blog podcast

Perhaps Metrics After All

In physics the observer effect stipulates that an observed system is disturbed by the act of observation. As someone that spent a considerable part of my adult life in corporate work life I can confirm that this effect is very real also in corporations.

It is one of the reasons why people caution against measuring individual performance based on metrics. For example in the case of a programmer, what if you measured performance based on the number of rows written? Obviously you incentivise writing more code than necessary (and updating the npm lock file as often as you can). If you base it on story points delivered? Obviously you incentivise to bloat estimates. More conscientious workers will suffer, those more okay to bend the rules will prosper. In the end you might not end up with the workforce composition you need.

The problem is that modern day corporations operate in very fuzzy circumstances. It is often hard if not impossible to create feedback loops to decide if a given action last Monday was a good or a bad action. Obviously the ultimate metric for a corporation is money, but the bigger your corporation gets, the more unclear the relationship between the individual’s actions and the profit of the company as a whole becomes.

Furthermore, the workforce in corporations are often specialized. You end up with different work roles with different responsibilities, and sometimes the local optimum of one role might not necessarily contribute to the global optimum of the corporation as a whole. A classic example which I have mentioned before is the birth of DevOps where it was recognized that the Devs department and the Ops department usually end up with conflicting interests which hamper overall quality and speed of the organization as a whole.

The problem expands past Devs and Ops, you can easily imagine how HR might end up with different interests compared to marketing that might have different interests compared to the devs etc.

It was conflict of interests like these that (likely) gave birth to the concept of OKRs (Objective Key Results). The idea of OKRs was that you try to negotiate strategic goals on a corporate level. You then go down into different subdivisions of the company and try to find strategic goals there that also align with the corporate strategic goals. You repeat this exercise down the hierarchy of the organization. You then also make sure to create “Key Results” i.e. metrics, that measure how well you are or are not reaching these goals that you have.

Evidently you are now observing and for that reason you are affecting the system you are measuring, so you have to be very very conscious to define key results in a manner that won’t backfire on you. Using these key results now you have an easy way for different parts of the organization to know if their actions are actually adding to, or detracting, from the overall goals of your organization.

None of this is very new or revolutionary, it was introduced already in the 70s according to Wikipedia. It does highlight something that I feel like my field (IT) sometimes forget. This is an amnesia that might take its form in people saying it’s of no worth to perform estimates. It is an amnesia that might take its form in rejecting McKinsey’s report on developer performance. It is an amnesia that might take its form in unwarranted confidence of knowing it all. What we sometimes forget is that IT is in itself only one aspect of a corporation. Software development never happens in a vacuum and whatever software is being created is only as valuable as it becomes together with the effort of other parts of the company. A program never sold or distributed carries no value.

In closing I would like to say that you might be completely right, the measurements that are trying to be imposed on you as a developer from above might be hampering the amount of good quality code you can create. It might very well be preventing you from reaching the local optimum of the best possible code. Just please remember that the metric of success for a company is complex and it is likely that there are things that look counterproductive to you, but that makes a lot of sense when seen from a higher perspective.