Blog

Alexander Bird Software • Blog

Learning to build useful, valuable software as part of a team.
About

Leaders ought to clarify the problem (and not solve it)

What I do to create an environment where everyone gets to do interesting, valuable work

In my experience, software work can be split into a few broad categories:

In general, folks covet the design work (it's fun, interesting, high impact, and looks great on a resume). There tends to be lots of implementation work. The ambiguous work is easy to neglect ("nobody pulled that work this sprint, again") and easy to excuse ("we started on that thing, but ran into this obstacle so we're blocked again").

I have seen groups of developers (I intentionally avoid the word "team" here) where the most senior members take the interesting design work. They let everyone else do the implementation work, and they do the bare minimum on the ambiguous work (just enough to show that they're making progress on it, not enough to see it through to completion).

I have found that a team is significantly more productive (and has more fun) if the most senior folks abstain from the coveted design work and instead create an environment where junior folks can safely and effectively complete design work. In Netflix's book No Rules Rules they call this "Lead with context, not control".

How a leader can enable inexperienced team members to do interesting design

1. Clarify risky design tasks

For risky design, redefine the problem so it's safe for someone inexperienced to complete.

Usually, it's possible to make a minor change to the problem statement so that any unsafe solutions are obviously invalid. For example: suppose your team mate has no experience evaluating open source licenses but the design task is to choose a library to meet a certain need. You can make it safe for them to do the work by clarifying at the start "the library must comply with our company's open source software policy. I don't know where it's written down, but I or someone else on the team can help you find that if you don't know where to find it". Or, if you're at a small company with no open source policy, then: "we need to be careful to choose a library with an appropriate license. I don't have a full list of acceptable licenses. Any solution must include a justification for why the license is appropriate for our company -- and let's have Jenny the Business Leader review that license justification". With a clarification like that, you can safely go on vacation and trust that the task will be completed safely. You don't need to micromanage the design by reviewing every alternative and helping choose one; you've created an environment where the inexperienced team member can safely make a choice.

2. Automate and optimize the implementation work

In the context of this blog post, implementation work refers to work that has a well defined solution. It's boring for builders to assemble something with a well defined solution. We want to invent. We want to innovate! But that's not just a selfish motivation; boring work is bad for business. If we could design a solution and have it instantaneously implemented, that would be just as valuable to a customer as one that takes months or years to implement. Actually, the one that's instantaneously implemented is more valuable because it's available sooner. So not only is boring work bad for builders (because we don't like being bored), it's bad for customers (because it delays the stuff they want from the business). It's bad for business to have unhappy builders and impatient customers. So nobody is helped by boring work.

Although implementation work is necessary, there's no benefit to having it take so long. Rather than having junior folks handle the boring implementation work, technical leaders should participate in the boring work and find opportunities to automate and optimize the work. If it normally takes two days, is there something simple we can change that will make it take one day? Sometimes these optimizations are architectural or tooling related. Other times, the optimization is related to training: if the junior team member knew how to use a certain technique they could complete the work in half the time. So let's take a few days and learn that technique so everyone can complete the boring work faster.

3. Convert ambiguous work into design work or implementation work

Be the first to take on the frustrating/ambiguous work which we agree is important but which nobody can summarize. This is the work with vague problem statements. Here are some examples:

The default approach that I've seen is to have one of the most experienced people complete this work. However, as soon as the problem is well defined, the less experienced people on the team could take over. Instead of having the technical leader solve the problem, let them define the problem.

In the example above with missing data from the Quux report after 10pm, the problem could be better defined like so:

  1. let the leader draw out a diagram of all the components that are involved in a customer requesting the Quux report and getting the results. They should share this with the team to make sure it's clear and meaningful to any builder who's interested, and to see if other experienced folks have anything to add or correct.
  2. let the leader identify what diagnostic information is available between the different components. Presumably some is missing because several people attempted to find the source of the problem but could not.
  3. the leader then describes the problem: "how can we change our system so that we have enough diagnostic information in our Quux report generation process so that we can pinpoint the source of any data integrity errors.

Now, the ambiguous problem is a desgin problem, and we're looking for someone with little experience who can take it on safely.

In my experience, it's common for experienced folks to do step 3 but not 1 or 2. The less experienced folks are lost because they don't have a map in their heads of how the Quux report is generated, and they don't have the reverse-engineering experience to create the map for themselves. Meanwhile, the experienced person considers this a trivial debugging task because the system diagram exists in their head, and they avoid trivial tasks because they want to do interesting work. This leads to the important but ambiguous task being neglected and postponed.

In the second example with the BazWizard integration, it's a bit different -- the ambiguity is more product ambiguity than technical ambiguity. I've seen folks dismiss this sort of problem with "when The Business has a clear ask of us we can work on it" or "this problem statement doesn't make any sense". Meanwhile, the person requesting the work knows that it's important but they don't know enough to ask for something that the builers will accept. It's easy for everyone to postpone or ignore the work, but there is someone who is convinced that it's important, valuable work to do. Instead of neglecting this ambiguous work, a technical leader can disambiguate it. For this example, a leader might do the following:

  1. Learn about BazWizard. Take a few days playing around with it, reading documentation, watching videos, and talking to anyone nearby who is familiar with it.
  2. Talk to the business leader who asked for the integration. Find out why it's important.
    • for the sake of the example, suppose the business leader clarified that there is a specific workflow that customers would take that would involve the BazWizard integration.
  3. Write a memo describing what they've learned so far:
    • the opportunity that the business leader has identified
    • the specific workflow that relates to BazWizard
    • what even is BazWizard? (Lessons learned from playing around with BazWizard)
  4. Review the memo with the team, and facilitate a discussion around what steps can be taken to create a proof of concept for the BazWizard integration. Generate a list of goals, like:
    • get a dev copy of BazWizard for us to experiment with
    • figure out (via experiment or analysis) what needs to change in the GlurbEngine to integrate with BazWizard

Reiteration: the alternative is stagnation

When the leader defines problems clearly so that inexperienced folks can safely operate outside of their comfort zone, then

  1. inexperienced folks grow quickly
  2. experienced folks are available to tackle the ambiguous work and to automate the boring work

Final thoughts: the biggest challenge is letting go

The hardest part about this approach is that the technical leader does just enough work to be deeply invested in something, and then (just before the really fun design work starts) they let someone else take over. That's a hard thing to do. But, it's an extremely rewarding thing to help create a team where everyone is working on interesting, valuable work.


Something to add? Continue the conversation on Mastodon


Written 2022-12