Original image by Luc Viatour / www.Lucnix.be


I have been at Twitter for almost two years and now, in the midst of my third full bi-annual “feedback” cycle—the process by which Twitter employees write self-evaluations, peer-evaluations, and manager evaluations, and ultimately get some feedback delivered by their manager that tries to distill it all—it strikes me that if you could chart my probability of leaving Twitter over time, it would probably spike every six months sometime during feedback. Now, I may be a particularly grumpy fellow but, based on people’s Tweets and conversations around the office during feedback, I’m pretty sure I’m not the only person who finds feedback at best a meaningless chore and at worst a toxic, off-putting experience. And I don’t think this is unique to Twitter—most companies beyond a certain size have some similar kind of regularly scheduled process and I’ve never heard anyone say they really love the one at their company.

As with any analysis of a process, the first step to analyzing these proceses is to understand what they are for. This can be tricky because most such processes are trying to do two things. On the one hand, they are the formal way we give each employee information on how they’re doing so they can do their job better and improve their professional skills. On the other hand, they are also often used to inform decisions around who gets promotions. These are actually, I’d argue, two quite different purposes. To tease them apart and see why it’s hard to achieve both at the same time, let’s look at how the process might look if it was purely about one or the other.

Feedback for professional growth

A feedback process 100% aimed at professional growth would, I suspect, be totally divorced from promotions and compensation bumps. Not because those things should be unrelated to professional growth but because truly reflecting on how you can do better and being open to feedback from your peers and managers is already tremendously difficult; when you are also worrying about whether or not you’re going to get that promotion or raise you were hoping for, it’s probably impossible.

Negative feedback—i.e. the most valuable kind—is hard to take even in the best circumstances. Even the most sensitively couched suggestions from people we know truly have our best interests at heart still provoke rationalizations and denial in all but the most enlightened of us. Negative feedback that comes as part of an explanation of why we didn’t get the promotion we had our heart set on is likely to trigger a full-on fight or flight reflex.

Also, when writing evaluations—or anything, really—the audience matters. If you’re writing a self-evaluation to be read by your manager and used as a starting point in a conversation about how you can improve, that’s very different than something that’s going to be read by a promotions committee that will determine whether you’re worthy of a title bump. And writing feedback intended to be read by a peer or their direct manager, aimed at helping them improve is, similarly, very different from writing something you know is going to be used to judge them.

Finally, a process focused on professional growth would ask questions that focus on the future rather than the past. Rather than asking, “What impact have you had?” or, “How have you done your work?” we’d ask: “How could you have a greater impact?” and “How could you improve the way you work?” Not only are these forward looking questions, they are also more open ended: a legit answer to “How could you have a greater impact?” might be, “I would be allowed to move to a team where my skills are more useful”. Whether or not that’s really the best way to increase a particular employee’s impact, it’s at least a good starting point for a conversation between them and their manager. The “what impact have you had?” question is also pretty useless for personal growth since so many factors outside a person’s control affect how much impact they’ll have.

Feedback for determining promotions

On the other hand, if we were designing a process solely to determine which employees most deserve promotions, we’d have to come at things differently. For starters we’d have to get clear about the purpose of promotions. Are promotions just another kind of reward, like a bigger salary or more stock, handed out to employees we most want to keep happy so they’ll stay at the company? Or are promotions a way to publicly reward the kind of behavior we want everyone to emulate? Or is there actually supposed to be some correspondence between title and responsibilities—a Director naturally has broader responsibilities than a line manager; should a Senior Staff Engineer likewise be expected to have broader responsibilities than a Software Engineer II?

There’s obviously a correlation between titles and responsibilities but it’s noisy and tenuous. This isn’t, as a friend of mine observed, the army, where Captains have to salute Majors and Majors have to salute Colonels. And as a signalling mechanism titles are kind of crap because even when you know someone is senior to you, unless you work very closely with them—and not necessarily even then—it’s difficult to know what about them you should be trying to emulate. And of course we don’t need to give someone a title in order to hold them up as a role model—a manager can tell you, “You notice how Sally always does X? If you could learn to do that, that’d really make you more effective.” Finally, I’ve never heard a clear articulation of the objective criteria for being promoted to any title.

The most parsimonious model, then, is that titles are just another reward mechanism like raises and extra stock. And that’s not all bad: since they are a publicly visible reward they force management to make their values explicit. If we see that only a certain kind of engineer is ever promoted (low-level systems hackers, user-facing feature developers, white males, whatever), then we can see what kind of engineers the company most values.

Of course their public nature is what makes them tricky—even if we did something really dumb like give somebody $1,000,000 worth of stock who then turns out to be a bozo, while unfortunate, the downside is limited to the $1,000,000. However, if we promote the same engineer to Principal Engineer, then we risk demoralizing everyone who actually knows the new Principal Engineer is a bozo and confusing them about what is valued by the company. The total cost of that mistake—in lost productivity, regretted attrition, and extra money spent to retain disgruntled employees—could be much higher than $1,000,000.

So the stakes are high. How shall we proceed? It doesn’t seem crazy to say, the more information the better, and to use some kind of 360-degree feedback system to gather evaluations from each individual, their manager, and their peers to use as the input to the promotion decision. Certainly, the self-evaluation can help fix the bug of having employees’ promotions depend entirely on how good their managers are at noticing, remembering, and communicating all the promotion-worthy things the employee has done. Also if the employee and the manager have very different estimations of how the employee is doing, it’s worthwhile for that to come out.

But if all we care about is correctly allocating the reward of titles, then we should recognize that we’re asking for a very different kind of self-evaluation than one aimed at personal growth. We should be honest about what we want and just ask: “Why do you think you should be promoted?” Let everyone engage in honest self promotion—listing their accomplishments and telling us about the subtle and not-so-subtle things they did to help their teams succeed. Let the peer feedback be focused on corroborating or contradicting the self assessment and pointing out other aspects of the employees behavior that might be worth rewarding with a new title. For people who aren’t ready to be promoted (which they hopefully know because they’ve been talking with their manager about it) we could ask, “What do you think you need to do to be ready to be promoted?” These straightforward questions would at least allow for a useful conversation between employees and managers about what it takes to get promoted.

But having such a process doesn’t let managers off the hook for the rest of the year. Good managers should be looking after their employees’ career growth if only because it’s one of the tools they have to keep their employees happy. And having an employee-involved promotion process makes it even more important that managers handle promotions well. That means being clear with employees about when they are likely to be eligible for promotion. While it is irritating enough to go through a feedback cycle knowing you aren’t even being considered for promotion when you think you should be, it’s even worse to know you’re up for promotion and to be passed over. The absolute worst is to have your boss tell you they think you should be promoted but you weren’t.

Who decides?

Which raises the question, who decides on promotions anyway? There are conflicting interests at play here. In a perfect world, where all managers were competent and trustworthy, we’d simply let managers decide when to promote their direct reports. This has the advantage that it allows managers to use promotions as positive feedback without the risk of demoralizing people by telling them they’re up for a promotion and then having to tell them it wasn’t approved.

However it’s easy to see how unchecked control over promotions could quickly lead to title inflation as every manager uses the relatively cheap reward of a fancy title to keep the troops happy and pretty soon the whole company is made up of Principal Engineers.

To control that, we could treat titles as a kind of currency and to grant each manager a “promoting authority” just as they have “spending authority”. So maybe a low-level manager can unilaterally promote a SWE-I to SWE-II but has to check with their own boss before moving a SWE-II to SWE-III. Promotions to the higher levels would have to be approved at correspondingly higher levels of the management chain.

The potential flaw with that system is that different managers may have very different capabilities when it comes to making the case for their people to get promoted. In the worst case, promotion decisions get dominated by politics and cronyism—if you work for someone who’s buddies with the VP of Engineering your promotion goes through but if you work for a less well-connected manager you’re stuck forever.

I see how that could be a problem but not that the solution needs to be any more complicated than better management. If the VP of Engineering or the CEO is engaging in cronyism in this, they’re probably doing so in other contexts as well and we might as well all start looking for a better place to work. Seems to me it’s up to the higher level managers to make sure their lower level managers know how to recognize who should be promoted and how to advocate for them effectively. To the extent different managers are still better and worse at this, it is up to the higher-level managers to adjust appropriately.

But for whatever reason, quite a few companies have decided to work around this problem by establishing “promotion committees” made up of individual contributors at or above the level for which promotions are being considered. These committees, sometimes likened to tenure committees in university, make the ultimate decision on all promotions.

Obviously such committees need to be carefully managed lest we replace one political, connection-based process with another one, just with a different cast of characters. Thus it’s imperative that the committees are only allowed to make their decisions based on some kind of objective information. Some companies even try to “blind” the committees, obscuring who the promotion candidates are, so the committee can focus only on the information about what the candidate did and how they did it.

There are all kinds of problems with punting the problem of promotions to a committee not least of which is that it makes promotion decisions—especially if the committee membership is secret—seem almost totally arbitrary.

But the real problem with promo committees is that they are rooted in the false notion that there are objective criteria to be applied to promotion decisions—that there is a linear scale of technical merit with titles acting as the tick marks on that scale. If that was the case, then it would be a technical problem to measure each candidate’s technical merit and place them at the appropriate point on the scale and it’s not crazy to think that the best people to perform that technical task would be technical workers, not managers.

On the other hand, if we accept that promotions are really a kind of reward the business can use to motivate certain employees, then it’s not a technical problem at all but a management problem. To take a couple rather coarse examples, it may well be that the business would do better to reward the people who take on unglamorous, less technically challenging tasks rather than the technically sweet bits of hacking that are going to earn their authors plenty of nerd cred anyway. Or maybe there’s a certain area of the product that needs extra love and a few promotions in that area will help keep key developers and encourage other developers to consider working in that area. More subtly, the ways an individual employee can contribute to the business are extremely multi dimensional and non-fungible—behavior that in one context might be hugely valuable and worthy of reward, in a different context might be counter productive and damaging. It is the job of managers to make those distinctions. If we don’t have managers who are up to it, we need some new managers.

The notion that titles are not awarded according to some elegant and rational scheme may be distasteful to some engineers. I only recently came around to accepting it, myself. But if you think about it, titles are an artificial business construct so it makes sense that they follow the subjective and ad hoc rules of business rather than the perfectly logical laws of the computer. And once you do accept that, you’ll see forming promotion committees as a gross abdication of management’s responsibilities to use one of the more powerful management tools at their disposal.

Why do we do everything in lockstep?

The last feature common to feedback processes at a lot of companies is that all feedback is done in lockstep across the company or department. Everyone writes their evaluations at the same time and when promotions are tied to feedback all the good and bad news about promotions is delivered in a couple week window.

If feedback were primarily about self-reflection and growth, then the lockstep schedule wouldn’t matter much. In that case it would be fundamentally a personal process and it wouldn’t matter that lots of other people happened to be going through the process at the same time. It might even foster a certain kind of camaraderie.

But when feedback is about who gets promotion, it changes things. Not only do the people up for promotion know who they are but likely the people around them know as well. When the cycle ends and promotions are announced, everyone knows who got one and who was left out. And the effects of those decisions can reach beyond the individuals up for promotion.

For instance, one of the most frustrating feedback cycles I experienced was one where I wasn’t even up for promotion. My manager had told me I hadn’t been at Twitter long enough to be promoted so I wasn’t worrying about it; it was a little irritating but whatever. However, one of the people I worked with was up for promotion and, as far as I could tell, richly deserved it. When he didn’t get promoted it spun us both out and made me super angry. Both my irritation at not being up for promotion myself and my anger over my teammate not being promoted were totally unnecessary, caused only by the lockstep feedback cycle.

On a bi-annual feedback schedule, twice a year every employee is forced to think about whether they think they should be promoted and to be dismayed when they see good people get passed over. And it’s worse for more senior employees because the funnel narrows at the top—many Staff engineers will never make Senior Staff and those that do will probably have to wait longer than the folks waiting to go from SWE-II to SWE-III so they will live through more six-monthly periods of disappointment. Personally, I’d far rather be told by my manager that the average time at my current level is X years with a standard deviation of Y months and, in the meantime, here are some of the things I should be working on to get better at my job. And even more than that, I’d like be able to trust that if I actually work at improving my skills and at doing what is valuable then promotions and raises and all the rest will take care of themselves.

Obviously there’s a risk that without an organization-wide process forcing the issue, promotions will fall to the bottom of managers’ todo lists. But that also seems like a totally manageable problem: if we can make everyone drop what they’re doing every six months to write feedback, presumably we could instead make managers go through a bi-annual exercise of documenting their current thinking on the promotion prospects of all their direct reports and to track how often managers have a serious conversation with each of their direct reports specifically about career development. When the time is right to actually put someone up for promotion, the manager can gather the self-evaluation and peer feedback and put it into a packet.

If managers are doing their job, most of these promotions should be approved; the few that don’t should come back with clear feedback about why not so both the manager and employee can improve. Directors and VPs should be making sure that all the managers below them are striking the right balance between trying to get their folks promoted and not putting people up for promotion who will be disappointed. While they’re doing that, the upper-level managers can also make sure that managers are not letting unconscious bias skew their choices of who to put up for promotion.

The biggest difficulty with such a system is that it relies heavily on having good managers throughout the organization. This is simultaneously a serious problem—truly good managers are hard to find and develop—and a total cop out. If we design our processes to try to compensate for mediocre people, we’ll definitely end up with mediocre people because good people won’t put up with the processes. And the result will almost certainly be mediocre in the end. The only other reason I can imagine breaking out of the lockstep is hard, is because doing things in lockstep makes some kind of financial accounting or planning easier. That seems like a dumb reason to demoralize a good chunk of your employees every six months, but what do I know?

This is obviously a complex topic and despite the length of this essay I’ve only scratched the surface. The issues are so tangled up that it’s hard to make definitive statements without explicitly stating a million assumptions. To close, here are a few things I think are probably true:

Thanks to @marcprecipice and @skamille and all the others who helped me explore these topics on Twitter. The views expressed here are my own but your insights were valuable in helping me figure out where I stand.