Measuring workload, an unsolved problem

The Court of Cassation, France's highest civil court, has repeatedly held, starting with a 2011 ruling, that annual-day arrangements are deprived of effect where the collective agreement authorising them provides no effective workload monitoring device.

That case law led to the rewriting of many sector and company agreements, then to a more detailed inclusion of those requirements in the labour code by the 2016 law. It requires monitoring without defining its content, which is left to collective bargaining.

Workload may be approached through time, which contradicts the logic of the arrangement; through the volume of tasks, which presupposes a stable classification; or through the employee's perception, which introduces a subjective variable.

Each measures something different, and the devices put in place generally combine all three without making explicit the weight of each. The same week may seem reasonable in hours, heavy in number of files and crushing in the perception of the person living it.

Ergonomists, for their part, distinguish prescribed work, what is asked, from real work, what is actually done to meet the demand. Workload arises largely from the gap between the two, which neither hours nor task counts bring to light.

Models of strain at work, in the line of the one proposed by Robert Karasek in 1979, distinguish demand, autonomy and support, and measure those dimensions through validated questionnaires. They describe a relationship between requirements and resources rather than a quantity of work.

That approach is the most used in national surveys. It does not supply a measure of workload as such, which explains why it answers the legal demand for individual monitoring poorly. Other instruments, such as the NASA-TLX scale published in 1988 by researchers at the American space agency, measure the load of a specific task and do not transfer to monitoring a post over a year.

Work on interruptions and on the dispersion of attention, such as that of Gloria Mark, measures an aspect that neither time nor task volume captures: the fragmentation of activity. Resuming an interrupted task has a cost that ordinary counts ignore.

Available measures rest on direct observation or on software usage traces, methods that are costly and applicable to small samples, which prevents their routine use in firms. Exploiting those traces on a large scale also raises surveillance issues that the law strictly regulates.

The most widespread devices rest on a dedicated annual interview and on self-reported alerts. Their effectiveness has not been the subject of published evaluations.

The gap between a precise legal obligation and the absence of a validated instrument is a frequent situation in labour law, and it passes to collective bargaining a problem research has not solved.