How many workers should be interviewed in a company / factory? 

How to check the validity of the completed interviews

Sampling strategy for Decent Work Check workers survey

How to identify which workers you want to interview in a company / factory? The aim of the survey is to provide a realistic picture of the working conditions of the workers in the factory. The survey therefore should not interview only the lowest paid workers, or not only the management. The survey should reflect the total workforce in the factory. The workforce includes workers on a fixed-term contract or on a casual contract. The workforce also includes cleaners, packers, and similar staff.

To specify the sampling of the factory workforce, four questions in the company / factory survey are relevant:

Question 1 - How many people are employed in the factory?

Question 2 - How many of the employees are women?

Question 3 - How many of these women have a permanent contract?

Question 4 - How many of the male employees have a permanent contract?

Based on the ansers you can specify how many workers have to be interviewed within each category. Let's assume that the factory has 500 workers (Question 1), of which 400 women (Question 2) and thus 100 men (Question 1 - Question 2). Of these 400 women, 150 have a permanent contract (Question 3). Of the 100 men, 75 have a permanent contract (Question4). If you target to interview 50 workers in this factory, the composition should be

F_perm = Q3 / Q1 = 150 / 500 = 30%

F_temp = (Q2 - Q3) / Q1  = (400 - 150) / 500 = 250 / 500 = 50%

M_perm = Q4 / Q1 = 75 / 500 = 15%

M_temp = (Q1 - Q2 - Q4) / Q1 = (500 - 400 - 75) / 500 = 25 / 500 = 5%

TOTAL 100%.


So, before starting interviewing the workers in a factory, the factory survey should be completed. Based on the answers, you will be able to compute the percentage workers to be interviewed in each category.

Of course, during the interview phase it may be difficult to reach the targeted percentage per category. Therefore, I advise to be a bit flexible with these percentages. For example if you target to interview 30% women, you could also be satisfied with interviewed women between 28-32 percentages.  

The sampling strategy is only based on gender and labour contract. However, other characteristics of the workers should preferably also reflect the workforce. For example, if 50% of the workforce is unionised, you should not only interview union members, but also an approximate equal number of non-union members in the workforce. Another example. If 50% of the workforce has a job as sewing machine operator, you should not only interview sewing machine operators, but also workers in other jobs. These examples are no strict requirements, but should preferably as much as possible be taken into account.

The number of workers to be interviewed depend on the factory size. The smaller the factory, the more workers should be interviewed. For example, in a factory of 100 workers, 30 workers should be interviewed. In a factory of 500 workers, up to 39 workers should be interviewed. In a factory of 1,000 workers, up to 41 workers should be interviewed. In a factory of 5,000 workers, 42 workers should be interviewed. These are soft targets, just see how far you get.


The main purpose of the validity tests is to check whether interviewers have faked an interview or have made data entry errors. The Decent Work Check  survey has a tool called questanalyser. This tool allows the coordinator to  monitor how many surveys have been completed by which interviewers. It allows also a breakdown by region, so the coordinator can monitor progress according to the agreed scheme.

The Decent Work Check survey provides also access to the surveyed data. This database is in Excel format. The coordinator can use this database to perform extra checks. For example, Excel allows to make a sum of a range of cells. So, make a sum all answers of an interview. Compare per interviewer the totals across the interviews. Conduct an extra check in case one or more sums are exact the same. This may point to a fake interview. Another check can be to sum empty cells, for example count per interview (thus per row) the empty cells for the last 20 columns of the database. Excel has a command to count empty cells within a range. Check if there are many empty cells. If so, the interviewer has skipped survey questions. Please, check the wages.  If wages are paid per month, are the reported wage realistic? For example, is the monthly wage between 1,000,000 and 5,000,000 Rupia for a full timer. So, an amount of 100,000 is invalid, and so is an amount of 50,000,000. If this is the case, the interviewer has not entered the wage data correctly. In all these cases, identify the interviewer in the column INTERVIEWER_API, and discuss with him/her the problems you’ve identified.




2020 JAN 11, Kea Tijdens, WageIndicator Foundation