Mean

some stories about my performance reviews at msft, and thoughts on how to be better

My Five for Friday yesterday included a link to this article discussing Microsoft’s review system (or one of them), that I thought would be fun to reflect on a bit.

When I joined Microsoft in 1995 (please, no age jokes), we did a full on 1-n stack rankings twice a year (I didn’t join in on the fun for a few years, but it wasn’t a secret). At the time, Microsoft had a scoring system that went (theoretically) from 2.5 to 5.0. The bulk of the org would receive a 3.0 or 3.5 rating (basically, meeting expectations, or doing slightly more). A certain percentage of the org could get a 4.0 or higher rating (higher scores were rarer, but would usually occur, although I never saw a 5.0 given). Then, 10% would have to get a 2.5 - which meant you were most likely getting fired. It’s worth noting that a 2.5 didn’t guarantee that you were done, there was a little bit of wiggle room and politics you could play, but I don’t remember anyone every getting a 2.5 in two consecutive reviews - it was an “up or out” number.

The “Fire 10% was a Jack Welch thing. He talked about it in (I think) all of his books. Here’s a discussion on it from Jack: Straight From the Gut

Some think it’s cruel or brutal to remove the bottom 10 per cent of our people. It isn’t. It’s just the opposite. What I think is brutal and “false kindness” is keeping people around who aren’t going to grow and prosper.

That was the inspiration, but as you can imagine, the process was stupid and painful. Bubble sorting an organization of different roles and levels of seniority was a waste of time, unhelpful, and mean. Even though we ranked from top to bottom, the only place it really mattered was where the lines were drawn. Pay raises, bonuses, and stock changed (with some variation) at each score ranking, and the arguments on who had to be a 3.0, and who was “definitely not” a 2.5 were legendary.

It’s worth mentioning that we had “guidance”, but our budget tools left wiggle room. That meant if I had employees who cared more about money than their score, I could give them a 3.5 with virtually identical rewards to someone who got a 4.0 if I needed to. This is obviously wide-open for abuse, so sometime in the early 2000s, we switched systems.

It was worse.

Nine Boxes

In the early 2000’s, Microsoft changed the performance system to an almost traditional nine-box system - a 3x3 grid of performance and potential. Nine box can be a really useful system for talking about employees (more on that later), but Microsoft Microsofted it. They couldn’t shake the ghost of Jack Welch, so they said that only 20% of people could be top performers, 70% would be middle performers, and that 10% must be poor performers and managed out. This system also put more strict restrictions on merit, bonus, and stock for each part of the grid. Again, sometimes you could wiggle out of getting let go if you were in the 10% bucket, but even if you weren’t fired, the 10% bucket was a curse. Many orgs would not allow you to change to another team if you were in the bottom bucket, so if you were a bottom 10% in a high performing org, you received no pay raise, and could not go to another team - even though you could be a top performer on other teams.

Numbers…

After that, I stopped paying attention. They killed the nine box system and changed to something else, but in the end, it was still stack ranking. We couldn’t call it stack ranking, but it was stack ranking. We didn’t have quotas, and we were no longer required to select people to be fired, but we had a budget, and now, zero flexibility with pay changes. The number you were given in the totally-not-a-stack-ranking meeting was mathed into your rewards by someone in HR or Finance.

…Stories…

Before anyone thinks that these complaints come from spite, I think it’s worth calling out that in 22 years at Microsoft, I received exactly one poor review. I was in a group I really liked, and I ended up in the top few people in the entire org of hundreds of people in the cross-org totally-not-a-stack-rank meeting - but our SVP stepped in and mandated that senior ICs were by definition, under performers, and gave me the lowest score he could give me without firing me. In the end, it was good, because it gave me a reason to leave that group and join the XBox One team, where I probably did some of the best work in my career. Worth noting that the ranking our SVP forced my manager to give me put me in the cursed-bucket where I had to get our HR business partner and an SVP in the XBox org to “approve” my transfer.

I just looked him up on linkedin, and that guy is still at Microsoft. I should (but won’t) tag him in this post.

… and Cheaters

The logistics of the systems are what they are. The horror came from the behavior of people when put into this system. At some point, Microsoft settled on a yearly merit cycle that was done in June. The joke was “peak in May” - meaning that May was a great month to just crush it at work, since proximity bias would have a huge say in your yearly pay increase - even if you did nothing for the rest of the year. As pointed out in the linked post, people would undermine others, suck up to leaders, coast on bad teams or avoid joining strong teams. In the end, we had a system with flaws, and a bunch of smart and driven people who easily figured out how to exploit those flaws. It was painful, but I survived there for over two decades, so it wasn’t that painful.

I’m told by my former colleagues that things are much better at the big M, and I’m truly happy to hear that. I learned to work with the system and keep my integrity and morals (and when I couldn’t was when I finally left). Books like Orbiting the Giant Hairball were helpful - but it was hard.

A Better Way?

That’s a lot of words to reflect on what’s wrong with performance review systems, but either despite that experience, or because of that experience, I’m actually a strong believer in performance review systems.

As I mentioned previously, Microsoft completely botched nine-box systems, but I actually love using a nine box system for talking about employee performance. To be clear, there’s little value in plotting people into boxes and making sure the numbers work and then moving on. The value comes from the leaders of the org talking about why someone is in a particular box. We can identify who needs help, who could take on bigger challenges, or who may be in the wrong role. It helps us think about impact and growth consistently across the org.

Traditionally, the nine box approach is a 3x3 grid of performance and potential.

Over years of using the nine box grid, I have two complaints. One is that the layout implies a somewhat equal distribution between boxes (which is completely wrong), and the second is that potential is extremely subjective and difficult to calibrate (e.g. a new hire out of college has huge potential, while a strong principal engineer has less potential???).

So instead, I often map the y axis to culture - as in, how well does this person exemplify the culture we want in our org. Because our culture is (or can be) defined, we can then give objective examples to help us understand where someone may fit.

I end up using something like this:

…which is only seven boxes, but can be valuable. In a healthy org, most people will end up in that big blue box in the middle. The interesting discussions happen when we discuss everyone else, why they’re where they are, and what their path is. It’s important to talk about growth paths for everyone - those that are struggling, as well as those who are your star performers.

A Culture of Feedback

I have a post exclusively on Feedback in my backlog, but it’s worth discussing in brief here. Another quote from Welch’s book above (bold mine).

“..(removing the bottom ten per cent of employees) works because we spent over a decade building a performance culture with candid feedback at every level. Candor and openness are the foundations of such a culture. I wouldn’t want to inject (this approach) cold turkey into an organisation without a performance culture already in place.”  

The vast majority of my performance reviews at Microsoft were the only time I ever received feedback. Of course, this meant that performance reviews were an anxiety inducing mess of emotions where I never remember what happened. I think it’s absolutely critical that leaders talk about how their org is performing and align on what good looks like and how we want to grow people. But we have to give feedback constantly. We have to normalize it by giving both positive and corrective feedback all the time. All. The. Time. Despite what (I think) sounds horrible above, it would have been 100x better if Microsoft (or any other company, because I feel like I’ve picked on old-Microsoft enough) also focused on building a culture of feedback. Start with that. Start letting people know how they’re doing, and how they can grow. Once you’re good at that, you can talk more about how they’re doing and how they can grow.

If you have a performance management system and you use it as the only time to give employee feedback, that’s horrible. And mean. Do better.

-A