← Back to all articles
Article 05Product management concepts & tools

Process: a tradeoff between speed and error resilience

Used daily, never thought in depth about

Process: used daily, never thought in depth about

In the life of a product manager, one cannot but stumble upon the mythical concept of process. “Process” is a word used by many in companies, often flaunted around vaguely, usually with connotations of there being too much or too little of it.
Whereas the concept of process is fairly well understood (definition will follow later) what objectives it actually serves are usually not. Unlike many other terms and concepts that I have or will cover, that have conflicting views and factions around them, what struck me regarding process is that rarely people stop and question what it is and what it serves, with everyone agreeing that “some but not too much” is needed. This realization led me to think in more depth about it and to write this article.
What is more scary compared to “something that everyone speaks about though no one agrees on” (the definition of platform for example), is “something that everyone uses daily but no one ponders about”.

Definition and examples

“Process is a set of procedures that defines how a certain output is obtained inside a company. Once something has been codified as a process, the assumption is that it should always be done that way across the whole company, unless there’s another process that states the contrary”

Usually processes include the how, the who, the tools, the how long of the output. Given though the exact definition of process is not really the scope of this article (nor something that people seem to care about), I’ll provide a few examples from my experience before moving on to the usual evolution of process inside companies and then my take on the framework on how to tackle processes from a product perspective.

Examples of processes from my experience:

  • At NVIDIA, there is a specific process to review and approve a PCB board (the piece inside electronics that has all the chips on it) for manufacturing, that included multiple reviews with all the involved functions (electrical engineering, mechanical engineering, thermal engineering, RF engineering) on the design itself to ensure functionality and manufacturing feasibility, review with ops to check for all the parts used and ensure the presence of alternates, compliance review to make sure the board will be able to import / export, review with ops finance on the quotes and a few other steps. This process was in place because the manufacturing of a PCB is very expensive and often there’s no way to know if it is correct until it comes out, so a mistake in the design has huge cost implications (including time implications, since manufacturing takes time)
  • At Mythical and pretty much a lot of other startups, there was an agile process that involves sprint planning and retros. At each sprint planning there were clear steps to review past tickets, close the sprint, review how much was done in the sprint, put tickets in the next sprint and start the next sprint. This was set up in order to make sure that we had a number for “velocity”, that is what we would use to track how much was being done (more on this later)
  • At Lyft we have a process of “vertical reviews” every 6 weeks (twice a quarter), where each “vertical” (a group of teams) presents their progress and plans to the SVP to provide updates on what is happening. This is to ensure exec visibility into each teams’ operations and to make sure they know what each team is up to. The executive assistant sets up the meetings and the teams have to show up and present. There is a template for what has to be covered to ensure consistency

The evolution of processes through the lifecycle of a company

In origin, the startup was created and in the startup, there were no processes.
And that makes sense! When the team is small, usually coordination happens through “mutual coordination”, which in practice means that when you need something approved or need to check in on something you turn around and chat with your colleague. You don’t need to codify processes, since usually they are just informal or taught through example by the leads of the company.

This leads to two important points:

  1. If startups with few people work without processes and (some) are successful, then it means that there is a way to make small teams work without processes!
  2. If that is the case, then why do big companies try to enforce processes even on small teams?

In small teams (usually around 5 - 8 people) a strong and experienced lead is enough to keep the team productive and effective. The lead can keep an eye on everyone, coordinate the work between people and ensure that the output quality is high. The lead steps in when needed though will let the team ideate and execute when they are doing well, keeping delivery speed up.
Similarly though, a not-strong lead will have the exact opposite effect, and that is when usually people start bringing up “processes”.

Usually, a catastrophic error, launch or release happens, so leadership starts requesting “processes” to prevent those from happening again. In the most extreme case, they bring in “experts” in processes to establish processes and keep everyone to them.

The more a company grows without processes, the more these “errors”, big or small, happen, the more the need for processes gets brought up. In some cases they get implemented well and they work, in some cases they get implemented badly and don’t work and in some cases, they don’t get implemented and the situation with errors gets bad to the point of non functionality of the company.

Once a company achieves great product market fit and becomes profitable, the impact of the errors becomes much bigger and this is usually when processes start pervading the company and the day to day life of employees. Why is that? Find out right below!

Process: a tradeoff between speed and error resilience

“Process sets a tradeoff between speed of the company and error resilience”

That’s how I approach processes from a product perspective.
The more processes there are in a company, the slower the company progresses in their product development, though the less errors, mistakes, catastrophes etc happen with their business.
Which explains perfectly the lifecycle presented above:

  1. For a startup which hasn’t found product market fit, speed is everything and if the output breaks often, it doesn’t really matter until the product market fit happens and there is significant customer adoption. So absolutely no process is the way to go. Teams should be small and coordinate between themselves flexibly
  2. As companies grow and find initial product-market fit, speed is important, though it has to be well directed. So processes make sense in specific parts of the company (especially the parts that work and make up the core business). When successful startups grow, teams start to independently aim in different directions, following the direction of the respective team leader (which may and usually does differ from the CEO’s, often NOT by design). That is when some processes that “direct and constrain” the energy of the core business teams make sense, while still allowing some teams to have top speed and autonomy to keep exploring for new product or more business breakthroughs
  3. Once companies start to have a sizable business, then usually the priority becomes to defend the business, not necessarily to innovate or evolve the business. Keeping existing customers is the priority, and the best way to do that is to NOT change anything. Keep the lights on and that’s it! If you just tell a team to “keep everything running” though, it is usually not going to be enough. The ambitious people on the team (who usually are also the best) will want to progress their career and innovate. So they will do changes to the codebase (or the hardware design in case of a hardware product). With a solid process that checks those changes and makes sure that they don’t break anything, the company protects itself from reckless changes that may affect the end user experience badly and influence the company’s reputation. So a company with a great business will introduce processes everywhere to not degrade the existing (and well functioning product).

Adding processes to everything though does slow the company down. In the startup-no-process phase, engineers can work on something and deploy it to production / add it to a design without any delays, immediately moving on to the next thing. In the big-company-stable-business-process-everywhere case instead, even very small modifications might take weeks (or months in the worst case) before they go through the release process and are officially added to the product.

Effect of process on morale and culture

Process has a big effect on culture, and it affects different types of people differently. Usually people can be divided into two groups based on how they react to process:

  1. People that don't like process and that feel that process is put in place to stop them from doing their best work. These are also the type of people that thrive in start ups or at the early stage, though then resent being in bigger companies. Especially if we are talking about engineers, these people are usually high performers, creative thinkers but are not necessarily the best collaborators. They have high output, though a high degree of autonomy, which leads them to work on what they want, not necessarily what the company needs.
  2. People that love processes and that think that all things that don’t work can be solved by processes. To the extreme, these people don’t actually do anything unless there is a process that defines how to do things. These are people who love retrospectives (aka retros), love drafting proposals about how to prevent a bad thing from happening again, write a lot of after the fact incident analyses. For processes that are in place, they follow them to the letter. These people are usually very predictable though I’ve found them to not be the “star” performers of teams.

OK, I'm being a little extreme and not everybody fits squarely into one of the buckets. However, if you think about your colleagues, you can most likely identify to which extreme each person leans more towards.
You can also imagine that no one team works with only one of the kinds of people above. You need a good balance of both and the balance shifts based on what the team main objectives are.
A team made up of people of type (1) will deliver quickly though likely break a lot of things and end up with non consistent output. A team made up of people of type (2) will likely spend most of the day in meeting discussing the best way to do things, without actually getting anything done.

Adding process ends up making people more tending towards the first category less happy and people that tend towards the second category more happy, if the process actually does end up being followed. Processes that are not followed decrease morale across the board. On the first group because they don’t understand why the company is wasting time on processes that no one follows, and on the second group because they’d like to see the process followed but nobody does.

If the people of the first kind start leaving, then people of the second kind will be adding more and more processes, making it hard to come back to a process-less state. Additionally, people of the second kind will hire more people of the second kind.
Having a lot of processes will also teach people to always seek approval, and ultimately removes independent decision making, resulting in slower companies.

Relationship between process and hierarchy

I’ve noticed that the deeper a hierarchy the more processes are in place. There is a specific causal direction in this. The more layers of management there are, the more people are not actually doing work and are just “administering” teams. In order to justify their roles they create processes to:

  1. Look like they are contributing something to the company
  2. Being the person in charge of administering the process, to make sure it is followed

This is why Google in 2025, following ChatGPT’s success, issued a wake up call inside the company, removing a lot of middle management. They had to shift from a slow moving, moat defending business back to innovation. In order to do that they needed less processes and removing middle managers was one of the steps they took to remove the amount of processes in the company. I never worked at Google so I don't know the specifics, however in the six months following the internal code red, Gemini (Google’s alternative to chatGPT) made huge progress and started seriously threatening chatGPT.

Good use case for processes

So, based on this, when should you absolutely have processes in place?

As we mentioned above, processes reduce errors, so the places where you need processes most are those where errors are the most expensive.

1 - Hardware product release to production or testing

In hardware, there are a lot of points during the development process from which there is no turning back. Crossing each of those milestones requires significant expenditures, or the promise of significant expenditures by the company.
After a device has been sent to testing, it can hardly be recovered. After a production batch has started, it can’t be edited. Everything that requires the creation or destruction of hard things is a one way gate. You really don't want to have any mistakes or errors in what you're sending to produce, test or certify. Hence why hardware companies have many processes consisting of many reviews by many teams before each of those gates. Those processes make sense, and the best companies in hardware have usually figured out the best set of processes to keep development fast while still ensuring that the quality of anything being sent to production is there.

The worst part of a production batch gone wrong because of a mistake is the time needed to evaluate, reorder all parts and restart it. Catching an error early doesn't just save money but it also saves development time which is often more scarce and valuable than money.

Even with processes, things still go wrong…

At NVIDIA once we had a first batch of edge developer kits arrive from the factory ready to be sent to distributors. We had followed the whole process, though as one of our marketing people was preparing a demo with it, they discovered that one of the camera connectors had been mounted wrong and didn't work. That was terrible. We figured out that the issue had stemmed from an alternative part to the design being used and the alternative part requiring different assembly instructions, which we had missed during the assembly manual compilation.
As a result we had to send all the units back to China to get reworked. By the time they came back fixed, 4 weeks had passed. Fortunately, we had one batch come in which had used the original part and hence was working and we could start distributing with those. However, the delay in seeding the distribution channel delayed our time to revenue. We amended the process to make sure that alternative parts also had similar assembly instructions.

2 – Anything related to short and specific timeframe sales

For companies that operate on very short time (or quantity) limited sales, having an error in their systems can be fatal and cause irrecuperable losses. Those are usually caused by high load on the system, usually by an overwhelming use by users.
Examples of this kind of errors are:

  • Ticketmaster selling tickets to Taylor Swift’s or Oasis’ tour. The website crashed and kept people waiting, resulting in very bad press. The website though did preserve the dynamic pricing aspect, not impacting revenue of the company
  • At Mythical, when we did blockchain based drops of items. Blockchain is irreversible, so you really don’t want to mess the sale of items on it. Early on we sold a Deadmau5 character, though had a little issue related to system load delaying sales for a few minutes. Everything sold in the end, so there wasn’t material loss
  • Auctions
  • At Lyft, New Years is a huge night where a lot of people go out and use rideshare. If the system fails then, Lyft will most likely will lose a lot of riders that will instead use the competition, since they can’t wait for Lyft to fix the system to get home. One year, there was an issue with surge pricing during New Year that resulted in a few hours of drivers mostly driving for the competition. After that, testing before key dates was made stricter.

For those kind of high visibility, low inventory events, having the system tested and simulated for loads that exceed the expectation can be a life saver. For this reason, putting processes in place that force employees to test before them and that prevent employees from making any changes before them is seen as prudent and good practice, often understood also by the biggest process detractors.

Why process doesn’t solve for slow execution

Another very commonly heard myth at companies is that process will make execution faster and enable teams to deliver projects faster! This train of thought is particularly popular at scale ups and rapid growth startups and usually results in the hiring of “specialists” that come in and set up the whole team with a very thorough project management software and usually agile approach. I’ll write another article on agile in practice as a follow up.

What does this process look like?

  • Usually tracking “tickets” or to do items in a spreadsheet or project management software
  • Detailed descriptions for what each of the tickets / items are needed for
  • Many meetings to discuss the order of the tickets and the relative priority of each
  • Many meetings that look back to the past week / two weeks about how the team progressed

The key reason management usually wants to set a process in place is to “measure” the output. That usually happens through story points or through the number of tickets completed.

All of this is mostly crap and gameable and does NOT speed up the team. At Mythical I worked with one of the people that I consider was the best at tracking all the projects and their statuses, and I don’t think that them being there helped the velocity of the team.

The false hopes of management are that:

  1. It will be clear who is a top contributor in terms of tickets / story points and who isn’t
  2. It will be clear what needs to be changed to get the team to deliver faster
  3. It will be clear how close we are to the deadline of each project

Unfortunately that rarely works.

What actually happens is the following:

  1. Smart and not smart people can and will game the system by writing more tickets and closing more tickets. Everybody looks like is doing a ton of tickets and work, without actually doing the work that matters
  2. People spend more time writing tickets
  3. Management (eng manager, project manager, product manager, designer) spends more time reviewing reprioritizing and debating tickets, while people are waiting around to see what gets decided
  4. Initially the planned work doesn’t get done and sprints have a lot of uncompleted work. Management gets dissatisfied with it. Project manager reacts by creating more retro and planning meetings, which take up more of everyone’s time
  5. In order to hit timelines, less tickets and work is scheduled per time frame (one or two weeks usually), extending timelines with fake buffer. If you give buffer, people feel less pressure and their work fills the space, resulting in lower velocity.
  6. Overall project timelines are extended to account for “team capacity” which is calculated based on the buffer extended timelines
  7. Management asks for more people to do the job, hiring more people or transferring them from other teams, which increases team size, increases collaboration burden and usually requires more project managers
  8. Repeat

Which leads us to the possible exit scenarios

  1. If company did not have product market fit and not a stable business, the company closes
  2. If the company had a very solid product and stable business, the product gets stuck in a making money phase, until a disruptor comes along and forces the company to change. This can take years or never happen, so it is not necessarily bad (error resilience!)
  3. If management realizes the issue (usually only in the business challenge at point 2) they either bring a leader in who has dealt with this before or they start doing what is needed. What’s need to be done is usually painful and requires laying off people, putting pressure on the team and letting leaders drive tight knit teams against very aggressive timelines to hit specific shipping milestones

This isn’t to say that process is not helpful. Beyond a certain number of people, some process is needed to ensure the team is working on helpful things. I’ll cover more on how to setup effective team management process in the article on agile.

Standardized processes and qualifications

One case in which you do have to follow processes is for highly regulated industries like car production, medical, financial etc. Regardless of how sensical (or not) the processes are, if you don’t have specific certifications your company will not be able to compete in those sectors, so you should follow whatever process the certification says and make sure to follow it well. Getting those certifications is also not really something that you can decide not to do. If you don’t have them you won’t have a business in those industries.
Following these processes will make you slower, though that is by design. In regulated industries a mistake is usually very costly (unlike in social media for example). If you are a startup you usually want to delay getting those certifications as long as possible, since usually (though depends on a case by case basis) once you obtain the certification your whole company has to adhere to its rules.

The product manager’s role in processes

So we covered all this content on process… though what role does the product manager play in this?
First and foremost, the product manager should be the one deciding how much process should be applied to a specific team, even before getting into the details of the specific process steps. In a company there are usually forces that push for more process and forces that push for less process. Usually they do so without actually understanding the tradeoff pointed out in this article and their arguments come from a more philosophical standpoint.
Process is not a philosophy. It is a tool. And the product manager has to decide how much to use it to pursue the company objectives, conscious of the fact that the tool’s effectiveness depends A LOT on the composition and tendencies of the team. The product manager has to read the situation and decide how much process the team has to follow (and then be the key person enforcing it).

Let’s give a few examples from my experience.

(1 - process was positive)
At Lyft, one of my teams was a platform team made up of very different individuals with a big difference in individual skill levels. They hadn’t had a product manager for a while and it was completely unclear what they were doing; management wanted some answers. One of my first directions (together with the engineering manager) was to try and set up a very stringent process. The speed of delivery of the team wasn’t a concern. We didn’t even understand what the team was doing, so speed of achieving that was really not important. First we had to establish with the team that they couldn’t just work on what they wanted or what random people at the company asked them to, they had to add ideas to the backlog and we would all prioritize them, which could result in ideas of person A being worked on by person B.
This was a very painful transition for the team and I don’t think I was fully successful. After some time though, I can say that the team was 50% of the time working towards important projects and was delivering those at a good enough speed. Adding the process also led to the discovery that some engineers were doing all the work and some others weren’t, though more because the higher performing engineers weren’t leaving them enough space to execute and learn. All things that we could then address. Without an established process of adding tickets to project management software, talking through which ticket everyone was doing, a few meetings a week where to review everything and forced tech spec shareouts we would not have achieved the platform progress that was needed and uncovered the uneven work distributions across people.

(2 - process was negative)
At Mythical, we were pursuing aggressive targets on multiple projects and the team had grown rapidly, so when some deadlines started slipping (I’m not going to go into the details on this here, though it was mostly for continuous reprioritization reasons, a topic I’m going to cover in another article around focus and rapid pivoting) management started getting concerned and hired a project manager. The teams had grown a lot so engineering managers were stretched. On the PM side, I was hiring though it was taking time (see the article on hiring and keeping a high bar), so we needed some help on the organization side.
The project manager that we hired was, out of all the ones that I have ever seen, one of the most organized and well versed in Jira; one of the best I’d say. He set up a very strict project management and agile cadences. Unfortunately, I don’t think that had a great effect on delivery. Suddenly we were spending a lot of time reviewing how tickets were written, discussing how an epic should be divided, identifying the size of story points etc. It was simply too much and it was taking too much time from the team and from myself, given I was PMing a few pods and hence had a lot of meetings just related to project management.
The ultimate goal of this was to understand the “team velocity” and from there decide on what to do to increase it. I can say with certainty that we didn’t learn anything new from the process. The issue was reprioritization, lack of focus on the main project and overhiring. We didn’t need story points to find that out since we already know from the onset. The process added burden to both the team and the engineering / product / project managers.
The team had worked well up to then, delivering aggressively, without that level of process. A little more discipline in ticket writing and breakdown of tasks helped, however going into too much detail on story points, demos and retros did not.

Echoing the earlier part of this article, the team was working on zero to one projects so the only thing that mattered was speed (except for the blockchain part, since that is irreversible, though we were very careful there). We didn't need to make sure we weren't getting things wrong so there wasn't too much of a need for process. If we had delivered faster with some bugs, we would have been able to iterate from there.

Conclusion

I hope that this gives you clarity on how and when to adopt processes in your company and on what role the product manager plays in them!
A helpful follow up to this will be the article on Agile and what I think is the best interpretation of it for software teams.