Transcript
Tina Wright: This session is called, “What Keeps Me Up at Night, Existential Questions for IC Leaders”. I’m Tina. I’ve been in this industry for about 15 years now. In that time, I started out as an idealistic new grad. I’ve progressed towards a somewhat jaded senior engineer, but with better photos at least. Maybe one day I’ll be able to press into principal or beyond. During this time, I’ve stayed as an individual contributor. If you’re thinking about existential questions for an individual contributor, you might think that what I worry about is, how do I do global data replication? Or, how do I do cache invalidations across synchronized services? Those can be really fun problems but they’re not existential questions.
For me, the existential questions are the ones that challenge my sense of identity, my sense of accomplishment, and my sense of belonging. These questions are the ones that haunt me. This is very loosely Halloween-themed. We’re going to go through these zombies that rise from the dead, over and over. We’re going to talk about the vampires that drain your confidence. We’re going to talk about the ghosts that haunt you in your weakest moments. We’re also going to not just talk about the problems, but some concrete ways that I, and hopefully you, can use to tackle them.
How Do I Have More Impact?
First up, how can I have more impact? If you’re a successful individual contributor, I’m sure that you’ve been rewarded with maybe promotions, maybe larger scope, and definitely more work. You might look at your peers. You might look at your leaders and think, these people spend a lot of time at work. They’re on-call, nights and weekends. This is really their entire identity. I’m not sure I want to be like that. How do I have more impact without spending more hours? Because it doesn’t matter how dedicated you are, if you want 10 times the impact, you don’t have 400 hours in a week. We have to start thinking differently. We want to work better, not longer. Let’s look at three different techniques for changing the way we work to be more effective. The theme through these is going to be utilizing other people and looking at collective activities that we can do. The first is delegation, which I’m sure you’ve all heard of. Second is distribution across your organization.
The third is elimination of deadweight. Delegation, you’re probably thinking, what could be talked about delegation? I take my work and I dump it on some other poor sucker. You could definitely do that, but your org probably won’t love you very much. When I think about delegation, the first thing I like to think about is matching the opportunity to the delegate. Find somebody for whom this task that you don’t want to do because you’ve done it 16 times before is an opportunity for them to learn and grow. This might be that they have a career aspiration that they’re not currently able to fulfill. How can you take the work that you’re doing and make it something that for them is exciting?
In order to do this sort of matching, you need to know about their career aspirations. For me, I do that through quarterly one-on-ones with my senior engineers. Sure, we’ll talk about the projects they’re working on. Maybe we’ll talk about problems I can help unblock them. I like to also talk about what they’re not working on, what they would like to be, because that’s where the gold is in terms of finding matches in the future.
We’ve matched the opportunity to the delegate. It’s a bit of a stretch for them, but not too much. Next up, we’re going to mentor them through the challenges. Because you were assigned that work presumably because people knew you could do it. Who better to guide them through it than somebody who’s done this before, who’s reliable for it? This is your way of giving back to them. You’re asking them to take their time to do a task for you. The way you give back to them is that you help them through their learning opportunity here. While you’re mentoring them through the challenges, you will be involved in this project.
One thing you want to be really careful of is not to overshadow them. To do that, you need to give them autonomy. You need to give them the space to operate so that they know what they can do without your intervention. For instance, you might say something like, I’d like you to design this project, work with these three people, and execute on it. However, once you start to hit this boundary of talking to this team or that team, I’d like to be involved again. I won’t monitor your execution very much, but if you’re more than two weeks late, I’m going to come in. This gives them a sense of where is the scope that they can control, and what is the scope that you will control? This is a really important part of not undermining your delegate and not micromanaging your delegate.
The last piece here is for you, not them, which is that you need to learn to accept good enough. Many people who rise through the ranks are quite good at what they do, for a good reason. When you’re delegating to somebody, this is going to be a stretch for them. They’re not going to do it the same way you’re going to do it. They’re a different person who makes different decisions. That different might not be worse, but it also might be worse. They will probably take longer to do this. They may deliver it at a lower quality than you were. That’s ok if it’s good enough. This is the cost that we’re paying in order to train our organization. You have to accept that cost. Let it go. Breathe in real deep and say, “That’s fine, this is good enough”. That’s delegation and some tips for how to get through that.
Next up, we’re going to distribute our work, which is essentially just delegation to lots of people. There’s a couple more things that come into play when you start to distribute your work across multiple people, teams, organizations.
The first is that you’re going to need to spend some time upfront. In that time upfront, you’re going to need to define your parallelizable and your dependent tasks. Your parallelizable tasks are the ones that you’ll be able to throw five engineers on and they can independently operate through those. You’ll also end up with dependent tasks where A must come before B. You have to have a clear sense of which tasks fall into what bucket and what your dependency graph is. Once you have that, you can then determine milestones. Everyone’s favorite word. Within a milestone is a great place to put all those parallelizable tasks. Between milestones is a fantastic place to put all your dependencies because they’re natural checkpoints in the project for which you can synchronize. With each milestone, it’s really crucial that you define success or your exit criteria. When will you move on from one milestone to the next?
One of the things that I really like to do for each of these milestones is to create some tracking mechanism. Some dashboard that people can look at and see the progress. While this can be a Jira epic or tickets, I actually encourage you to think about what tracking mechanisms can you do that are actually at the root of the thing that you’re trying to solve. For instance, maybe you’ll have a database burndown chart where you track all of the databases that you need to migrate from one system to another. Because that ties it down to what the actual value is we’re bringing as opposed to the activities that we’re doing for that.
Once we’ve created the milestones and the dashboards, we’ve done our parallelization, our dependencies, now we actually get to do the work. You’re not going to be doing the work, they’re going to be doing the work. What do you do? You coordinate the working group. This is something that maybe a program manager might do, but also, I highly encourage you to lean in as a technical person because, frequently, not all the decisions have been made yet. By being active in the coordination of the working group, you can guide those decisions in the direction that you want. This could look like biweekly meetings. It could look like, like I said, the Jira tracking. Whatever it is, what you’re trying to make sure is that the working group is collaborating effectively and that when a decision is made, it is rapidly communicated to the other portions of that working group. While your working group is incredibly important, it’s also really important to communicate broadly to your peers and leadership.
The reason this is so important is that when you are distributing work, you’re going to be taking a non-trivial portion of your organization’s resources towards your project and your approach. Your leadership at some point is going to go, how are you using those? Are you using them effectively? You can preempt this by communicating broadly the things that they need to know. Where are you in your milestones? Are you on track? Personally, I really like to do this through a newsletter that no one ever reads, but they could. That’s the important thing. They have the accountability mechanism that they could check you. After the first one or two, they usually just trust you, but once you’ve built up that muscle, now you have the trust of your leadership and you’ve shown that you are accountable, that you’re responsible. That’s distribution. We’ve dropped our work onto a senior engineer. We’ve distributed it onto a team.
The third thing I want to talk to you about is elimination of deadweight. You probably all think, yes, I’ve seen the to-do list. I’ve seen the efficiency self-help books. I know about deadweight. I know about efficiency. That’s great. Save yourself 10 minutes. I encourage you to do things like that. Today I want you to think about, how can you go beyond your personal efficiency and instead think about the efficiency of your engineering organization. We’re going to shift to collective efficiency. This might be finding a process that’s not working and streamlining or eliminating it. This could be finding a tool and making it faster, like your build or test system. It could be things like lowering your maintenance costs by going to managed solutions and spending money on it instead. The theme of this is that you’re not worried about your own efficiency in these cases. You want multiplicative efficiency of your entire organization. Similarly, focus. I’m sure you’ve all been told to focus before.
Again, we’re going to think not just of our own focus and the things that we want to do personally, but what kind of focus can we provide to the entire engineering organization? As an IC, you might not define the OKRs. You might not define the product direction. What you can define is the technical strategy and the pieces that are absolutely required to be able to get you the next product capability or a faster engineering organization. You can come up with a focused list of the things that you need people to be working on.
More importantly, you can defer everything else and accept that they won’t be worked on, and encourage people not to work on that. This is hard, because all of those things you’re deferring, they probably have value. They’re probably legitimately good things to do. An engineer probably is coming to you saying they want to do them. Are they high enough value to meet your focus? If they’re not, you need to sacrifice them.
Finally, we’re going to look to eliminate work theater. I’m on a stage right now, feels very theatrical. Work theater is the things that you do throughout your day that look like work, but aren’t. They don’t really, at the end of the day, have that much impact. This could be for you going into the office if you’re more effective at home. My favorite one is checking Slack messages that I don’t really need to know about. I’ll say that the most dangerous work theater is the theater you’re putting on for yourself. It’s the ones where you’re convincing yourself that you’re doing a full day’s work when you’re not really.
Instead, you’re avoiding the ambiguous problem you’re supposed to be solving, the challenge that you’re just stuck on. Be really careful when you see yourself doing activities that are work theater, and try to figure out why you’re doing it, and address the root cause of your productive procrastination. Those are three tips for us to work differently. We want to work better, not longer, in order to preserve our sanity, while allowing us to continue to progress and have more impact in our organization.
Does What I Do Even Matter?
The next question, does what I do around here even matter? When I delegate to somebody else and they succeed in their project, one thing that you might ask yourself when you’re feeling particularly down is, would that project have succeeded anyway? Delegate was pretty good, I trusted them. Maybe I didn’t need to be there at all, and it would have been just fine. How much of their success can I attribute to my mentorship or my help? I no longer directly control the output of the work that I’m doing, I’m giving it to somebody else. How then do I know if what I’m doing is mattering versus what they’re doing? This is really dangerous because you might have a hypothesis, but if you don’t have evidence, then when you’re feeling insecure, your confidence will go away as well. We’re going to look at ways to find evidence. In fact, we’re going to treat this scientifically. I have a hypothesis that at work, what I do matters positively.
More specifically, let’s go through an example. We’re going to use an example of, I hypothesize that I can improve our incident response at the company. I encourage you to think of an example for yourself. Is there something in your work that you’re doing that you’re not sure is really landing or working? For that, keep that example in mind as we go through this whole section. We have a hypothesis, I can improve the incident response at my company. The next thing that we’re going to do is run experiments against that hypothesis to validate it. Those experiments have to be falsifiable, which is to say we have to actually be able to prove them true or false. Those experiments should be divided into small pieces so that we can actually run them within our lifetimes. We have to be able to repeat those experiments because there’ll be variation in the system and we want to eliminate that or reduce it.
Let’s start with falsification. Falsification is the concept that something can be proven true or false. If you’ve heard of string theory, this is the problem with this theory, and it’s why it’s still called theory, not fact, because it has yet to produce a single experiment that is falsifiable. We don’t want to be like string theory, we want to be falsifiable. How do we create falsifiable experiments? We’re going to start by being very clear about what our output signals that we’re trying to affect are. In the example I was giving, improved incident response. Let’s go even more concrete though. What am I trying to do? I’m going to find the specific measurable proxies that give me a sense of what my signal is because I cannot directly measure my signal. You can’t see a black hole, but we can measure the stars that orbit around it and know that it has influence.
The same way you are the black hole in your organization and you’re hoping to find what your stars are that are showing that you matter. Those proxies for incident response could be something like the number of dashboards that we have for a service. Let’s say we have at least one dashboard per service, or it could be the time it takes to fully resolve an incident, or it could be the quality of the incident report afterwards. All of these are measurable proxies that give us a sense of how mature our organization is in terms of incident response. We know where we’re trying to affect. We know how we’re going to look at it. Now we have to consider what we do that could affect that signal, our controllable inputs. This is just another name for your work. The controllable inputs in this case formulate our experiment.
In the case of incident response, I could run a seminar on how to do good incident response. I could work with people to try and improve the response reports. I could help create automated dashboards so that services don’t have to do this one at a time manually. These are my controllable inputs that I’m going to test to see if they have an effect on the output signal. I’m going to take these three pieces and I’m going to close that feedback loop by evaluating it. In this example, I could say, at the end of the quarter, I’m going to look and see how long did it take us to have our first call to action for an incident. I’ll do some activities that I think might affect that. Then at the end, I’ll take a look and say, how was it? Did I improve it or not?
Now I gave an example of a quarter, but it turns out humans aren’t very good at feedback loops that are long. Ideally, one or two seconds is how we learn best. It’s a pretty big gap. We’re going to try and shrink it. We’re not going to get down to one or two seconds, but we can probably get to maybe days, or hours, weeks. The idea is to get smaller. How are we going to get smaller? Two ways. First way is that we’re going to take a look at the thing that we want to do, and we’re going to split it into independent component skills. Think of this like cross-training.
If I want to be a 100-meter butterfly swimmer in the Olympics, I don’t just swim 100 meters once every 4 years and see how I do. I don’t just swim 100 meters every day and see how I do. I will also identify different things that I can train independently. I might be able to fix my response time. I might lift weights to change some musculature. I might work on my flexibility. These are independent components that contribute to the skill that I’m looking to build that I can train separately. That’s one way we’re dividing it. Once we’ve gotten a component skill, though, we’re going to divide it yet again.
The way we’re going to do that is that we’re going to divide it into smaller steps that we want to do. Take the next actionable step for a skill. Let’s go back to our incident. Anybody have a favorite sort, do you want to do the dashboards, the incident itself, or the post-incident? Dashboards. What is the next achievable step for our dashboards? We can start by taking a look at where they are now. We can potentially unify those dashboards. We can spread that unification to one more service. We can train people into using the unified dashboards and services. All of these are narrower, actionable steps along the progression of that skill. For each of them, we’ll be able to do some actions and run some experiments.
The nice thing about looking for the next achievable step is it makes it reachable. It’s nearby. You can conceive of it. It’s very easy to then construct your experiment because you’ve anchored it so well. One of the problems with trying to grow skills and move into new places that people come to me with is they’re like, “Tina, I really want to be doing x, but no one’s given me the opportunity yet”. That’s a real problem, because who would give you the opportunity if you haven’t proven yet that you can even do it? This is the way that we find opportunities in our current position to be able to practice those opportunities. I didn’t start speaking and practicing my speaking by coming here today. I first went to my team meeting and tried not to stutter. That for me was my next achievable step when I was very far away from being able to give talks in front of people. For you within your current role, there are achievable steps for each skill that you want to grow.
We’ve divided our experiments in two ways. We’ve split them into independent skills that can be trained, and for each skill, we’ve gotten a training regime or progression. That’s really important because now we’re going to have to repeat some experiments in order to tell what our impact is. Why do we need to do that? Because we’ve introduced unavoidable variance into the system. Our hypothesis is that our input has positive effects on our output at work. In the middle, we’ve introduced some confounding factors. Let’s look back at part one. Our delegate is a human being. They may have motivations. They may have variance in the way that they operate, good days and bad days. They may not have the skills required. All of that are things that are outside of your direct control. You didn’t do anything to make those varieties happen.
Similarly, in delegation, we’ve added complexity to the system. We’ve added more failure points that could go awry, and none of that has to do with the work that you’re doing. This has all introduced noise into our signal. Lucky for us, we have the math to try and get around noise and signals, and that’s aggregation. We can aggregate our results into a distribution. We look at doing several repeated experiments that are small so that we can observe several different cases of it happening and then aggregate those results into a distribution for which you can then take an average. For example, if I wanted to improve the dashboards, I could write a template and share that template with several teams, and have them implement those dashboards.
Then I look at them and see, how’d they do? Did they mess up? Were they close enough? Some teams will do a really great job and some teams will do a poor job. That’s differences you’ll know are the variations that you don’t control. If you look on the average, has it gone up overall, or stayed the same, or maybe gotten worse? We’re going to aggregate our results through many repetitions. In order to do this effectively, we need to consistently apply our inputs, what we do.
In this case, I need to share the same template with them and I need to give them the same guidance because otherwise I’ve introduced even more variance in the system. You might say, if I tried to run a project twice, I think my manager would kill me and I don’t have a time machine. I can’t go back and do Q1 again. How am I going to consistently do my job and measure the outputs through differences? We are all living a series of one-shot experiments in our lives that we never get to repeat. We’re going to just accept that similar enough is good enough. Yes, I’m not going to do it again, but I can do it for different people, different teams, and call it close enough, and accept that I’m never getting a perfect signal here.
How do we determine if we have impact when we’re no longer directly controlling and contributing work? We’re going to hypothesize and run experiments on ourselves and our work, and intentionally observe the outputs. I liken this to listening for birdsong. You might have to go to a new place. You might have to learn to listen to different signals. You might have to be very patient. In the end, you’ll hear the birds singing. In the same way, you might need to learn to look in different places for your validation. It might be hard and it might take time. In the end, there are signals of whether or not you’re effective in your organization, and it’s worthwhile to try to learn how to listen for them.
How Do I Influence Without Authority?
Third question, how do I influence without authority? Yes, get this one a lot. If you looked at part one and you said, I’d love to delegate, but why would anybody listen to me? Or I’d love to distribute this, but I don’t have the authority to do that. How is this ever going to work for me? Am I just limited to not scaling in those ways? That’s hard. You might think, I’ll go to my management ladder. They have the authority to tell people what to do. Maybe if I just convince them, then they can just command it to be so. They can, kind of. The problem is I don’t really believe that authority is enough. Authority gets you the minimum, maybe a little bit more than the minimum.
In the end, even managers, even people with authority, they have to intentionally motivate their teams to work effectively together. Here we’re wondering, how do I influence without authority? Why would anybody listen? I’m going to shift the question a little bit. I think we should be asking, how do I gain respect? Because people who respect you will want to work with you and will want to have the things that you want happen. We’re going to look at three specific types of respect, although I acknowledge there’s probably more, but these are ones that are near and dear to my heart. First is the respect of a role model. This is somebody who you look up to and want to be like. The next is the respect of a conductor of an orchestra who aggregates a number of people to contribute to something where they can be better as a whole than they could individually. The last is of a gardener, somebody who sets the environment that allows people or plants to thrive.
I want to start with role model because this is the one that you can do without really thinking about it, at least to begin with. Because you start off by demonstrating an enviable skill. This is just your work. You code faster than somebody else. You build a cleaner system than someone. You do something that others admire, and they see a little bit of it in themselves. They want to be like that. They need alignment because, for instance, if I started juggling, that’s not exactly the skill set that you might admire. I need to have a similarity in that.
The second thing that I’ll need is I need to have it be observable. People have to see that I’m actually good at this. Otherwise, whatever, don’t know. I encourage you then to work in public. Find ways to have observable work. It’s so easy these days to put everything into DMs or just to do things under the hood because we’re working remotely sometimes. If you can find ways to intentionally demonstrate your work publicly for the organization, you can set yourself as a role model, and people can see the ways that you do things and maybe hope to imitate it, and thus gaining you respect. You can even sometimes make up this. Maybe you do something that you normally wouldn’t do just for the sake of gaining role model respect.
This last thing relies on you to have a little bit of respect to start with, but you can also expose hidden challenges to a select group of people, say a mentee of yours. Not all work can be done in public. It turns out if an eng manager decided that they wanted to demonstrate their firing abilities in public, that would be bad. There are things that we are not going to do in public. Maybe in the case of an IC, I don’t necessarily want to expose my political climate to the rest of the engineering org and admit that that is even there. For a trusted individual, you can expose to them the hidden challenges of the role that when they reach it, they will have to contend with, and therefore deepen the trust and respect that they have in you. Most people start off as a role model, but, over time, we can look at other forms of respect as well.
The next one I want to talk about is a conductor. I really love this analogy because a conductor auditions their musicians. They select the work to be played. They organize rehearsals. Then they coordinate the execution of it in the concert. We do the same things. We interview. We set OKRs. We execute on projects. Very similar. Because both of them are fundamentally figuring out how to get people to work as a collective. In order to do that, you need the people to be willing to lend their autonomy to you. You are now controlling some of their time, maybe some of their career. Then, it’s important to them to know that you have good judgment. The way that you gain this respect is you demonstrate good judgment through your decisions. I am not going to talk about how to have good judgment. That would take a while. Let’s assume you have good judgment and instead go into demonstration of it. To start with, you need to have a really clear articulation of your decisions. You have just entered into a really big game of telephone.
People are only half listening to you at the best and most times, and so you need to be super clear about what your decision was and what the outcome was. Once people can understand what your decision is, the next question is, how did you make that decision? You can show your work by being transparent. You could say, I looked at these inputs, I considered this with this balance, and as a result came to this decision. This is the way by which they know that you have a good process for applying tenets, which is our next thing. The next thing people need to know is how your current decisions may correlate with your future decisions. You can do this by being principled in how you make your decisions. If they trust the underlying principle, then they will trust the resulting decisions.
Finally, at least from the things I could think of, is being consistent, knowing that you will consistently apply your principles to each decision, and that you’re not just some loose cannon that’s going to suddenly be like, let’s go do that. Because they need to trust you. They’re putting their goodwill in your hands. You need, in return, to provide them opportunity. You need to give them something back, because no relationship is one way, at least no healthy relationship is one way. What’s your end of the contract look like? Frequently, your end of the contract here looks like an opportunity, an opportunity to grow, an opportunity to be part of something successful, or maybe even just the opportunity to have a stable job that they know that they can do. It varies based on the person, but for every person, there should be some amount of give and take.
We’ve conducted our orchestra. It sounds very nice. Next, let’s talk about a garden. Because if you were looking closely, I didn’t say anything about a conductor needing to be nice or pleasant. Just needs to have an opportunity and good judgment. A gardener is more about the things that we think about when we think about humans. A gardener defines the environment for their plants to grow. In the same way, you define the values for your organization that will nurture the types of people that you want to have in your organization. Just as there are different environments that are valid for gardening, you could have arid cacti and you could have wet tropical plants. They’re totally valid. If you want a tropical plant, you shouldn’t nurture an arid environment. Once we’ve defined our values and what we would like our organization to look like, we then need to reinforce it. We’re going to have to weed out the behaviors that we don’t like, the things that are drawing us away from our values.
Then we’re going to reward consistently the things that are aligned with our values. By this we maintain the environment, because otherwise it just slowly degrades. This is the point at which you start to see if you’re growing prickly cacti but you think your environment is tropical, you might want to reconsider how you’re reinforcing things. Are you actually rewarding the wrong behaviors? I’ve been talking a lot about collectives, but I also want to acknowledge here that each person is also an individual and a human. The way that we can guide our individuals to reach their potential is through pruning. This is the activity of helping an individual become the best version of themselves. Just like a plant sometimes tries to grow too many leaves instead of fruit, you can prune to help direct it in a certain way.
The final thing I’ll note about gardening is that when you transplant a seedling, it experiences transplant shock and takes some time to reestablish its roots. It can still grow to be a healthy plant, but for time being there will be a setback. People are like this too. Once people establish roots, moving them around becomes costly. I personally don’t believe that people are fully transferable, no matter how flexible you are. I think that in order to get the best out of people, you need to give them time to establish their roots.
I’ll leave the respect section on the note, which is, the reason that I’m thinking about this a lot lately is that I found myself losing role model respect and struggling with how to deal with that personally. It was really a hit to my ego. I’m not doing the kind of activities that I used to do. I’m not doing the kind of activities that my new grad self admired. Am I just becoming a worst-end engineer? Maybe. If you’re like new grad me and have very strong tenets, you might say yes. That’s ok. I’ve come to terms with it. The way that I’ve come to terms with it is that now I think about the other types of respect and shifting into those forms as I lose role model respect because I’m no longer aligning as strongly with some of the engineers in my organization on that level. I can manufacture work to do it, but what I’d rather do instead is lean into these other forms of respect because they can get me further. They can let me work differently and have influence.
Recap
In summary, we went through three questions today. How do I have more impact? Through working differently. How do I determine if what I do matters? Create experiments that you can verify. How do I influence without authority? Cultivate respect in your organization. I hope that this helps you to battle those monsters in your mind a little bit, but remember that these will keep coming back over and over because they are challenging questions. They are existential. They will always be there.
Questions and Answers
Participant: I was wondering while you were talking that, for example, you spoke about delegation and distribute, as well as for respect model and conductor. I’m highlighting those four as trying to understand what are the lines or responsibilities between a role as an engineering manager versus a principal engineer or a staff engineer. How those lines are blurry or not blurry, or how do we draw those lines? Where do we draw those lines?
Tina Wright: I like that you mentioned blurry because I’m not particularly interested in drawing those lines. I think that it’s just a distribution of how much of one versus another. An engineering manager should be more of a conductor and a gardener and probably will be less of a role model because the things that they are spending their time on are not the things that engineers might want to spend their time on. That’s ok. The way that I personally think about it is there’s no hard boundary here. I don’t think that just because you’re running a project you can’t set values, and just because you’re setting values you can’t be a role model. I think there’s pieces of each of them. I would encourage you to think about for you what’s important, and lean into that.
See more presentations with transcripts
