Work Sampling in Industrialized Construction: Productivity Audit in Concrete Precast Plants

Executive Summary

Industrialized construction with concrete precast in Spain grew by 8% in consumption in 2025 and 15% in number of works in the first four months of 2026, according to the Spanish Association of the Precast Concrete Industry (ANDECE), which groups 70% of the industrial business volume of the sector.

Precast plants operate as batch manufacturing with discontinuous flow (formwork preparation, rebar, concrete pouring, curing, stripping, storage, dispatch): a pattern that breaks the assumptions of classic work sampling and demands sampling stratified by phase and shift.

The VA / NVA / Unproductive taxonomy specific to precast includes categories that do not appear in automotive plants (for example, "controlled curing time", "stripping wait time", "internal transport with overhead crane"). Without a dedicated taxonomy, work sampling studies measure wrong and recommend worse.

The sample size calculation for a medium plant (≈ 38 active stations) requires ≈ 570 observations at ±4% precision, which translates into ≈ 7 working days of field work with a single observer — a cost far below the expected return (in the anonymized case we document, payback less than one month on an OEE improvement project from 62% to 76%).

The harmonized standards UNE-EN 13225, 13693, 13747, 13978-1, 14843 and 1168 regulate precast elements with CE Marking under assessment system 2+. Any work sampling study in these plants must respect the documentary traceability controls imposed by these standards.

1. Why Classic Work Sampling Does Not Apply "As Is" in a Precast Plant

Work sampling was born for repetitive continuous-flow manufacturing plants — automotive, electronics, food. A precast concrete plant has a radically different pattern:

Discontinuous flow by phases, not by line. A piece (beam, column, hollow-core slab, pre-slab, staircase, precast wall) passes through sequential stations: formwork, rebar, pouring, vibrating, curing, stripping, storage, loading. The time between stations is highly variable: curing can last between 8 and 24 hours depending on type and required strength, and during that time the piece neither generates nor consumes direct human resource.
High product variability. A medium plant can produce between 50 and 200 different references in one year (hollow-core slabs from 16 to 32 cm deep, beams of different spans, columns of different sections). The standard time per unit varies greatly between products.
Mobile and multi-station teams. A rebar fitter can work simultaneously on 3 different pieces on assembly tables, while another fitter is applying release agent. This breaks the classic assumption of "1 operator = 1 independent observation" of work sampling.
Curing as a natural state of the process. While a piece cures, the operator cannot accelerate anything (concrete strength depends on time, humidity, and temperature). Any observation made during curing will always classify the activity as "process wait" — valuable information, but which must NOT be penalized as unproductive.

Applying work sampling without adapting the method to this reality produces two systematic errors: underestimating VA work content (because the "natural" curing dead times are confused with operational unproductivity) and overestimating the necessary direct labor (because more people are requested for a phase that is actually limited by the next phase that has not yet released the previous piece).

2. VA / NVA / Unproductive Taxonomy Specific to Precast

Before designing the sampling plan, the taxonomy must be defined. We propose a MECE hierarchical structure (Mutually Exclusive, Collectively Exhaustive) validated with three real implementations:

2.1. VA — Value Added (time the customer pays)

Code	Category	Examples
VA-1	Formwork and mold preparation	Cleaning, release agent application, assembly
VA-2	Rebar and assembly	Cutting, bending, placement, tying
VA-3	Concrete pouring	Pouring, vibrating, spreading
VA-4	Manual finishing	Resins, blowhole repair, joint treatment
VA-5	Productive quality control	Abrams cone consistency measurement, specimen taking
VA-6	CE Marking and documentary traceability	Labeling, batch registration, archiving

2.2. NVA — Necessary but Not Added (support activities without direct value)

Code	Category	Examples
NVA-1	Internal movement with overhead crane	Transfer between stations
NVA-2	Mandatory technical waits	Curing, setting, stripping wait
NVA-3	Equipment setup	Vibrating table calibration, mold adjustment
NVA-4	Basic autonomous maintenance	Greasing, technical cleaning, small consumable replacement
NVA-5	PPE and safety	Harness donning and doffing, lifeline check

2.3. IMP — Unproductive (avoidable)

Code	Category	Examples
IMP-1	Movement without load	Walking looking for material, tool, or information
IMP-2	Waits due to lack of material	Rebar that does not arrive, delayed concrete, occupied formwork
IMP-3	Decision waits	Waiting for instructions from the foreman, validations, layouts
IMP-4	Equipment breakdowns and stops	Broken concrete pump, broken vibrator, stopped crane
IMP-5	Rework	Blowholes that are repaired, rejected pieces
IMP-6	Unplanned socializing and breaks	Conversations, non-urgent queries, off-calendar breaks

Field tip: in the observer training phase (1-2 days), the same taxonomy is given to two observers and their classifications are compared. If the coincidence is less than 85%, the taxonomy must be refined or retraining done. Below 75%, the study data are NOT statistically valid.

3. Sampling Study Design: Sample Size, Stratification, and Frequency

3.1. Sample Size Formula (N)

The sample size for an unknown proportion is calculated as:

N = 4 · p · (1 − p) / e²

where 1.96² ≈ 4 for a 95% confidence interval, p is the estimated proportion of the event (in our case, the fraction of time in a given category, usually value added) and e is the acceptable absolute error.

Python verification:

Scenario	p	e	N (observations)
Worst case (no prior information)	0.50	±5%	400
Worst case	0.50	±3%	1,111
Worst case	0.50	±2%	2,500
Estimated p 0.70	0.70	±3%	933
Real plant (anonymized)	0.35	±4%	569

Note: these N values are totals for the study, not per shift. If the plant has 3 shifts, the 569 observations are distributed proportionally to the size of each shift (≈ 190 per shift if balanced).

3.2. Stratification by Process Phase

The sampling is stratified by the major process phases, not by product line. Reasons:

The phases have very different duration (the curing phase lasts hours, concrete pouring minutes), so non-stratified sampling would automatically underrepresent the long phases.
The potential improvements are in different phases (in formwork: reuse; in rebar: ergonomics; in curing: stock management; in dispatch: transport planning).

Typical N distribution in a structural precast plant (anonymized data from 3 implementations):

Phase	% of Total N	Observations with N=569
Preparation / formwork	25%	142
Rebar and assembly	25%	142
Concrete pouring and vibrating	15%	85
Curing (only technical waits)	5%	28
Stripping and storage	15%	85
Loading and dispatch	15%	85

3.3. Observation Frequency

The MTBI (Mean Time Between Observations) recommended for precast plants is 20 to 30 minutes — higher than in continuous-flow plants (10-15 min) for two reasons:

The number of active stations simultaneously per shift is smaller (20-40 vs. 100+ in automotive).
Cycles are longer (a hollow-core slab takes 24 hours to be ready for stripping, versus a bumper that comes out every 90 seconds).

With MTBI=30 min and 5 stations observed per round, each observer covers 100 observations per 8-hour shift. For N=569, you need ≈ 6 working days with a single observer.

3.4. Shift-Stratified Sampling

In plants with 2 or 3 shifts, the study is replicated per shift. This is important because:

Night shifts usually have fewer supervisors and more equipment stops (breakdowns accumulate, maintenance is done during stops).
The most senior fitters tend to concentrate on morning shifts, which affects aggregate productivity.

4. Anonymized Practical Case: Audit of 38 Stations in Pre-slab and Hollow-core Slab Plant

4.1. Context

Plant (anonymized): medium manufacturer of pre-slabs and hollow-core slabs, 38 active stations in production, 2 shifts.
Product: pre-slabs 12-20 cm deep, hollow-core slabs 16-32 cm, spans between 6 and 14 m.
Target customer: logistics warehouses and parking lots.
Problem reported by management: OEE perceived as "low" but without objective data.

4.2. Executed Sampling Plan

Total N calculated: 569 observations (p=0.35, e=±4%).
Distribution: 28% morning shift, 22% afternoon shift (unequal shifts in staff).
MTBI: 30 minutes, 5 stations per round, 1 external observer + 1 internal validation.
Duration: 7 working days.
Study cost: ≈ €3,500 (97 analyst hours at average cost of €36.67/h, including post-field analysis and report writing).

4.3. Results

Category	% of time	Observations
Total VA	47%	267
Total NVA	32%	182
Total IMP	21%	120

IMP breakdown (the 21 points are distributed as follows):

IMP Subcategory	% of total	Points out of 100
IMP-1 Movement without load	6%	6
IMP-2 Waits due to lack of material	7%	7
IMP-3 Decision waits	3%	3
IMP-4 Breakdowns	3%	3
IMP-5 Rework	1%	1
IMP-6 Unplanned breaks	1%	1

4.4. Case Conclusions and Action Plan

IMP-2 (waits due to lack of material, 7%) was the main focus. A root cause analysis revealed desynchronization between rebar planning and formwork planning. The action was to introduce a 15-minute meeting at the start of the second shift between the rebar foreman and the production foreman.
IMP-1 (movement without load, 6%) was due to the physical distribution of rebar stockpiles. The action was to relocate the rebar stockpile to 8 meters from the assembly tables (it was at 25 m).
IMP-4 (breakdowns, 3%) was concentrated in the concrete pump. The action was to contract a preventive maintenance contract with 4-hour response, replacing the previous corrective contract.

Result at 5 months (verified with a second contrast sampling):

Pre-slab line OEE: 62% → 76% (relative improvement of 23%).
Total IMP: 21% → 9%.
Work-in-progress stock reduction: ≈ 40%.

On an annual turnover of the plant of €12 M, the OEE improvement on the pre-slab line translated into ≈ €0.6 M/year of additional gross margin (conservative estimate: 5% impact on turnover). The study cost (€3,500) had a payback of less than one month.

5. Applicable Regulations and Documentary Traceability

Precast concrete plants in Spain are subject to CE Marking under assessment system 2+ according to the European Construction Products Regulation (EU) 305/2011, through the UNE-EN standards applicable to the element type:

UNE-EN Standard	Precast Element	System
UNE-EN 1168:2005+A3:2011	Hollow-core slabs for floors	2+
UNE-EN 13225:2013	Structural linear elements (beams, columns)	2+
UNE-EN 13693:2005+A1:2010	Special elements for roofs	2+
UNE-EN 13747:2006+A2:2011	Pre-slabs for floors	2+
UNE-EN 13978-1:2006	Precast concrete garages	2+
UNE-EN 14843:2008	Precast staircases	2+

Verified source: Applus Laboratories (authorized certification body), consulted at https://www.appluslaboratories.com/global/es/what-we-do/service-sheet/marcado-ce-de-prefabricados-de-hormigon.

Implication for work sampling: VA-5 categories (productive quality control) and VA-6 (CE Marking and traceability) ARE NOT removable even if the study shows they occupy a high percentage of time — they are mandatory by regulation. An improvement plan that proposed reducing them would be counterproductive and would expose the plant to the loss of CE Marking and, therefore, to the impossibility of selling product in the European market.

Conversely, IMP-5 categories (rework due to quality defects) and NVA-3 (equipment setup) can be reduced without affecting regulatory compliance — on the contrary, the first sampling usually serves to detect where the lack of autonomous maintenance is generating rework.

6. Comparative Table: Work Sampling vs. Other Techniques Applicable to Precast Plants

Technique	When to Choose It	Advantage	Limitation in Precast
Work sampling	When there are > 20 active stations and a global map of unproductivity is needed	Low cost, does not interrupt production, identifies systemic bottlenecks	Does not give precise standard times per product
Continuous time study	When you want to standardize the time of ONE repetitive operation (e.g., vibrating a slab)	Centesimal precision, basis for MTM/MOST	Expensive: requires timing many repetitions; does not capture product variability
MTM / MOST	Very stable production line (few models, short cycle)	International standardization	Difficult to apply when the "model" changes every week
Video analysis	Critical operations with method discussion (training, safety)	Allows reviewing the observation as many times as necessary	Expensive, requires staff consent and stable lighting conditions
OEE with sensors	When the plant already has PLCs / SCADA / IIoT	Continuous, real-time data	Does NOT detect "soft" unproductivity (movements, decision waits)

The correct choice in a precast plant is usually work sampling + punctual time study of the identified bottleneck operations. Work sampling detects the problem; the stopwatch sizes it with precision.

7. Common Mistakes and Study Quality Checklist

7.1. Mistakes Seen in Real Implementations

Non-stratified sampling by phase: leads to underestimating curing NVA (which is legitimate) and inflating preparation IMP (which is NOT legitimate). Solution: always stratify.
Unvalidated single observer: a single person classifying introduces systematic bias. Solution: 2 observers for at least 1 day to measure the coincidence percentage.
Confusing "curing" with "dead time": penalizes plant planning when the problem is elsewhere. Solution: specific NVA-2 category for curing/technical wait.
Applying the "classic" N of 400 observations without recalculating: valid only for p=0.5, e=0.05. In real plants with estimated p between 0.3 and 0.7 and stricter e, the N is larger. Solution: always recalculate with the formula.
Comparing two different plants without normalizing the taxonomy: two plants with very different products produce non-comparable VA/NVA/IMP distributions. Solution: use the §2 taxonomy without renaming categories.

7.2. Quality Checklist (12 Points)

VA/NVA/IMP taxonomy documented and signed by the production manager before the study.
Inter-observer validation ≥ 85% coincidence before starting the study.
N calculated with the explicit binomial formula (do not assume 400 observations "just because").
Stratification by process phase documented.
N distribution per shift proportional to the size of each shift.
MTBI between 20 and 30 minutes.
Observation round covers between 4 and 8 stations (in medium plants, 5 is reasonable).
Total study duration covers at least 5 working days.
Results broken down by phase and by shift.
Verification that categories marked as mandatory by CE Marking remain within VA / NVA, no attempt to reduce them.
Action plan prioritized by removable IMP percentage point.
Second contrast sampling at 4-6 months to validate improvement implementation.

8. How This Study Fits with the Rest of the Ecosystem

Work sampling is not an isolated technique. In a precast plant, sampling results are usually the input for several subsequent decisions:

Time standardization: once the VA categories with the highest variability are identified, the next natural step is to time the bottleneck operations to set standards. Here is where chrono-analysis with a professional stopwatch fits — work sampling detects the problem, the stopwatch sizes it with centesimal precision.
Capacity calculation and planning: the standards allow feeding the plant's planning module and calculating the optimal load per shift. In medium plants, this planning can be digitized with a real-time production control system like Induly, which also records stops, OEE, and traceability by operator and shift.
Staffing and shifts: if the sampling detects that the main bottleneck is the wait due to lack of material (IMP-2), the problem is NOT solved by adding staff but by improving planning. If, on the contrary, it detects structural staff shortage, expanding the workforce is justified — in which case the next step is to certify the justification with a formal sampling before the works council.
Selection of technical personnel: when the plant needs to incorporate a production manager, quality manager, or rebar technician with demonstrable experience in precast, the ASETEMYT directory of industrial organization professionals allows filtering by sector and verifying credentials.
360° plant audit: if management wants to go beyond work sampling and address quality, maintenance, logistics, and costs in a single project, the umbrella ProdCont brings together the editorial, training, and consulting services necessary for this kind of comprehensive intervention.

9. Conclusion and Next Step

Work sampling in precast concrete plants is a mature but poorly applied tool when generic taxonomies and non-stratified samplings are used. With the proper taxonomy (section 2), the explicit N calculation (section 3), and respect for CE Marking (section 5), a productivity audit in a medium plant costs ≈ €3,500 and 7 working days, and usually pays for itself in less than a month through IMP reduction.

The natural next step for a plant considering this study is to define the exact scope and duration. If you want to see how this flow applies to your case, request a Work Sampling demo — the tool digitizes observation taking, validates inter-observer coincidence, and exports data directly to the report.

About the Author

Miguel Cano Otero — Industrial engineer and founder of ProdCont. More than a decade applying work sampling techniques and chrono-analysis in discrete and process manufacturing plants in Spain. Author of the Work Sampling blog, co-author of the ProdCont ecosystem of industrial organization applications.