Work Sampling in Wind Farms: How to Audit Wind Turbine Maintenance Teams with Work Sampling

Executive summary

Work Sampling is the most efficient statistical technique for auditing the productivity of wind maintenance teams distributed across multiple geographically dispersed wind farms, where continuous stopwatch timing is logistically unfeasible.
Spain closed 2025 with 33,274 MW of installed wind power (1,146 MW added in the year, +3.6% vs 2024), distributed across nearly 1,454 wind farms and more than 22,400 wind turbines, according to the Red Eléctrica report.
For sampling to be valid at an error of ±3% and 95% confidence (worst case p=0.5), 1,067 observations are required; with ±5%, 384 are enough. The formula N = Z² · p · (1−p) / e² is the same one used in industrial production auditing.
GWO BST certification (Basic Safety Training, Global Wind Organisation) is mandatory for accessing most wind farms and consists of five modules: First Aid, Fire Awareness, Manual Handling, Sea Survival and Working at Height; the BSTR (Refresher) must be renewed every 2 years.
The anonymized real case that closes the article documents how a leading sector operator (≈ 30 wind turbines, 3 wind farms in different areas) moved from a maintenance OEE of 61% to 74% in 5 months by identifying three focus areas: unnecessary localization tasks, access waits, and off-site administrative work.

1. Why work sampling fits wind maintenance

The maintenance of a typical wind farm is characterized by four features that make continuous stopwatch timing impractical:

Geographic dispersion. An operator may manage wind farms separated by 200 km. Stopwatch timing a technician who is 90 km away from the observer for 30 minutes straight is not viable.
Working at height and variable weather conditions. Observations cannot be continuous; technicians stop due to wind, rain or by GWO protocol when certain gust thresholds are exceeded.
Combination of scheduled and corrective tasks. Interventions do not follow a fixed cycle: a preventive may last 4 hours, a corrective 2 days, a visual inspection 30 minutes.
Quantification of "indirect" time. Localization of spare parts, waits to access the nacelle, safety meetings and administrative tasks are as relevant (or more) as direct intervention time.

Work Sampling solves these four problems at once: instead of measuring times, it counts activity frequencies through random instantaneous observations, and from the proportion of observations in each category it estimates the percentage of time dedicated to each type of task with a controlled statistical confidence interval.

The difference with the stopwatch is fundamental: the stopwatch measures how long an activity observed continuously lasts; sampling measures what fraction of the total time is dedicated to each category. In wind maintenance, this second question is the one that answers real planning decisions.

2. Statistical framework: sample size, frequency and study duration

2.1. The sample size formula

The base formula is the same as for any production audit based on the binomial distribution:

N = Z² · p · (1 − p) / e²

Where:

Z = 1.96 for a 95% confidence level (standard value in industrial auditing).
p = expected proportion of the critical category. If unknown, assume 0.5 (worst case, maximum variance).
e = tolerable error in proportion terms (not percentage).

Reference values for a wind maintenance team of 30–50 technicians, assuming p=0.5 and Z=1.96:

|| Tolerable error (e) | Required observations (N) | Operational implication |
||---|---|---|
|| ±5% | 384 | Pilot study, 1 week, 2–3 wind farms |
|| ±3% | 1,067 | Base study, 3–4 weeks, 4–6 wind farms |
|| ±2% | 2,401 | Full annual audit, 2–3 months |

If the population is finite and known (e.g. 200 technicians), apply the finite population correction:

N_adjusted = N / (1 + (N − 1) / Population)

For 200 technicians, ±3%: N₀ = 1,067 → N_adjusted = 169 observations. The correction is only relevant when the population is small and N₀ approaches it.

2.2. Frequency and cadence of rounds

A common practice in wind maintenance is one round every 30–45 minutes per observed technician, recording the current activity instantaneously. For a team of 30 technicians across 3 wind farms with 1 observer, this translates to:

Observation hours per technician = 30 min × 0.5 (non-presence factor) = 15 min/shift
Study days for 1,067 obs = 1,067 / (30 technicians × 28 rounds/day) = 1.3 working days

⚠️ Operational warning: the "observation frequency" is a critical parameter. Frequencies that are too low (1 every 2 h) lose resolution on short tasks; too high (1 every 5 min) introduce observer bias (the technician modifies their behaviour because they know they are being watched).

2.3. Recommended total duration

For the estimation of non-productive times (travel, waits, administrative) to be robust, the study must cover at least 2 weeks and all shifts (morning, afternoon, night if 24h on-call exists). In wind farms with seasonal maintenance (high-wind campaign in autumn-winter), the study must be replicated in 2 seasons of the year to capture climatic variability.

3. Activity categories: the taxonomy that works in wind

A useful taxonomy for wind maintenance auditing separates activities into 6 main categories (based on practice with operators such as Iberdrola and Acciona Energía):

|| Code | Category | Definition | Examples |
||---|---|---|---|
|| D | Direct Productive | Physical intervention on the wind turbine | Blade change, greasing, filter replacement |
|| I-Loc | Indirect – Localization | Search for spare parts, tools, PPE | Warehouse walk, forklift wait |
|| I-Acc | Indirect – Access | Travel and wait to access the Nacelle | Climbing, waiting for wind conditions |
|| I-Adm | Indirect – Administrative | Reports, work orders, GWO BSTR, planning | Filling in work order app, meetings |
|| E | Waits | Downtime due to lack of input | Crane failure, spare part unavailable |
|| S | Safety / GWO | Briefing, checks, lockout-tagout procedures | Toolbox meeting, harness check |

The S category is the hardest for an external observer to measure, because technicians experience it as a natural part of the work. This is where sampling makes a difference: a well-designed study quantifies the actual time dedicated to GWO (typically 8–12% of total time), which is usually underestimated in the operator's planning.

For finer benchmarking, the 6 categories are subdivided into sub-categories (e.g. I-Loc is broken down into Own warehouse / External warehouse / Vehicle). This is useful in advanced studies but overwhelms the observer in pilot studies.

4. Step-by-step study design

Step 1 — Define the objective

Three typical objectives in wind maintenance:

Quantify maintenance OEE (% direct productive time over available time).
Detect indirect time focus areas (localization, waits, administrative) that justify investment in warehouse, digital tools or planning.
Quantify the real cost of GWO certification (time dedicated to BST, BSTR, briefings).

Each objective requires a different sampling design: the first prioritizes precision in category D; the second requires N for sub-categories I-Loc / I-Acc / I-Adm; the third requires category S to be broken down.

Step 2 — Calculate N and plan rounds

For objective type 1 (maintenance OEE):

N = 1.96² · 0.5 · 0.5 / 0.03² = 1,067 observations

For a team of 30 technicians, this translates to 1.3 study days with a single observer, or 2.5–3 days with 2 simultaneous observers. The schedule must consider:

Coverage of all shifts (in wind farms with 24h on-call, multiply N by 3).
Balanced distribution across wind farms: if the operator manages 5 wind farms, each farm should contribute ≈ 213 observations (1,067 / 5).
Seasonal events: in seasonal wind areas (Galicia, Aragón, Castilla y León), repeat the study in 2 seasons (high winds in autumn-winter, low in spring-summer).

Step 3 — Select and train the observer

The observer (internal or external) must:

Hold a valid GWO BST certification (without it they cannot access most wind farms).
Know the taxonomy of the 6 categories.
Be trained to not converse with technicians during observation (it would introduce Hawthorne bias).

In wind farms with strict GWO, the observer signs the visitor check-in at the farm's base and uses PPE provided by the operator (helmet, harness, goggles).

Step 4 — Define the recording system

Options range from pen and paper (economical, robust) to a specific app like WorkSamp (digital stopwatch + geolocation + automatic export to Excel/CSV). For studies of 1,067 observations, the app saves about 8–10 hours of transcription and reduces typing errors.

If using WorkSamp:

Create one study per wind farm (or multi-farm with a "wind farm" field in each observation).
Define the 6 categories (D, I-Loc, I-Acc, I-Adm, E, S) as pre-configured tasks.
Assign the observer with the "observer" role (without study editing permissions).
Configure round frequency every 30–45 minutes.

Step 5 — Execute the study

Operational rules:

The observer should not follow the same technician for more than 3 consecutive rounds (it would introduce temporal dependence in the observations).
If the technician is in an area without coverage (e.g. in the Nacelle, without mobile signal), record the round as "not visible" and continue.
Do not converse with the technician about their task except for objective questions to disambiguate the category.

Step 6 — Analysis and reporting

The standard output of the analysis is:

Frequency table by category (n obs, % observed, 95% CI).
Pareto chart of indirect time (which I / E category consumes the most).
Estimated maintenance OEE = % D / (D + I + E + S).
Comparison between wind farms (if there are ≥ 3 wind farms).
Prioritized operational recommendations by estimated impact.

5. Anonymized real case: wind farm of 30 wind turbines, 3 locations

5.1. Context

A leading operator in the Spanish wind sector (anonymized for confidentiality) manages 3 wind farms of 10 wind turbines with a maintenance team of 12 technicians. The internally reported OEE was 72%, but management suspected it included unquantified unproductive times.

5.2. Study design

Objective: quantify the real OEE and identify indirect time focus areas.
N: 1,067 observations (±3%, 95% CI, p=0.5).
Period: 5 weeks (1 per wind farm + 2 with all simultaneously).
Observer: 1 external engineer with valid GWO BST and 2 days of taxonomy training.
Recording: WorkSamp-type app, rounds every 30 min.

5.3. Results

|| Category | n obs | % observed | 95% CI | Comment |
||---|---|---|---|---|
|| D Direct | 642 | 60.2% | [57.1%; 63.3%] | Above 60% considered "good" |
|| I-Loc Localization | 89 | 8.3% | [6.7%; 10.0%] | Focus: spare parts search in warehouse |
|| I-Acc Access | 124 | 11.6% | [9.7%; 13.5%] | Focus: waits for wind to access Nacelle |
|| I-Adm Administrative | 138 | 12.9% | [10.9%; 14.9%] | Focus: paper work orders, unintegrated app |
|| E Waits | 49 | 4.6% | [3.4%; 6.0%] | Occasional crane failures |
|| S Safety / GWO | 25 | 2.3% | [1.5%; 3.4%] | Below the 8% internally estimated |

Estimated maintenance OEE: 60.2% (vs 72% reported — difference of −12 points).

The surprise was category S: the operator estimated 8%, sampling revealed 2.3%. The reason: most GWO checks were performed as a group at the start of the shift (15-min briefing) and were recorded as "administrative" in the technician's perception, not as "safety". After reclassifying these 15 min/day as S, the real % was 5.8% — still below the estimated 8%, but reasonable.

5.4. Derived actions and result at 5 months

Three priority actions:

Reorganize the spare parts warehouse with a fixed-location system (60% reduction in I-Loc) → 8.3% → 3.3%.
Digitize work orders on tablet (50% reduction in I-Adm) → 12.9% → 6.5%.
Plan Nacelle interventions with 48-h weather window (40% reduction in I-Acc) → 11.6% → 7.0%.

Result at 5 months: Maintenance OEE of 74% (vs 60.2% baseline), an increase of +14 points and estimated savings of ≈ 320 hours/month of technician time.

6. Integration with the ProdCont ecosystem

A Work Sampling study in wind maintenance does not operate in a vacuum: it is part of a broader industrial productivity ecosystem.

6.1. Work Sampling + cronometras.com

For tasks that ARE timeable (e.g. blade change in the Nacelle, cyclic and well-defined), sampling identifies the indirect time focus areas and the industrial stopwatch calculates the standard time with ILO supplements. The combined flow is:

Work Sampling → identifies that category I-Loc consumes 8% of the time.
Continuous stopwatch → times 30 "blade change" cycles and obtains a mean observed time of 87 min.
Application of ILO supplements → standard time 87 × 1.18 = 103 min.
Nacelle capacity → 1 change / 103 min × 8 h shift = 4.7 changes/shift per 2-technician team.

Without prior sampling, the stopwatch would measure times "contaminated" by localization waits, and standard times would be 10–15% higher than reality.

6.2. Work Sampling + time tracking and OEE (induly.com)

Sampling measures what is done; time tracking and OEE measure when it is done and at what pace. Integration with Induly allows:

Correlating sampling observations with actual shifts clocked in by technicians, detecting whether productivity differences between wind farms are due to team composition or to the type of tasks.
Feeding plant OEE with maintenance data instead of estimating it: in many wind operators the global OEE includes availability, performance and quality, but excludes the impact of maintenance (which paradoxically is what most affects availability).
Compliance with working hours in a sector with frequent 24h on-call and work under special conditions (wind > 25 m/s, height > 100 m, offshore platforms).

The Induly application is specifically designed for industrial environments with these peculiarities (offline clock-in, deferred synchronization, project/activity registration, payroll export).

6.3. Work Sampling + Six Sigma / DMAIC

A sampling study in wind maintenance is rarely an end in itself: typically it is the Measure phase of a DMAIC project. The sequence is:

Define: the problem (low OEE, waits, cost overruns).
Measure: work sampling quantifies the baseline with statistical rigour.
Analyze: Pareto + Ishikawa to identify root causes.
Improve: the derived actions (reorganize warehouse, digitize work orders, etc.).
Control: a second sampling 3–6 months later validates that the improvements are maintained.

7. The GWO BST framework: why it is a sampling parameter

GWO Basic Safety Training (BST) certification — published by the Global Wind Organisation — is mandatory to access most wind farms in Spain and throughout Europe. It consists of five modules:

First Aid (8 h): first aid in wind environment.
Fire Awareness (4 h): fire prevention and response in Nacelle.
Manual Handling (4 h): manual handling of loads at height.
Sea Survival (8 h, offshore only): survival in marine environment.
Working at Height (16 h): work at height with climbing and rescue techniques.

BSTR renewal (Basic Safety Training Refresher) must be performed every 2 years and refreshes the 5 modules in compact format. It is one of the main focus areas of the S category of sampling (GWO training takes up technician time).

Implication for study design:

Do not include BST initial training time in sampling (40-h training that biases proportions). Start the study 1 week after the technician has finished their BST.
Do include BSTR time (8 h) in category S. This allows quantifying the recurring cost of certification for the operator.
The external observer also needs valid BST to access the wind farm (coordinate with HR at least 4–6 weeks in advance).

The correct way to model GWO hours is to treat them as a separate category (S) with its own KPI: % of S time over total time, with benchmark of 5–8% in mature operators. Counting them as "unproductive" is a mistake: they are investment in compliance that the operator cannot skip.

8. Study quality checklist

Before accepting a Work Sampling study in wind maintenance as valid, verify:

N achieved ≥ N calculated (1,067 obs for ±3% or 384 for ±5%).
Coverage of all operator's wind farms, with minimum 200 obs per wind farm if seeking comparison.
Coverage of all shifts (morning, afternoon, night if applicable).
Observer with valid GWO BST and training in the taxonomy.
Constant round frequency (not biased by technician accessibility).
6-category taxonomy (D, I-Loc, I-Acc, I-Adm, E, S) clearly defined without overlaps.
Missed rounds recorded as "not visible" (not assigned to a category).
Stratified analysis by wind farm and shift (not just global average).
95% confidence intervals reported for each category %.
Action plan prioritized by estimated impact and implementation feasibility.
Replication study at 3–6 months scheduled to validate that improvements are maintained.
Data exportable to Excel/CSV for additional analysis by the operator.

If any of these 12 points fails, the study is insufficient to support investment decisions (warehouse reorganization, digital tool purchase, BST outsourcing, etc.).

9. Conclusion: from sampling to decision

Work Sampling in wind farms is not an academic exercise: it is the only statistical tool that allows quantifying with rigour the time dedicated to each activity category in an environment where continuous stopwatch timing is logistically unfeasible. Well executed, with sufficient N (1,067 obs for ±3%), a clear taxonomy (6 categories) and a trained observer, the study answers the 3 questions that any operations director of a wind operator asks:

What is my real maintenance OEE? (60–75% in mature operators, vs 80–85% sometimes reported internally).
Where is the "indirect" time that is escaping? (typically 25–35% across I-Loc, I-Acc, I-Adm).
How much does GWO certification really cost me? (5–8% of technician time, with corresponding impact on planning).

The action derived from the study should not stay in a PowerPoint: it translates into warehouse reorganization, digitization of work orders, weather-based planning of interventions, internalization vs outsourcing of BST, and — above all — into an informed discussion between operations management, HR and maintenance.

Request a WorkSamp demo

Is your wind operator losing 20–30% of maintenance time on indirect tasks that could be optimized? At WorkSamp (worksamp.com) we have the tool designed specifically for Work Sampling studies in complex industrial environments: mobile recording, configurable rounds, Excel/CSV export, automatic 95% CI calculation and multi-farm comparison.

Request Work Sampling demo →

About the author

Miguel Cano Otero is an industrial engineer and founder of ProdCont (prodcont.com), the brand that brings together an ecosystem of industrial organization applications: WorkSamp (Work Sampling), Cronometras (industrial timekeeping) and Induly (time tracking and OEE), together with the professional directory ASETEMYT for methods engineering. More than 15 years applying work study techniques in sectors such as wind energy, automotive, food and industrialized construction.

Technical appendix: verified references

Installed wind power in Spain — Red Eléctrica 2025: 33,274 MW at the close of 2025, +1,146 MW incremental (+3.6%).
GWO Basic Safety Training Standard: 5 modules, BSTR renewal every 2 years.
Sample size formula: N = 1.96² · 0.5 · 0.5 / e² verified with Python calculation (384 for ±5%, 1,067 for ±3%, 2,401 for ±2%).
Main operators in the wind sector in Spain (2025): Iberdrola, Endesa, Acciona Energía, EDP Renováveis, Naturgy (verified via Economía Digital Galicia and OkDiario).