How can you improve the quality of your promotion decisions, increase trust in them, and decrease bias and politicization?
When I joined Mailchimp two years ago as a Senior Director of Engineering, the company was expanding rapidly. During a growth spurt at any company, it’s common for job titles to get out of whack. New hires from different kinds of companies, acquisitions, and additional cities can skew the expectations of titles and expertise. Existing staff may take on responsibilities not reflected in their titles. Plus, promotion processes that evolved in an era where everyone knew each other don’t hold up in an environment with little or no overlap among people – and can leave qualified staff behind. As I took on reports, Mailchimp’s engineering department was experiencing all of these trends, so it wasn’t surprising that several of my team seemed nearly ready for promotion.
Shortly after I joined, the department introduced engineering ‘levels’. Also known as a career ladder, the document succinctly laid out the impact we expected our nearly-400 engineers to have at each title from Eng I to Distinguished Engineer (a parallel ladder for engineering managers followed some months later). This kind of framework was popularized by Camille Fournier about five years ago and has become common in engineering departments. It’s an important tool for ensuring that you have fair, clear, and transparent criteria for promotion, making it less likely that people will receive promotions on the basis of political factors or will be denied promotions on the basis of discrimination.
At Mailchimp, two engineering managers had led the development of our department’s levels, methodically testing and socializing them. Knowing the thoughtful work that had gone into the framework, I was confident that my reports and I could use it to assess their readiness, and I could move ahead with their promotions from there.
I was half right.
Career ladders aren’t enough
Although career ladders like ours offer common touchpoints for assessing an engineer’s readiness for promotion, they aren’t designed to provide a process for ratifying promotions. Put another way: once a manager has used the ladder to determine somebody is ready to be promoted, how do they demonstrate that to other people? Who else needs to be convinced? In what timeframe? With what data? How will the department ensure a similar interpretation of the levels framework across managers? How will disagreements get sorted out? Without a standard approach to these things, it’s easy for promotions to lose the foundation of transparency and fairness that a levels system can build.
After it became clear that many of us had different ideas about how to use levels (making some promotion decisions nearly impossible to finalize), I paired up with another department leader, Vicki Bodman, to create a standard process with an appropriate amount of structure for a department with hundreds of people. We started by focusing on senior ICs (promotions to Staff Engineer and above) where we had a number of candidates who had been approved by their individual managers but for which the next step wasn’t clear, and for which there was no deadline for getting out of limbo. In addition, because promotions at higher levels can have a big impact on the organization and on individual careers, we wanted to be particularly intentional with these decisions.
To understand what our colleagues wanted out of a more structured system, we interviewed the rest of the engineering leadership team, plus other managers and existing senior ICs. We also talked with people from other companies to broaden our thinking. Our interviews distilled to one basic idea: the department needed to make good promotion decisions that any staff member could readily trust. This led us to a handful of requirements, including:
- The process should have clear and published timelines;
- The process should have relevant, clearly-defined decision-makers;
- The decision-makers should have consistent information about candidates;
- The decision-makers should use a standard and collaborative process for discussing and deciding on promotions.
We hoped that if we could establish something that addressed these needs, it would improve the quality of promotion decisions, increase trust in them, and decrease bias and politicization. (I say ‘decrease’ intentionally. Bias and politics are difficult, if not impossible, to fully eradicate. We were aiming to reduce the risk of those factors influencing outcomes.)
A popular pilot
We tested our new system in the first quarter of 2020 with the senior individual contributor (IC) promotions that had stalled in the previous months. We knew we’d created something useful when, before the cycle was even over, a senior leader in the department wanted to promote a manager, and all of the other leaders immediately asked that we use the new process. Here are the basic elements of it.
Set a schedule. As it happened, we developed our system just as Mailchimp HR shifted the company from allowing promotions at any time to a quarterly cadence across departments. That gave us a predictable cycle we could use to set our schedule. At first, I thought quarterly might be too long between decisions, but I’ve come to see that it nicely balances the need to advance people with enough time to do it thoughtfully.
Relevant decision-makers. We implemented a two-tiered system. First, every promotion proposal is reviewed by a unique panel of three to five senior engineering ICs and managers, which must include people who know the candidate’s work and those who don’t (it cannot include anyone in the candidate’s reporting chain). That group makes a yes/no promotion recommendation to the department senior leadership team, which reviews all the recommendations and decides whether to ratify them.
Consistent information. Within a quarter ahead, managers with candidates in mind must fill out a short ‘intent to promote’ form. It automatically notifies senior ICs and managers in the department about potential promotions, giving them a chance to let the promoting manager know about any concerns regarding the candidate. This has been a good mechanism for surfacing problems. With new, timely information, managers now often decide to hold proposals until they can work with their reports on issues that come up.
For candidates they want to advance, managers must write a proposal using a standard three-page template. Two pages ask for examples of the candidate’s impact that map to our levels document. There are also small sections for discussing future areas of growth for the candidate and for describing why the business needs an additional person at the proposed level. The third page must comprise quotes from people who work closely with the candidate, including peers, plus people more junior and more senior. The quotes are important, in part because they ensure the manager is getting input from people who have varying power relationships with the candidate.
Standard and collaborative process. We put in place simple systems for the review panels and the senior leadership team. Each review panel gets the written proposal a week ahead, and then they meet for an hour. During the first leg of the meeting, they ask questions of the candidate’s manager. Then the manager leaves, and the panelists each vote on whether to: recommend the candidate for promotion without reservations; recommend promotion with reservations; or not recommend promotion at this time. Finally, the panel discusses their votes, reaches consensus and submits their overall recommendation to the senior leadership team, along with any notes about the candidate. Sometimes, there’s easy consensus on the panels; sometimes, it takes hard discussion. We’ve been surprised to find that in nearly all cases, the panel notes about the candidates are a rich source of feedback.
Once all the recommendations are in, the senior leadership team meets to review the whole batch for a quarter. While the group may discuss individual candidates, the idea is not to relitigate the recommendations. Instead, we focus on business needs and whether this set of promotions, along with any new hires, aligns with the ‘shape of the department’ we’d like to see. That is, approximately what percent of people we want in learning roles, execution roles, and teaching-focused roles. In practice, several of the recommendations over the past three quarters have sparked deep discussions among the senior leadership team, but we’ve ratified all of the recommendations. Plus, those intense conversations have flushed out needed clarifications.
Adjustments and additions we want to make
Overall, the new system has been a success. It’s predictable and relatively transparent. It involves appropriate people. It surfaces good information and fosters important communication. As a result of all this, the decisions tend to be trusted by many people. But there’s still room for improvement. In fact, as the company evolves, the process will always need to change. In addition to tinkering around the margins, here are a few of the bigger things we’re looking to add and adjust in the coming year. As you’ll see, they’re all related to improving equity in the system.
Training for managers and review panelists. Not all managers are equally good at writing proposals or answering questions during review panels. And not all panelists understand their roles in the same way or know how to listen well. In several cases, panels have not recommended candidates for promotion, because they didn’t think the manager was able to back up their proposal. We’re particularly concerned that managers who are underrepresented in tech may not be heard by panels that are primed to listen primarily to white men.
Self-nomination and peer-nomination options. As it stands, our process relies on managers to propose their reports for promotion. But what about people who have had three or more managers in the recent past? Or people who believe their manager isn’t recognizing their growth and contributions – be it because they're matrixed, they have a personality conflict, or the manager’s biases are in play? In addition, a number of people have asked for a way to formally flag peers whom they believe are under-leveled. We’d like to find straightforward mechanisms to bridge these gaps.
Data collection and use. Currently, we have very little visibility into promotion trends across the department. A few examples include:
- What’s the distribution of races and genders that have been proposed, and approved, for promotion?
- By managers, what is the race and gender of the people proposed for promotion? What is the race and gender of the managers proposing promotions? What are the approval rates by managers?
- What’s the distribution of roles that have been proposed and approved for promotion?
If we could answer these questions, we’d be in a better position to recommend additional changes to the process, aiming to make it fairer and more trusted all around.
I don’t know of any companies where promotions aren’t at least somewhat fraught. Having built a process that appears to be reducing negative effects and increasing trustworthy promotions, I’m hopeful that the combination of career ladders and decision-making with a transparent structure can help many organizations grow more smoothly.