It happens all the time: as soon as we find a solution for a problem, the solution becomes a problem itself. The division of labor is not an exception: it increases the productivity indeed, but it also decreases in other cases.
Then he or she graduates, finds a job and becomes “installed” to some company and department. And what’s interesting about people is we tend to identify ourselves with a small group at first – in our case it’s department vs. company. We accept the interests of our near colleagues and our department much closer than interests of the company as a whole, letting alone customer’s interests. We are all great professionals for sure, we are able to demonstrate great productivity in our area but it turns out that department’s productivity doesn’t automatically guarantee the end-to-end productivity of the company. Besides, the distance between “we are able” and “we do” may be significant.
It wasn’t that crucial at the early days of Adam Smith and later at Frederick Taylor’s because it was mostly about the division of industry workers’ labor. As long as each worker performs a single operation and the sequence of operations is predetermined, coordinating them is an easy job. Just measure the time spent for each operation and calculate the conveyor speed and headcount for the given productivity. This is how the scientific process management was born.
Problems arise when we turn away from Adam Smith’s sewing needles and the legendary Ford’s T Model (which could be “painted any color as long as it is black”) to something more complicated and diverse. The greater the range of finished goods and parts, the greater the range of manufacturing operations – the more complicated is job coordination. To cope with this problem western businesses rely on computers and develop MRP, MRP-II, ERP, APS algorithms, while the Japanese invent “Just in Time” and “Kanban” (a kind of “analogue computer”). One way or another (or rather combining both) the problem can be solved.
It becomes worse when we switch from the shop floor and manufacturing processes to the office and business processes. Factors that add complexity are: multitasking, creativity and cross-functionality.
Multitasking means that we switch between tasks many times during the day or even within an hour. Conveyer workers complain about the dullness of their job, yet the office work is another extreme: the more skilled and responsible is an employee the more processes he/she would be involve into.
There are two possible ways for an employee to react. First, he or she can minimize the amount of switches between tasks. E.g. Finance processes payment orders after 4PM because processing them as soon as they arrive would mean a “productivity decrease”. This is a classic example of so-called “sub optimization”: the performance of Finance clerk would increase while the company’s efficiency from a customer’s perspective will decrease.
The second option is a tricky one: do not let anyone figure out how productive you really are. In many cases, it doesn’t make sense for an employee to do the best: the more you do – the more load you get from the boss. As for the boss, he/she is probably a seasoned professional very able to evaluate subordinate’s true performance. But is it in his/her best interest to get the most from the staff? In so many cases it’s a better strategy for a line manager to ask for extra workforce arguing that the men are overloaded. After all let’s not forget that the more the headcount, the more power and weight within the organization the manager has. Therefore, pressing subordinates would mean not only spending emotions and efforts, but also a chance to lose the career race to other managers.
Now let’s talk about creativity. It’s relatively easy to measure performance of a manufacturing worker doing routine job and to set up performance targets accordingly. But how would you measure e.g. a software developer’s performance? The number of lines of code is a very bad metric indeed, but there is hardly anything better. In fact, this is the case with all knowledge workers: there is no reliable way to measure the result.
Over one hundred years ago, a French professor Maximilien Ringelmann discovered the effect later called by his name. He performed a set of experiments in which men were pulling a rope alone or in a team. Professor has found out that performance in a team decreases: whereas a single man can pull say 100 kg, a team of two pulls 80 kg each and a team of eight – only a pitiful 50 kg. If a man is certain that no one can determine that he isn’t doing his best, then he saves his efforts, consciously or not.
The famous Parkinson’s Law says about the same: “work extends so as to fill the time available for its completion”.
Dan Ariely explored the problem with a series of experiments at MIT. Being a behavioral economist, he demonstrated that the vast majority of people are unable to resist cheating if they are sure that they will not be caught or when their cheating would cause harm or damage only in the distant future.
The rope pullers example shows how even a homogeneous team may become inefficient. Now what should we expect when a coordination of several departments’ efforts is necessary? The problems above would seem nothing compared to cross-functional coordination.
Every time an organization faces a problem that can only be addressed by joint efforts of several departments, the purely hierarchical organization is in deep trouble. A classic example is “design to order” business: the company obtains Request to Proposal; doing it right requires committed participation of a) sales manager, who communicates with the client, b) an engineer, who designs the product requested, c) Purchasing department which knows from where to buy the parts needed, d) Manufacturing who schedules the production, e) Accounting who calculates costs. A purely hierarchical organization has no chances to do the job in time and with acceptable quality. Because why should Manufacturing obey Sales? Each team has its own chief, budget, performance measures… What did you say – customers? Nobody cares.
And this isn’t the most complicated case yet. At least the sequence of activities is the same from one customer’s request to another. It becomes worse when the sequence is unpredictable: geological research at the construction site, law firm actions in court, patient at the hospital emergency room etc.
Getting back to the topic, please note that all these issues result from the separation of labor: a medieval guild master never hit anything like this.
What we see on the positive side is: humanity could raise a nominal productivity manifold thanks to the division of labor. Yet it also brings fuzzy performance measurements, blurred responsibility and poor coordination that decrease overall performance. The larger the organization, the larger are these negative side effects. The effect is non-linear, so both absolute and relative losses increase. There is a certain scale limit where benefits of the separation of labor are negated by increasing losses so the net productivity doesn’t grow anymore and starts to decrease.
How can this limit be estimated? Let’s define the metric first. Keeping in mind that manufacturing process is easier to manage because there is no multitasking, performance measures are clear and it’s a single department, it looks reasonable to use the “white collar” employee headcount as the measure of the organization scale.
This limit depends on a multitude of subjective factors like CEO personality, corporate culture, company age etc. so there is no a single number. Supposedly, the line after crossing which a company should look for the ways to compensate growing negative effects is somewhere between 20 to 100 “white collar” employees with a mean of around 50.
