N random observations are made over a representative period.
The proportion of observations falling into each category is calculated.
Confidence intervals based on the binomial distribution are applied.

The result is not a single data point, but an interval with a specified confidence level. This is fundamental for credibility at a negotiation table.

Difference between measuring people and measuring processes

A critical distinction that must be clear from the start:

Measuring people implies evaluating individual performance, which generates legitimate resistance and raises ethical and legal issues.
Measuring processes implies analysing how time is distributed across activity categories at an aggregate level, without identifying individuals.

Work Sampling measures processes. Each observation captures the state of an activity at a given moment, not the performance of a person. This makes it inherently compatible with data protection and union acceptance.

Statistical foundations that give credibility to the data

The strength of Work Sampling in a negotiation is not in the consultant's opinion. It is in the mathematics. And mathematics has no side.

How to calculate the correct sample size

The fundamental Work Sampling formula is:

N = (Z² × p × (1-p)) / E²

Where:

N = number of observations required
Z = Z-value of the confidence level (1.96 for 95%)
p = estimated proportion of the phenomenon
E = desired margin of error

Practical example:

A plant wants to determine, with 95% confidence and ±3% margin of error, the percentage of direct productive time. The prior estimate is 35%.

N = (1.96² × 0.35 × 0.65) / 0.03²
N = (3.8416 × 0.2275) / 0.0009
N = 0.8739 / 0.0009
N ≈ 971 observations

At a rate of 4 observations per analyst hour, this is equivalent to about 243 analyst-hours, i.e. approximately 6 weeks with a team of 2 analysts dedicating 20 hours per week each.

This calculation must be shared with the works council before starting the study. That both parties understand and accept the sample size is a legitimacy requirement.

Confidence level, margin of error and binomial distribution

Each Work Sampling observation is a binomial trial: the observed activity either belongs or does not belong to a given category. With thousands of observations, the binomial distribution approximates the normal (Gauss curve), which allows:

Calculating confidence intervals for each proportion.
Determining whether differences between periods or shifts are statistically significant.
Rejecting or confirming hypotheses about productivity with objective rigour.

At a negotiation table, this translates into statements like: "With 95% confidence, direct productive time lies between 28% and 35%." That statement carries a weight that no subjective estimate can match.

The Gauss curve as a framework for interpreting results

The Gauss curve does not appear directly in Work Sampling, but it does in its interpretation. When we repeat the study multiple times or analyse data subsets, the distribution of results follows a normal bell.

This allows us to:

Identify outliers (shifts or days with anomalous behaviour).
Separate normal variability of the process from real deviations.
Communicate results intuitively: "68% of the days analysed fall within this range."

For operations directors and plant engineers, this framework is familiar and credible. For union representatives, it is verifiable.

Why Work Sampling avoids the Hawthorne effect

The Hawthorne effect describes how workers modify their behaviour when they know they are being continuously observed. It is the main enemy of any productivity study.

Work Sampling minimises it by design:

Observations are random and instantaneous (Snap Readings).
The observer does not stay next to the worker for prolonged periods.
After a few sessions, the observer's presence normalises.
Randomness prevents workers from anticipating when they will be observed.

Empirical studies show that the Hawthorne effect in Work Sampling is reduced to insignificant levels after the first 50–100 observations. This is a huge advantage over classic continuous time study.

MECE taxonomy for classifying activities in the plant

The heart of a Work Sampling study is its category taxonomy. If the categories are poorly defined, the data is useless. That is why the MECE principle is non-negotiable.

What Mutually Exclusive and Collectively Exhaustive means

MECE stands for Mutually Exclusive, Collectively Exhaustive. Applied to Work Sampling:

Mutually Exclusive: each observation can be classified into only one category. There must be no ambiguity about where an activity fits.
Collectively Exhaustive: the categories must cover 100% of possible activities. There cannot be a significant residual "other".

If an observer hesitates more than 2 seconds over which category to assign an observation to, the taxonomy needs revision.

Example of categorisation for Spanish industrial sectors

A typical MECE taxonomy for a manufacturing plant might include:

Direct productive work: machine operation, assembly, packaging.
Indirect productive work: material preparation, zone cleaning, preventive maintenance.
Waiting time: waiting for material, waiting for instructions, waiting for machine.
Movement: moving between workstations, picking up components.
Unproductive time: unjustified absence, personal use, unplanned inactivity.
Support time: training, meetings, communication with supervisor.

The key is that this taxonomy must be designed jointly with union representation. Not imposed.

How to involve the works council in the study design

Union participation in the design is not a courtesy gesture. It is a methodological legitimacy requirement. The concrete actions are:

Category definition workshop: session where company and council review and validate the MECE taxonomy.
Joint pilot test: both parties conduct trial observations to verify that the categories are clear.
Agreement on confidence level and margin of error: both parties accept the statistical parameters before collecting data.
Access to intermediate results: the council may request periodic updates during the study.

When the council participates in the design, the results are uncontestable because the party that could question them helped build them.

Wrench Time: the metric no one wants to measure but everyone needs

What direct productive time is and why it matters

Wrench Time — or direct productive time — is the proportion of the working day during which the worker performs the activity that constitutes their main function. In maintenance, it is time with the wrench on the machine. In production, it is time operating the process.

Everything else: movements, waiting, tool searches, paperwork, meetings, is time that does not generate direct value.

This metric matters because it is the denominator of every productivity improvement. If you do not know how much productive time you have, you cannot size real improvements.

The empirical data: 31.4% average across 14 Spanish plants

The data from the longitudinal study already mentioned deserves to be broken down:

Overall average: 31.4% Wrench Time.
Standard deviation: 7.2 percentage points.
Observed range: from 19% in plants with highly fragmented processes to 48% in automated lines with specialised operators.
Sectors: metallurgical, agri-food, automotive.

These data are not a Spanish anomaly. International studies place Wrench Time in industrial maintenance between 25% and 35% on global average.

The 23 to 47 percentage point gap between estimation and reality

The most revealing thing is not the absolute figure, but the perception gap:

Union estimate: 55% (subjective, unmeasured).
Company estimate: 78% (subjective, unmeasured).
Measured reality: 31.4% (Work Sampling with 95% confidence).

Both parties were wrong, but in opposite directions. The company overestimated available productivity. The union overestimated current effort.

The objective data allows a different conversation: not about who works more or less, but about what organisational barriers reduce productive time to 31%.

How to use Wrench Time without generating confrontation

The framing of the result is decisive. A Work Sampling study must not be presented as an accusation but as a shared diagnostic:

Correct language: "The study reveals that 68.6% of time is spent on support, waiting and movement activities. We need to understand together why this happens and how to improve it."
Incorrect language: "Workers are only productive 31% of the time."

The same data, with a different communicative framing, produces completely different outcomes at the negotiation table. Responsibility for this framing lies both with management and with the consultants presenting the results.

OEE without sensors: measurable productivity through random observation

What OEE is and how it is calculated by sampling

OEE (Overall Equipment Effectiveness) is the standard efficiency indicator in manufacturing. It is composed of three factors:

Availability: proportion of time the equipment is operational vs planned time.
Performance: actual production speed vs theoretical speed.
Quality: proportion of conforming parts vs total produced.

Traditionally, OEE requires sensors, PLCs and SCADA systems. But with Work Sampling, it is possible to estimate availability and performance components through random observation, without installing additional hardware.

The proportion of observations where the machine is operating, waiting, broken down or in changeover provides a direct estimate of availability and performance, with calculable confidence intervals.

Advantages over continuous monitoring with hardware

Sampling-based OEE offers specific advantages in union negotiation contexts:

No hardware investment required: no sensors to install, cables to lay or systems to integrate.
It is temporary: the study has a defined start and end. It is not permanent surveillance.
It is anonymous: data is aggregated by category, not by individual or workstation.
It is union-acceptable: it does not imply electronic worker monitoring.

Platforms like Induly offer real-time production control with hardware when the organisation is ready for it. But as an initial diagnostic and negotiation tool, sampling-based OEE has a cost-effectiveness ratio and social acceptance incomparably superior.

Practical case in an agri-food plant

An agri-food processing plant in Andalusia needed to justify to the works council the need to reorganise shifts to increase capacity without hiring additional staff.

A Work Sampling study was carried out over 4 weeks with 1,200 random observations. The results showed:

Measured availability: 72% (prior management estimate: 85%).
Main causes of non-availability: changeovers (14%), intermediate cleaning (8%), lack of upstream material (6%).
Measured performance: 81% of theoretical speed.

With these data, the company and council agreed on a plan to reduce changeover times through SMED and an improvement in supply scheduling. The result was an increase in OEE from 72% to 84% in 6 months, without additional hiring and with full union agreement.

Spanish regulatory framework 2025 for observational data collection

Data protection and anonymisation in Work Sampling

The collection of observational data must comply with the European GDPR and its Spanish transposition. The key principles applied to Work Sampling are:

Minimisation: only what is strictly necessary is collected (activity category, not worker identity).
Proportionality: the study has a legitimate and proportionate purpose (productivity diagnosis).
Aggregation: data is analysed at category and period level, not at individual level.
Temporality: raw data is deleted after analysis. Only aggregated results are retained.

New features of Organic Law 1/2025 in the labour context

The entry into force of Organic Law 1/2025 has reinforced guarantees in data processing in the labour context, with special emphasis on:

Reinforced transparency: obligation to inform in detail about the methodology, scope and purpose of any study.
Prior information right: workers must know about the existence of the study before observations begin.
Prohibition of disciplinary use: individual data cannot be used as a basis for sanctions.

Work Sampling, due to its statistical and aggregated nature, meets these requirements inherently. But formal communication to the works council and the workforce must take place before starting the observations.

Information and consultation rights of the works council

Labour transparency regulations grant workers' legal representation the right to:

Be informed about the existence, methodology and purpose of the study.
Consult on the design (categories, periods, observation zones).
Access the aggregated results.
Participate in the interpretation of the data.

These rights are not an obstacle. They are an opportunity to build shared credibility.

Why statistical sampling is compatible with current regulations

Unlike continuous tracking by sensors, GPS or monitoring software, Work Sampling:

Does not generate individual performance profiles.
Does not allow identification of workers in the results.
Is temporary (not permanent).
Has a shared diagnostic purpose, not disciplinary.

This compatibility is a competitive advantage of Work Sampling over other productivity measurement tools that can generate legal and union conflicts.

How to incorporate objective metrics in collective agreements

The methodological gap in Spanish sectoral agreements

Less than 15% of sectoral agreements in Spain include productivity clauses linked to verifiable objective metrics. Most use generic indicators (CPI, turnover, gross production) that do not reflect real operational efficiency.

This gap creates two problems:

Salary increases become decoupled from real productivity.
There is no accepted mechanism to measure and certify efficiency improvements.

Productivity clauses linked to verifiable data

Work Sampling allows building concrete clauses such as:

"Productivity will be measured through a Work Sampling study conducted with the methodology agreed in Annex X, with a confidence level of 95% and a margin of error not exceeding ±3%. The study will be designed and validated jointly by the company and the works council."

Example wording for a productivity annex

A typical annex might include:

Methodology: Work Sampling with Tippett's technique (Snap Reading).
Taxonomy: MECE categories defined in a joint workshop (Annex Y).
Statistical parameters: 95% confidence level, ±3% margin of error.
Frequency: annual study with minimum duration of 4 weeks.
Participation: the works council participates in the design, has access to intermediate results and co-signs the final report.
Linkage: the results are used as a basis for negotiating productivity supplements.

Step-by-step implementation in an industrial plant

Phase 1 — Diagnosis and scope definition

Identify the scope of the study: complete plant, line, functional area.
Define the purpose: productivity diagnosis, justification of organisational changes, basis for negotiation.
Select the representative shifts and periods.
Estimate the prior proportion of each category to calculate the sample size.

Phase 2 — Study design with union participation

Conduct a joint workshop to define the MECE taxonomy of categories.
Run a pilot test of 50–100 observations to validate the clarity of the categories.
Agree on the statistical parameters (confidence level, margin of error).
Formally communicate to the entire workforce the existence, methodology and purpose of the study.

Phase 3 — Execution of random observations

Generate a random sequence of routes and observation times.
Take Snap Readings recording only the observed activity category.
Maintain a steady rate of observations (typically 4–6 per hour).
Supervise the quality of capture and resolve classification questions in real time.

Phase 4 — Statistical analysis and presentation of results

Calculate the proportions of each category with their confidence intervals.
Verify that the achieved sample size is sufficient (recalculate if necessary).
Analyse relevant segmentations: by shift, by area, by day type.
Prepare an executive report with clear visualisations and accessible explanation of the statistical results.

Phase 5 — Use of the data in collective bargaining

Present the results in a joint session with the company and council.
Interpret the data as a shared diagnostic, not as an accusation.
Identify improvement opportunities with impact on productivity and working conditions.
Incorporate the results as an objective basis for productivity agreements in the collective agreement.

Common errors that invalidate a Work Sampling study

Non-random observations and selection bias

If the observer consciously chooses when and whom to observe, the results are biased. Randomness is not optional: it is the statistical foundation of the method.

Using random sequence generators or specific applications like WorkSamp eliminates this risk.

Taxonomies that are not MECE

If categories overlap or leave uncovered areas, observations cannot be classified consistently. This invalidates the count and the confidence intervals.

Insufficient sample size

A study with 200 observations when 1,000 are needed has no statistical validity. The sample size must be calculated before the study and verified during its execution.

Confusing observation with surveillance

If workers perceive the study as a covert surveillance system, the Hawthorne effect spikes and union acceptance is lost. Transparent communication and participation in the design are the best vaccine against this perception.

Frequently asked questions about union negotiation with objective data

Is Work Sampling legal in Spain?

Yes. It complies with the GDPR, the LOPDGDD and Organic Law 1/2025 provided that observations are anonymous, aggregated and previously communicated to the works council.

How long does a complete study take?

It depends on the calculated sample size. For a medium plant, between 4 and 8 weeks of active observations with a team of 2–3 analysts.

Can the works council reject a Work Sampling study?

It may request modifications to the design, but cannot prevent a productivity diagnostic study if the legal information and participation requirements are met. Collaboration is always preferable to imposition.

Does Work Sampling replace time study?

No. They are complementary tools. Work Sampling is ideal for activity distribution diagnostics. Time study is necessary to establish standard times for specific operations. Both methods remain the basis of modern productivity, and directories like ASETEMYT make it easy to find qualified professionals for both.

What confidence level is recommended for a negotiation?

The accepted standard is a 95% confidence level with a margin of error not exceeding ±3%. These parameters must be agreed with the council before the study.

Resources and Tools

WorkSamp: Specialists in Work Sampling. Complete methodology for productivity diagnosis with random observation.
Cronometras: Digital tool for time and motion analysis. Complements Work Sampling when detailed standard times are required.
Induly: Production Control and Industrial Time & Attendance software. For when the organisation needs continuous monitoring after the initial diagnostic.
ASETEMYT Directory: Find professionals and companies specialised in industrial time study, methods engineering and time studies.
ASETEMYT Blog: Technical articles on productivity, industrial engineering and operations management.
Add your company: If you offer time study or methods engineering services, become part of the directory.