The world’s leading publication for data science, AI, and ML professionals.

Goal-setting isn’t worth it

Engineering and data science work doesn't benefit from SMART goals, OKRs, or similar management favorites.

The goal of making baskets isn't what makes you good at making baskets.
The goal of making baskets isn’t what makes you good at making baskets.

I don’t believe in goal-setting. Over the course of my career, I’ve occasionally been pressured into setting goals, always as an initiative from an executive or manager who was hoping to inject new life into an old process. I’ve done SMART goals. I’ve done OKRs. I’ve done KPIs. I dislike them all.

I don’t like goal-setting in the workplace – particularly in creative jobs like Data Science and engineering – because I’ve consistently found it to be ineffective, I’ve often found it to be harmful, I’ve always found it to more trouble than it’s worth, and I’ve never seen it provide me with information I actually needed to achieve results. As far as I’m concerned, goal-setting is pretty much a failure all around. And there’s a better way (I’ll talk about that at the end of the post).

Goal-setting is ineffective

It’s not hard to find mountains of research claiming to show that goal-setting is effective in achieving results. I’m not aware of any study from that body of research that was conducted in a way that justifies those claims. Any variant of "goal-setting achieves results" is a causal claim. Causal claims require (at least) two things: support for the claim, and lack of support for any plausible alternative explanations.

James Clear voiced a very plausible alternative explanation. I’ll use one of his examples: say you’re a sports team and you want to win a championship. Say you set a goal of winning the championship by a certain date. Now, let’s say you make a lot of changes in order to progress towards your goal: you recruit new players, hire and manage assistant coaches, schedule and conduct practices, and even get old recordings of competitor’s games and study them.

All these things are systems: specific behaviors you commit to do on a regular schedule. If you set up these systems, and you stick with them, you improve your chances of winning the championship. If you also set the goal of winning the championship, your chances of actually winning don’t change. The systems are necessary to the accomplishment of the goal. Actually setting the goal, and even measuring progress towards the goal, is not.

Plausible alternative explanation. The goal research doesn’t address it. Therefore, the goal research doesn’t support the claim that goal-setting achieves results.

Goal-setting carries hidden costs

A few years ago, researchers at several major business schools published a working paper on the dangers of goal setting. Their paper doesn’t prove goal-setting doesn’t work – I refer to it because it tells a few illustrative stories (all of which I quote here):

  • "In the early 1990s…Sears imposed a sales quota on its auto repair staff of $147/hour. This specific, challenging goal prompted staff to overcharge for work and to complete unnecessary repairs on a companywide basis… Ultimately, Sears’ Chairman Edward Brennan acknowledged that goal setting had motivated Sears’ employees to deceive customers."
  • "In the late 1990s, specific, challenging goals fueled energy-trading company Enron’s rapid financial success… Enron’s incentive system [involved] ‘paying a salesman a commission based on the volume of sales and letting him set the price of goods sold.’ Even during Enron’s final days, Enron executives were rewarded with large bonuses for meeting specific revenue goals… By focusing on revenue rather than profit, Enron executives drove the company into the ground."
  • "In the late 1960s, the Ford Motor Company was losing market share to foreign competitors that were selling small, fuel-efficient cars. CEO Lee Iacocca announced the specific, challenging goal of producing a new car that would be ‘under 2000 pounds and under $2,000’ and would be available for purchase in 1970. This goal, coupled with a tight deadline, meant that many levels of management signed off on unperformed safety checks to expedite the development of the car – the Ford Pinto. Investigations revealed that after Ford finally discovered [a safety] hazard, executives remained committed to their goal and instead of repairing the faulty design, calculated that the costs of lawsuits associated with Pinto fires (which involved 53 deaths and many injuries) would be less than the cost of fixing the design. In this case, the specific, challenging goals were met (speed to market, fuel efficiency, and cost) at the expense of other important features that were not specified (safety, ethical behavior, and company reputation)."

I really like the author’s summary argument:

The beneficial effects of goal setting have been overstated and that systematic harm caused by goal setting has been largely ignored… Rather than dispensing goal setting as a benign, over-the-counter treatment for motivation, managers…need to conceptualize goal setting as a prescription-strength medication that requires careful dosing, consideration of harmful side effects, and close supervision.

In other words, goal-setting is a fragile way of trying to achieve results. Not only are goal unnecessary to achieve desired results (because they don’t do anything systems don’t already do), but they also create incentives that make it easy to achieve undesired results.

Goal-setting is more trouble than it’s worth

When I make arguments like those I’ve laid out above, I often get a response of "Well, yeah, if you’re stupid about how you set goals then of coarse you get into trouble. Goals work when you do them right."

It’s so very easy to say that.

Take a look at one example of what it takes to do OKRs "right":

  • Objectives need to be the right balance between general and specific, but there are no clear ways of telling where that balance lies.
  • Metrics measuring progress towards goals need to have a clear definition and people need to have access to those metrics and those metrics need to be frequently updated. All of this assumes the existence of measurement systems (a good assumption for sales teams, but a bad assumption for almost everyone else).
  • You have to clearly define the relationships between goals across different teams, but there are no clear ways of telling how to get those definitions.
  • You have to be able to recognize when a goal will increase productivity at the cost of stifling innovation so you can avoid overly-prescriptive OKRs in those cases. Again, there’s no clear way to recognize those situations.

When it’s that easy to misuse a tool, it may be time to consider that the problem is the tool itself, not the users. The amount of time it takes to set goals in a way that are both meaningful and not harmful is simply not worth it. Time I spend defining and implementing and measuring progress towards goals is time I don’t spend building things that actually add value.

I think this is especially true when you consider the amount of instrumentation necessary to meaningfully and explicitly measure progress towards an end state in an Engineering project. If your goal is a certain sales figure, the instrumentation is already built. If your goal is anything more ambiguous, and it usually is, building a way to measure it effectively will usually take more time and resources than it will take to actually do the thing that will actually accomplish the goal.

Goal-setting doesn’t yield actionable information

Let’s go back to the analogy of a sports team. Let’s say you set the goal of winning the championship, and let’s ignore the fact that you’ve set up all sorts of systems that will move you towards the goal without you ever needing to set the goal in the first place. Now, let’s say you reach the date of the championship you’ve wanted to win. Let’s say you lose.

What should you now do because you failed to meet your goal? How should your behavior change? Let’s reverse the scenario and say you actually won – you met your goal. Same question: therefore, what?

Whether you accomplish the goal or not does nothing but tell you whether you accomplished the goal. If you’ve already put together systems to lead you toward a desired future state, you already know what you need to do. Clearly-defined goal with measurable progress markers don’t tell you next steps. They don’t contain any information that allows you to reach your desired future state any faster or more efficiently.

I found this post by a former Google Product Lead insightful. He references Paul Graham’s "Maker’s Schedule, Manager’s Schedule". Makers (engineers, data scientists) are people whose job performance is measured by what they build. Managers are people whose job performance is measured by the meetings they have. Both are necessary functions (as much as makers sometimes hate to admit it, meetings are the glue that coordinate and hold together the stuff that gets built).

Talking about his experience using OKRs at Google, the author of the post said:

We consider ourselves a company founded and driven by Makers (our engineers), but somehow we settled into a Manager planning rhythm, one which mimicked accounting cycles rather than how things actually get built. "Quarterly goals?" Why are three months the right duration for building features, why not two months or four months? And there was the amusing "last week of quarter" push to try and ship all the features you’d committed to ~90 days earlier.

In a business context, most goal-setting benefits managers more than makers. If a makers need to prove their value, they can point to things they’ve built. The nature of Management is such that has a less explicit set of accomplishments, and goal accomplishment is a way to prove managerial value. It’s like assessing a teacher’s value based on student test scores – and just about as valid. But because the accomplishment (or not) or goals doesn’t tell us what to do next, that’s the best we can hope for: goals don’t tell us about anything but themselves.

In my experience, goal-setting is often the refuge of weak management. Goal-setting allows organizations to pretend that getting results less messy and more predictable than it really is. The cost of that make-believe is that both makers and managers spend an inordinate amount of time managing the illusion, and to add insult to injury, they don’t even get actionable information from doing so.

What to do instead

I’ve found the following process effective:

  1. Define a desire future state, and go ahead and call it a goal if you want. Ask yourself: "What do I want to be true about this project (or product or company) by the end of next quarter (or week or year)?"
  2. Define obstacles that prevent you from moving toward this future end state – not things that prevent you from accomplishing it, but rather things that prevent you from moving towards it. It’s a lot less work, and generally more effective, to measure direction than distance covered.
  3. Define tasks that mitigate the obstacles. Prioritize those tasks in terms of what you think will have the most impact on the obstacle, as well as what tasks are necessary for other tasks to be accomplished.
  4. Tackle the tasks in order of priority, as time and resources allow. If you’re accomplishing the tasks, you’re moving towards your goal, because you’ve already defined how the task mitigates an obstacle that prevents progress towards the goal.

Is this similar to setting goals, OKRs and things like that? Yeah, it is. But this focuses on "where do I want to be?" and "what’s going to move me toward that place?" instead of "what can I commit to accomplish?" and "what can I measure?" It requires us to set up systems regardless of goals, doesn’t incentivize counter-productive behavior, takes remarkably little time to do reasonably well, yields actionable information at every step. It’s a little shift in perspective that makes a big difference in both job satisfaction and results.


Related Articles