Joël has a fascinating discovery! He learned a new nuance around working with dependency graphs. Stephanie just finished playing a 100-hour video game on Nintendo Switch: a Japanese role-playing game called Octopath Traveler II. On the work front, she is struggling with a lot of churn in acceptance criteria and ideas about how features should work.
- How do these get documented?
- What happens when they change?
- What happens when people lose this context over time?
JOËL: You're the one who controls the pacing here.
STEPHANIE: Oh, I am. Okay, great.
Hello and welcome to another episode of The Bike Shed, a weekly podcast from your friends at thoughtbot about developing great software. I'm Stephanie Minn.
JOËL: And I'm Joël Quenneville. And together, we're here to share a bit of what we've learned along the way.
STEPHANIE: So, Joël, what's new in your world?
JOËL: So long-time Bike Shed listeners will know that I'm a huge fan of dependency graphs for modeling all sorts of problems and particularly when trying to figure out how to work in an iterative fashion where you can do a bunch of small chunks of work that are independent, that can be shipped one at a time without having your software be in a breaking state in all of these intermediate steps. And I recently made a really exciting discovery, or I learned a new nuance around working with dependency graphs.
So the idea is that if you have a series of entities that have dependencies on each other, so maybe you're trying to build, let's say, some kind of object model or maybe a series of database tables that will reference each other, that kind of thing, if you draw a dependency graph where each bubble on your graph points to other bubbles that it depends on, that means that it can't be created without those other things already existing. Then, in order to create all of those entities for the first time, let's say they're database tables, you need to work your way from kind of the outside in.
You start with any bubbles on your graph that have no arrows going out from them. That means they have no dependencies. They can be safely built on their own, and then you kind of work your way backwards up the arrows. And that's how I've sort of thought about working with dependency graphs for a long time.
Recently, I've been doing some work that involves deleting entities in such a graph. So, again, let's say we're talking about database tables. What I came to realize is that deleting works in the opposite order. So, if you have a table that have other tables that depend on it, but it doesn't depend on anything, that's the first one you want to create. But it's also the last one you want to delete. So, when you're deleting, you want to start with the table that maybe has dependencies on other tables, but no other tables depend on it. It is going to be kind of like the root node of your dependency graph.
So I guess the short guideline here is when you're creating, work from the bottom up or work from the leaves inward, and when you're deleting, work from the top-down or work from the root outward or roots because a graph can have multiple roots; it's not a tree.
STEPHANIE: That is interesting. I'm wondering, did you have a mental model for managing deleting of dependencies prior?
JOËL: No. I've always worked with creating new things. And I went into this task thinking that deleting would be just like creating and then was like, wait a minute, that doesn't work. And then, you know, a few cycles later, realized, oh, wait, deleting is the opposite of creating when you're navigating the graph. And, all of a sudden, I feel like I've got a much clearer mental model or just another way of thinking about how to work with something like this.
STEPHANIE: Cool. That actually got me thinking about a case where you might have a circular dependency. Is that something you've considered yet?
JOËL: Yes. So, when you have a dependency graph, and you've got a circular dependency, that's a big problem because...so, in the creating model, there is no leaf node, if you will, because they both reference each other. So that means that each of these entities cannot be created on its own, the entire cycle. And maybe you've got only two, but maybe your cycle is, you know, ten entities big. The entire cycle is going to be shipped as one massive change.
So something that I often try to do is if I draw a dependency graph out and notice, wait a minute, I do have cyclical dependencies, the question then becomes, can I break that cycle to allow myself to work iteratively? Because otherwise, I know that there's a big chunk that can't be done iteratively. It just has to be done all at once.
STEPHANIE: Yeah, that's really interesting because I've certainly been in that situation where I don't realize until it's too late, where I've started going down the path thinking that, you know, I could just remove this one thing, or make this one change, and then find myself suddenly, you know, coming to the realization, oh, this other thing is now going to have to change.
And then, at that point, there's almost kind of like the sunken cost fallacy [laughs] a little bit where you're like, well, I'm already in it. So, why don't I keep going? But your strategy of trying to find a way to break that cyclica...that is two words combined. [laughs] I meant to say circular dependency [laughs] is the right way to avoid just having to do it all in one go. Have you had to break up a cycle like that before?
JOËL: Yes. I do it on a semi-frequent basis. The fancy term here for what I'm looking for when I'm building out a dependency graph is a directed acyclic graph. That's a graph theory or a computer science term that you'll hear thrown around a lot, DAG. I often like to...when building out a series of tasks that might also form a graph because you don't just model entities in your system; you might model a series of tasks as a graph.
If there's a cycle in the graph, typically, I can break that using something like the strangler fig pattern, which is a way to kind of have some intermediate steps that are non-breaking that then lead you to the refactor that you want. And I've used the strangler fig pattern for a long time, never realizing until later that, oh, what I'm actually doing is breaking cycles in my task dependency graph.
STEPHANIE: Hmm. I'm curious if you have noticed how these cycles come to be because I almost imagine that they get introduced over time, where you maybe did start with a parent and then you, you know, had dependencies. But then, over time, somehow, that circular dependency gets introduced. And I'm wondering if part of figuring out how to break that cycle is determining how things were introduced, like, over time.
JOËL: In my experience, this happens in a lot of different ways because I'm using dependency graphs like this to give myself a mental model for a lot of different kinds of things. So maybe I'm thinking in terms of database tables. And so those might get a circular dependency that gets added over time as the system grows.
But I'm also using it sometimes to model maybe a series of tasks. So I take a large task, and I break it down into subtasks that are all connected to each other. And that doesn't tend to sort of evolve over time in the same way that a series of database tables do. So I think it's very context-dependent. But there are definitely situations where it will be like you said, something that kind of evolves over time.
STEPHANIE: That makes sense. Well, I'm excited for you to get to deleting some potential code or database tables that are no longer in use. That sounds like a developer's dream [laughs] to clean up all that stuff.
JOËL: It's interesting because it's...a move operation is effectively what's happening. So I'm recreating tables in another system, pointing the ActiveRecord to this new system, and then deleting the existing ones in the local database. So, in a sense, I'm kind of traveling up this dependency graph from the leaf nodes into the root and then back down from the root to the leaves as I'm creating and then deleting everything or creating in one system, and then going back and deleting in the other system.
STEPHANIE: Got it. Okay, so not necessarily a net negative but, like you said, a move or just having to gradually replace to use a new system.
JOËL: That's right. And we're trying not to do this as, you know, okay, we're going to take the system down and move 50 tables from one system to another. But instead, saying, like, you know, one at a time, we're going to move these things over. And it's going to be small, incremental change over the course of a couple of weeks. And they're all pretty safe to deploy, and we feel good about them.
STEPHANIE: That's good. I'm glad you feel good. [laughs] We should all be able to feel good when we make changes like that.
JOËL: It's going to make my Fridays just so much more low-key just, like, yeah, hit that deploy button. It's okay.
So, Stephanie, what is new in your world?
STEPHANIE: So this is not work-related at all. But I just finished playing a 100-hour video game on my Nintendo Switch. [laughs] I finished a Japanese role-playing game called Octopath Traveler II. And I have never really played a game like this before. I've not, you know, put in many, many hours into something that then had an end, like, a completion.
So, at the end of this very long game that had a very, you know, compelling and engaging story and I was invested in all of these characters, and by the time the credits were rolling, I felt a little sad to be leaving this world that I have been in many evenings over the last couple of months. Yeah, I don't know, I'm feeling both a little sad because, you know like I said, I got really invested in this game, but now I'm also kind of glad to have some free time back in [laughs] my life because that has definitely been the primary, like, evening activity that I've been doing to relax.
JOËL: It sounds like this game had a very, like, a particularly immersive world that really pulled you in.
STEPHANIE: It did. It did. It has these eight, like, different characters that you follow, like, different chapters and all of their stories, and then they all kind of come together as well. And the world was huge in this game. There were so many little towns to explore. And I didn't realize I was a completionist type. But I found myself running around opening every chest, talking to every NPC, and making sure that I, you know, collected all of my items [chuckles] before moving on.
I also finished all of the side quests, which is, I think, you know, how I managed to put in over 100 hours into it. But yeah, it was very immersive, and I really enjoyed it. I don't know if this will become a norm for me. I know there are some people who are, you know, JRPG diehards and play a lot of these kinds of games, but they're a real, like, time investment for sure.
JOËL: Are there achievements for completing everything?
STEPHANIE: Not that I can tell on the Switch. I do know that, like, on other systems, you can see your progress on having done all of the things there are to do. But I think it's actually kind of better for me to just play [laughs] to just, like, think that I've done it all but not really, like, have something that tells me whether or not I've done it because then I would feel a lot more neurotic, I think, about being able to let it go where I am now. [laughs]
JOËL: Right. If we've got, like, an explicit checklist of things or a progress bar, then it feels like you got to get to all the things.
STEPHANIE: Yeah, exactly. I think there are still, you know, a couple more things that I wrote down on my little checklist of tasks that I would want to do once I feel like I want to come back to the game. But for now, like I said, I watched the credits roll. I teared up a little bit, you know, thinking about and reminiscing on my adventure with these characters, and I'm ready to put it down for a bit.
JOËL: Did I hear correctly that you made a checklist for this game of things you wanted to do?
STEPHANIE: Yes, [laughs] I did.
JOËL: That's amazing. I love that.
STEPHANIE: Yeah, you know, there are just so many things almost kind of like work where I had to, like, break down some of my goals. I wanted to, like, hit a certain level. I wanted to, you know, make sure I defeat these bosses that would help me get to those levels. And yeah, I got very into it. It was definitely a big part of my life for a couple of months.
I got it originally because I needed a game to play on my flight to Asia back when I went to Japan. And I'm like, oh, like, this looks, you know, fun and engaging, and it will distract me for my, you know, over 10-hour flight. Turns out it distracted me for many, many more hours over several months [laughs] since then. But I had a great time. So yeah, that's what's new for me. Again, it's something I'd never really done before. I will say though I am very behind on my reading goal as a result. [chuckles]
JOËL: I feel like this is a classic developer thing to do is, like, use the tools that we're used to in our job and then apply them to other parts of our life. And now it's just like, okay, well, I made a Kanban board to track my progress in this video game. You know, or, in my case, I'm definitely guilty of having drawn a dependency graph for the crafting tree for some video game. So I feel you really strongly there.
STEPHANIE: Yes, I'm nodding heavily in agreement. I think it just scratches the same kind of itch of, you know, achieving, like, little things and then achieving one big thing.
JOËL: So, speaking of places that are nice to have checklists and, like, well-defined requirements, you and I were talking earlier, and you have recently found some frustration around having user stories be defined well on your current project.
STEPHANIE: Yes. So I've been reflecting a little bit about my current project and noticing what I think I might call product smells; I'm not quite sure, just some things I'm seeing in our day-to-day workflow that is getting me thinking. And I'm curious to hear if you've experienced something similar.
But I find myself being tasked with a ticket that is quite vague. And maybe this was written by a product owner, or maybe it was written by another developer. And it is not quite actionable yet, so I have to go through the process of figuring out what I'm really needing to do here.
I think another thing that has been quite frustrating is, you know, maybe we do find out what we want to do. And, like, I'll go back into the ticket, write down the requirements that I gathered, and do the ticket. I'll ship whatever change was required, and then I'll hear back from someone in a meeting or either as a one-off request in Slack. And it'll be like, "Hey, like, actually, you know, we want this to be different." And maybe you previously said that "Oh, the value for something would be 30. But now we found out more information; it should be 20. And so could you, like, make that change?"
And then I'm not really sure what the best way to document a change like that is because it, you know, maybe existed in the previous ticket, but now it has changed. And do I create a new ticket for this, or do I just go ahead and make the code change? Like, who would know this information that we're now carrying about 20 being the value for, let's say, like, days or not meaning something in the code that we're writing?
And I guess I've just been really curious about how to make sure that this doesn't become the norm where a lot of these conversations are just happening, and, you know, the people who happen to be in them know that this change happened. But then later on, someone is asking questions about, like, hey, like, when did this change? Or I expected this to be 30. But is this, you know, behaving as expected?
So that was [laughs] a bit of a nebulous way of describing just, like, this churn that I feel with being the executor of work. But then, like, a lot of these things changing above me or separate from me and figuring out how to manage that.
JOËL: When you were describing this scenario where you've done the work, and then someone's like, "Oh, could we change this value from, like, 30 to 20?" I'm thinking in my mind of the sort of beam that a lot of our designers face where it's like, you know, they have a design. They work on it; they do it. And then show it to a client, and the client is like, "I love this design. But could we just shift this box over, like, one pixel?"
Like, they're, like, tiny, tiny, little changes that are kind of requested for change after you've done, like, this big thing. And, oftentimes, those pile-up. It's like, you shift it one pixel. It's like, oh, actually, you know what? Why don't we do it two pixels? And then it's like never-ending cycles, sometimes of, like, minute little changes.
STEPHANIE: Yeah. But the minute changes really add up into, I think, really different behavior than what you maybe had decided as a team originally. And in the process of changing and evolving, I don't really know where documentation fits in.
I've been working on this project that had a pretty comprehensive product design doc, where they had decided upfront on, you know, how the application is going to behave in many different scenarios. But again, like, that has changed over time. And when I recently had to onboard someone new to this project, you know, we sent over this document, and we're like, yeah, you can, you know, feel free to peruse it. But it's actually quite outdated.
And then, similarly, right now, since the features that I'm working on are going through QA, there's been a lot of back and forth about, I'm seeing this, but the doc said that Y is supposed to happen, and I'm not sure if that's a bug or not. And I or someone else has to respond with that context that we were holding in our head about when that change happened.
JOËL: That's really interesting. And I think it varies a lot based off the size of the organization. In a smaller organization, you're probably doing a lot of the requirements gathering yourself. You're talking to all the stakeholders. You're probably doing the QA yourself, or you're walking somebody else through QA. Versus a large organization, there might be an entirely separate product team, and a separate QA team, and a separate dev team.
And a danger that I've often seen is where all of these teams are just kind of tossing work over the fence. And all you're given is a, you know, a ticket of, like, execute on this. Basically, turn these specs into code. And then you do that, and then you toss it over the fence to the QA team. And they check does the code do these things? And there's so much context that can easily get lost from one step to another. That being said, I think a lot of devs find it frustrating to do some of the requirements gathering work.
How do you feel in general about scoping out a ticket or doing follow-up conversations with the product team about, like, "Hey, your idea for the ticket is this. How do you feel about doing these things? Or what if we cut these things?" Are those conversations that you enjoy having? Is that a fun part of the developer role for you? Or do you kind of wish that, like, somebody else did all of that so that you could, like, go heads down just writing code?
STEPHANIE: I think it depends. That's a great question. Actually, I have so many thoughts in response. So let me try to figure out where I want to go from here.
But I think I used to not like it. I used to be stressed out by it, and sometimes I still am. But when I thought my role was purely executing, to receive a ticket that is a bit vague, you know, I might have been left feeling, like, stuck, like, not knowing where to go from there.
But now that has changed a bit because I received some really helpful feedback from an old manager of mine who was kind of invested in my growth. And she really suggested learning to become more comfortable with ambiguity because that just becomes more and more your job, I think, as you progress in your career. And so now I at least know what information I need to go get and have, you know, strategies for doing so.
And also knowing that it's my job, like, knowing that no one else might be doing it, and it might just be me so that I can therefore get this ticket done. Because, like you said, that problem of throwing the work over the fence to someone else, at some point, that doesn't work because everyone has too much on their plates. And you have to just decide to be the one to seek the information that you need.
JOËL: I think one way that, as developers, we bring a lot of value is that we help to cut through a lot of that ambiguity. I think if we see our role as merely translating a requirements document into code, that's a very simplistic point of view of what a talented developer does. So, like you said, as we grow in our careers, we start dealing with less and less defined things. We often have to start defining the problems that we're given.
And we have to have these conversations with other teams to figure out what exactly we want to do. And maybe better understand why is it that we want to do this thing. What is the purpose of it? How are we going to get there? And my favorite: Do we have to do all of these things to hit the minimum value of this goal? Can I split this into multiple tickets? I love breaking down work. If I can make the ticket smaller, I'm all about that.
STEPHANIE: Yes. I'm well aware. It's interesting about what you said, though, is that, like, yes, that becomes, in some ways, our superpower. But, for me, where the pain comes in is when that's not part of the expectations, where I am maybe tasked with something that is not clear enough, and yet, the time that I need to find that clarity is not given the respect that it, I think, deserves to build a good product because the expectation is that I should already be making progress on this ticket and that it will be delivered soon.
You know, in that situation, I wish I had been in the room earlier. I wish I had been part of the process for developing the product strategy, or even just, like, have come in earlier to be able to ask, you know, why are we building this? And, like, what are some of the limitations on the technical side that we have? Because often, I find that it is a little too...not necessarily too late, but it is quite down the road that we then have to have these conversations, and it doesn't feel good.
JOËL: I think that's one of the powerful things that came out of the agile movement was the idea that you have these cross-functional teams, that you don't have a separate product team, a separate dev team, a separate QA team, a separate design team that are all these isolated islands. But instead, you say, okay, we have a cross-functional team that is working on this aspect of the product. And it will be some product people, some dev people, some designers kind of all working together and communicating with each other. I know, shocking concept.
And even depending on the context, a big idea is that the client or the customer is a part of that team. So, when we at thoughtbot work with a client, especially when they are maybe a smaller client like a startup founder, we make sure that they feel like they are a part of the team. They are involved in various meetings where we decide things. They have input. You know, they're part of that feedback cycle that we build. But that can also be the case for a larger company where your internal stakeholders are kind of built-in to be sort of part of your team.
STEPHANIE: I've seen so many different flavors of trying to do Agile [laughs] that it has lost a little bit of meaning for me these days. And maybe we've incorporated some aspects of it. But then that idea of the tight feedback loops and then a cross-functional team where everyone is communicating that part has gotten a little bit lost, at least on my project. And I imagine that this is common, and our listeners might be finding themselves in a similar situation where things are starting to feel a little more like handing off and a little more like waterfall. [laughs]
I'm curious, though, if you found yourself being requested to make a change from what the original decision was, how would you go about documenting that or not documenting it? Where do you think the best place for that information about how this feature now is supposed to work where should that live?
JOËL: Are you talking about where do we document that a decision was made to change the original requirements of a task?
JOËL: In general, I think that should live on the ticket just because as long as the ticket is live, I think it's good to have all the context on that ticket for whoever's working on it to have access at a glance.
Sometimes it's worth it to say, you know what? We don't want to just keep this ticket live for weeks or maybe months on end. Let's ship this ticket, and create a follow-up to make a change later, especially if it's a change that's less important where it's like, you know what? It would be nice to have if...but, again, like, scope creep is a real danger. And so, again, me with the aggressive breaking up of tickets, I love to say, "That's a great idea. It would make a great change, not part of this ticket." So oftentimes, those changes I will push them into another ticket.
STEPHANIE: That's interesting. What about documentation beyond the current work? So I'm thinking about once, you know, a feature is delivered, how do people in the organization then know how this feature is supposed to work? Like moving forward as something that is customer-facing.
JOËL: That can vary a lot by organization, I think because there's a couple of different aspects to this. You have maybe some internal-facing documentation; maybe some customer support people need to know about the way the interface has changed. And then you also have customer-facing documentation where maybe you want some sort of, you know, you want a blog post talking about the new feature or some kind of release notes or something like that to be shared with your customers. And compiling that might look very different than what you do for your internal service reps.
STEPHANIE: Yeah, I like that. It's true that the customer documentation is really helpful. At least for, the product that I'm working on, it has very comprehensive documentation about how to use that for its customers. And that has been really helpful because, hopefully, that should be the truest [laughs] information out there.
But sometimes, you know, I find myself in meetings where none of us really know what happens. For example, a question that was asked recently is our product has a free trial capability. But it was unclear what happens to all of the data that the customer is getting access to as a feature. Like, what happens to that data after the free trial ends? Like, if they then have purchased a license, do they still have access to their free trial data? If, you know, there's a lapse between then, does it just get deleted, or will it show up again? And no one really knew the answer to that.
And I think that was another area that got my spidey senses tingling a little bit; I think because it reminded me of...there was a definition I read somewhere of legacy code that is basically when the person who has the most context about how a piece of code works and then they leave the company and that institutional knowledge no longer exists, like, that is legacy code. And I almost think that that also applies to product a little bit where a legacy product is something where no one quite knows what is supposed to happen, but it's still being used by users.
JOËL: That's a really fun definition there. I think there's sort of two related questions that are slightly different here, which is, one, how does the code behave? So, what happens when someone's trial period expires? And it's quite possible that no one on the team knows what actually happens when that time expires.
And then the second question is, what should happen when a trial expires? And it's possible, again, that the product team didn't think through any of the edge cases. They only went for the happy path. And so it's possible if that is also fully undefined and no one knows.
STEPHANIE: Yeah, I like that distinction you made a lot because they definitely go hand in hand, where someone realizes that some weird edge case happened, and then suddenly, they're asking those questions. And, you know, we realized, like, oh, like, that just didn't have enough, like, intention or thought behind how it was coded. So, like, it really is; who knows, right? Just whatever seems to happen.
And I think that this actually kind of reminds me of a previous episode we did about empowering other departments in the company because, ultimately, a lot of those questions about, like, how does this work? What happens? Ends up going to a developer who has to go and read the code and report back. And while, you know, we do have that power, it can also be a bit of a curse, I think. [laughs]
JOËL: I think this is an area where, as developers, we're maybe particularly skilled. Because of the work that we do, our brains are kind of wired to think about all of the edge cases, and sometimes they can be really annoying.
But I think there's a lot of value sometimes when maybe the product team comes to us with a maybe somewhat nebulously scoped ticket or a series of tickets for, let's say, a free trial period feature that only goes through the happy path. And then sometimes it's up to us to push back or to follow up and say, "Okay, great. We've got a bunch of tickets for a free trial period. Have you thought about what happens after a trial expires but the person hasn't converted to a paying customer?" And then, oftentimes, the answer is like, "Oh, no, we didn't think about that."
And I think oftentimes, as developers, our job is to kind of, like, seek out a lot of those edge cases. And we have a lot of techniques and methodologies that we use to try to find edge cases, things like test-driven development, various modeling tools that we'll try to use to make sure that we don't just crash or do something bad in our code.
But what should the actual behavior be? That's a conversation that we need to have. And hopefully, that's one that maybe the product team has already had on their own. But oftentimes, the benefit of having that cross-functional team is the ability to kind of have that back and forth and say, "Hey, what about this edge case? Have we thought about that? How do we want that to behave?"
STEPHANIE: Yeah, that actually made me think about the idea of tech debt but almost at a product level, where, hey, it turns out that we have all of these things that we didn't quite think through, and it's now causing problems. But how much do we invest in revisiting it? Because, you know, maybe this feature is several years old, and it was working just okay enough for it to, you know, be valuable. But we're now discovering these things and, you know, like, do we invest in them? Or are we more focused on, you know, coming up with new things and new features for our customers?
JOËL: That's a classic prioritization problem. It also kind of reminds me of the idea of an MVP. What are the actual, like, minimum set of features that you need in order to try out something or to ship something to customers? And, you know, maybe we don't need some special behavior if your trial account doesn't convert. Maybe we're okay [laughs] that you log in, and the app just crashes. Probably not, because we would probably want you to convert to a paying customer at some point. But maybe we're okay if you just get a screen that says, "You have no projects," when, in fact, you did have projects. It's just that you're no longer on the free trial.
Again, for business reasons, probably we want a call to action there that says, "You have five projects. They are not available to you. Please pay to unlock your projects again." That probably converts better. But, again, now that is a business decision. And that becomes a prioritization question that the team as a whole gets to address.
Sometimes it can also be some really fun prioritization things where if you're on a really tight schedule, you might ship some features live knowing that you have a time limit, but you don't have to necessarily ship other things. So let's say you've got a 30-day trial, and maybe you ship that before you've even implemented what the dashboard will look like after your free trial has expired, and that's fine because no one's going to hit that condition for 30 days. So now you've got 30 days to go out and handle that condition.
And maybe that's okay because it allowed you to get to market a little bit faster, allowed you to cut scope, break those tickets, yes, and just move that much faster. But it does require discipline because now you're on the clock. You've got 30 days to fix that edge case or potentially face some unhappy customers.
STEPHANIE: Yeah, I think that's quite a funny way to handle it. It's really ruthless prioritization [laughs] there.
But what you said was very interesting to me because I was thinking about how there is such a focus on new feature development and that being the thing that will attract customers or generate more money. But there is something to be said about investigating some of our old features of our existing system and finding opportunities there. And oftentimes, revisiting them will reduce the amount of pain [chuckles] that, you know, developers feel having to kind of keep track or have an eye on, like, where things are airing out, but then don't have the time to really invest in making it better or making that part of the product better.
JOËL: I think that's a great opportunity then to have a conversation with other parts of the team. Typically, I think you have to convert some of those into more of a business case. So the business people in the company or the product people might not care about the sort of raw metrics that you see as a developer. Oh, we got an exception with a stack trace in this part of our app. What does that even mean?
But if you say, hey, people who signed up for a free trial and then didn't immediately convert within 30 days who want to come back a month later and convert are unable to do so. And we've seen that that's about 10% of the people who signed up for a free trial. Well, now that's an interesting business question.
Are we losing out on potentially 10% of customer acquisition? I'll bet the sales and marketing people care a lot about that. I'll bet the business people care a lot about that. The product people probably care a lot about that. And now we can have a conversation about should we prioritize this thing? Are these metrics that we should improve? Is this a part of our code that's worth investing in?
STEPHANIE: Yeah, I like that because, in some ways, asking those questions about how does it work? Like, that is really an opportunity because then you can find out, and then you can make decisions about whether it's currently providing enough value as is or if there is something hiding under there to leverage.
JOËL: And I think that's one of the other places where, as developers, we provide value to clients is that we can sort of talk both languages. We can talk product language. We can talk business language. But we can also talk code. And so when we see things like that in code, sort of translate that into, like, what are the business impacts of this code change? Which then allows everyone to make the best possible decisions for the mission of the organization that you're a part of.
So we've talked about a variety of sort of patterns and anti-patterns that surround working through some of this churn on a product. I'm curious, Stephanie, for you, what's maybe one concrete thing that you've done recently that you've found has really helped you navigate this and maybe help reduce some of the stress that you feel as you navigate through this?
STEPHANIE: Yeah, I think, for me, one of the worst things is when that discussion is had in a meeting or a [inaudible 35:45] and then is not put anywhere. And so, one thing I've been making sure to do is either asking the person who made the request to write it down, either on the ticket or in Slack. Or I will write it down, you know, I will document the outcomes of what we talked about and putting it in a public space so that people are aware.
I think that small action has been helpful because we hold so much of this in our heads. And I've been finding that it ends up being hard for people to rotate onto different projects and, you know, get onboarded and up to speed effectively because there's so much knowledge and context transfer happening. But even just putting it in a place where maybe it's not relevant to everyone, but at least they see it. And then the next time that they're asked or maybe, like, do come around to working on this, they, like, have some fragment of a memory that they saw something about this. So that has been really helpful.
It actually dovetails really nicely into what we were talking about with opportunities, too, because once it's out there, like, maybe someone else will see it and have an idea about how it could be better or that change not being what they expected, and they can weigh in a little more. So that's what I'm trying to do.
And I think it's also nice to see how often that happens, right? If we're constantly seeing things changing because we have a written record of it, that could be helpful in bringing up and investigating further as to, like, why is this happening? Like, why do we experience this churn? And is that something we want to address?
JOËL: Yeah, because an element that we haven't talked at all about is any sort of feedback cycle or retrospective, where we can talk about these things and having that written trail and saying, "Oh, we changed this decision five times in the past week, like, really churned there." Now maybe that prioritizes it to be an important thing to talk about and to improve for the next cycle.
STEPHANIE: What I feel really strongly about is when, you know, each individual on a team is feeling this pain, but it not being known that it's actually a collective issue. Because maybe these things are happening in one-on-one conversations, and we don't realize that, like, oh, maybe there is something bigger here that we could improve on. And so the more eyes on it there are, the more visible it is, I think, that the easier it is to address.
JOËL: I love that, the power of writing things down. On that note, shall we wrap up?
STEPHANIE: Let's wrap up. Show notes for this episode can be found at bikeshed.fm.
JOËL: This show has been produced and edited by Mandy Moore.
STEPHANIE: If you enjoyed listening, one really easy way to support the show is to leave us a quick rating or even a review in iTunes. It really helps other folks find the show.
JOËL: If you have any feedback for this or any of our other episodes, you can reach us @_bikeshed, or you can reach me @joelquen on Twitter.
STEPHANIE: Or reach both of us at firstname.lastname@example.org via email.
JOËL: Thanks so much for listening to The Bike Shed, and we'll see you next week.
ANNOUNCER: This podcast is brought to you by thoughtbot, your expert strategy, design, development, and product management partner. We bring digital products from idea to success and teach you how because we care. Learn more at thoughtbot.com.Support The Bike Shed