Hardware teams now move three times faster with the same old tools
Hardware products that once took five years to reach market now ship in 12 to 24 months, yet many teams still pass designs around as PDFs, spreadsheets and email threads. Valentina Ratner, co-founder and CEO of AllSpice.io and a former Amazon PM, sees that gap as costly, because a single wrong board revision can mean anything from rework to a new factory. Which software habits can hardware teams borrow, and which ones break?
See how Git, design reviews and early automated tests translate from software into hardware work
Valentina Ratner, co-founder and CEO of AllSpice.io, argues that hardware engineering is following the path software took: first revision control, then collaboration, then automation. She explains why faster prototyping has outgrown legacy hardware tools, how Git-style workflows and design reviews translate across disciplines, and where hardware differs, from visual data to the far higher cost of mistakes.
“And nowadays they have to do things 3x as fast, but they have the same tools to do so than they did when things took longer.”
“The cost of an error in hardware is exponentially higher because we're talking about physical and real world manufacturing.”
What you will learn
Explain why shorter hardware product cycles expose the limits of legacy data management tools.
Map software practices such as pull requests, commits, diffs and CI/CD to their hardware design review equivalents.
Spot error-prone handoffs between electrical, mechanical, firmware and operations teams.
Plan to shift hardware tests from the end of the design process to the beginning.
Weigh the higher cost of hardware mistakes when bringing agile practices into physical product development.
Key takeaways from the speaker
Gain a deeper understanding of the relationship between software and hardware engineering.
See how cross-disciplinary innovation is shaping the hardware industry.
Learn how applying software principles to hardware design will shape the future of engineering.
Details
Level
Beginner
Format
Talk
Technical depth
Introductory technical
For
Mid-career, Senior individual contributor, People manager
Length
17:03
Published
June 2023
Questions this talk answers
Why do hardware teams need software-style workflows now?
Prototyping has become cheaper and faster, so products reach market in 12 to 24 months with many more iterations. Yet teams still use data management systems built for five-year product cycles.
Why do disconnected tools hurt hardware teams more than software teams?
Software tools connect through APIs without information loss, while hardware information gets reduced to PDFs, email threads and spreadsheets. That forces manual work, such as listing design changes by hand, which is slow and error prone.
What is the biggest difference between software and hardware workflows?
Software developers work in text-based code, while hardware developers work in visual 2D and 3D vector drawings. Hardware mistakes are also far more expensive because they involve physical manufacturing.
How does a software pull request translate to hardware?
It becomes a design review, where inline feedback maps to snippets on drawings, commits map to links to locations, and diffs and CI/CD tests run on visual schematics.
What comes next for hardware development?
Automation: setting up tests at the start of the design process so they run continuously, instead of testing only at the end when changes are hard to make.
About this talk
This presentation explores trends from software that have made their way into hardware engineering and vice versa, focusing on the popular version control system, git, and other notable innovations.
We will examine the factors that have driven this cross-pollination of technologies and how they have impacted the engineering landscape. The presentation will also cover aspects that have yet to carry over, including challenges with scaling, compatibility, and design considerations.
Hi, everyone. Welcome. My name is Valentina Ratner. I'll be host for here today, and we're going to talk about software and hardware, two sides of the same coin. To get us started, a little bit about my background. So I'm the co-founder and CEO of AllSpice.io. I have a background in mechanical engineering and also work experience as a PM where when I was at Amazon, before I went back to grad school, and I did a dual degree program at Harvard and amassing computer science and an MBA. That's where my co-founder, our CTO here at AllSpice, and our conversations about the industry and the problems in the space and the opportunity led us to launch the company.
So AllSpice is a state-founded startup. So we're venture-backed. We've been in market about a year and a half, and we focus on applying software principles to Harvard development. It's a B2B company, SaaS model, and we have customers from aerospace to medical devices and a variety of industries in between. A little bit more of what we are doing is we think of the software and the hardware world converging and hardware products becoming software products as well. And hardware companies becoming software companies at the same time, which also means hardware engineers and software engineers becoming more collaborative with each other and sharing a lot of principles and best practices between the two. So the way we see that come together is through revision control, through collaboration, and through automation.
This is inspired because the way we built hardware has changed over time. So traditional hardware design timelines were used to be very long. So maybe a couple of decades ago, taking a product to market will take about five years, and people will do three, four iterations prototypes in between. So a year-long cycle for each of those to launch a product, a hardware product at the end of it.
Recently, what we've seen is the cost and the time required to prototype has gone significantly down, which means engineers and hardware engineering teams are doing more and more product iterations at a higher frequency. So products today take anywhere from 12 to 24 months to get to market, and sometimes even faster than that. And teams do five, seven, eight, nine iterations for prototypes in between those two.
The problem we see is that a lot of these iterations are still happening in legacy or obsolete data management systems that were put in place for when teams had five years to take something to market. And nowadays they have to do things 3x as fast, but they have the same tools to do so than they did when things took longer.
So there's a variety of tools, anything from design to work processors, PDF and emails and PLM tools. And the problem is that none of these talk to each other. So we see kind of on the other side of the coin in the software space. There's also an incredible ecosystem of lots of tools and best in class specialty tools.
The difference between these two is that those tools have perfect connections with each other. So having a variety of tools isn't as much of a problem because they can talk to each other perfectly. There's no information loss, they use APIs, and all of that is transferred smoothly. When you go to the hardware side, the difference is the way you transfer information, for example, from a design tool to your next tool that you need to use it, everything gets produced down to the lowest denominator, which typically is a PDF, an email thread, a spreadsheet, right?
And this is causing a lot of problems for these teams. And there's different trends that are forcing these teams to become more and more agile. A similar transformation that software went through about 10 years ago. This process has a lot of manual tasks, which are error prone. Think about it when you're trying to, when a typical example is people have to manually create a list of all the changes between their last design and their new design.
This could take hours of time and it becomes a little bit of a playing where's wilder game where you have to put two images side by side and try to identify what's different, which you can imagine it makes it very easy for things to fall through the cracks. There's a lot of manual exports, there's disconnected, disconnected data, and all of these teams live in isolation.
There's also the iteration speed, which is, um, this, a lot of the systems were established for waterfront methodologies. And as teams become more and more agile over time, there's not a lot of real-time feedback and they have a hard time adapting to, um, to changes in the environment. So a typical example for that will be in the past, it takes you, it took you three weeks to validate a new component.
So you have a design and one of your components goes out of stock or you need a, um, you need to qualify a new part. It may take you three weeks to set up a meeting with someone, get everyone's input, go through all of your checks and then release that at the end of it. Nowadays, it's like your entire cycle may take that long.
So if you're spending that much time to qualify a part, you can't really keep pace. And before maybe you didn't need to because manufacturing will take that long. But nowadays with 3D printing and with fast PCB shops, teams have to respond more quickly to changes in the environment. And the third side is the communication. So different teams and schedules, people working remotely, or we see a lot of distributed teams with specialties.
So, um, sometimes by design and sometimes by, by chance, if you think about it, like having to design a board or having to, to work on a design that, um, requires a very specific skill and the expert on that leaves in 3000 miles away from you, you suddenly may have a distributed team, even if it's not how your company was set up to do.
So we see a lot of these specialty design getting outsourced or going to experts and they need to kind of keep everyone in the same, same pace, same place. So as these teams converge, we see the hardware industry looking at software for a lot of best practices. And software has some of the similar problems and they figure out ways to manage those issues.
And the biggest influence we see between these two teams is the share of workflow methodologies in which in, in our particular case, we see it as like the adoption of Git based workflows. So Git is a protocol that's very, um, well understood and adopted in the software space that is becoming more and more popular with hardware engineers as hardware engineers start working with software teams more and more, or sometimes being themselves part-time software engineers, right? So some people may be doing it 5% of their time for a small project, or some people may be doing it 50% of their time for writing firmware.
So this kind of workflows, the, what they allow these teams to do is to better organize their own information and make it easy to collaborate with the rest of the stakeholders. So hardware is a very collaborative process where there is a variety of stakeholders that participate. So we work with electrical engineers. There's also mechanical engineers, tests, manufacturing, and then there's the other side of the house where once things are ready to get, um, manufactured, they go into purchasing, planning, planning, CMs, applications, suppliers, and everything in between.
So when we think about the, um, we, we see the, there's a lot of influences between the two disciplines, but the, there's some things that are similar and there's some things that are different. So some of the things that we see, uh, similar is the trend towards agile development, which has been influenced by first the technology that allows these teams to move quickly and also consumer demands and products. So think of like, uh, if a product, a new hardware product, let's say the new iPhone took five years to, to get to market. By the time you launch it, the technology that's in it, it's already obsolete. If it's, if it's such a slow and long process to get to market.
So these companies are becoming more agile. They need to move faster. They need to release faster. And there's also the need for flexible and quick iterations. And the biggest impact here has been coming from the, um, supply chain disruptions from the past couple of years. We, we hear from a lot of teams that feel like they're playing whack-a-mole where, um, the, when you start a design, all of the components or all the parts that you're putting inside of it may or may not be available by the time you complete your design. So you almost have to start designing with enough flexibility to be able to accommodate like three, four or five, seven different changes along the way. So having that flexible infrastructure that helps you kind of navigate
those processes is really, really important. And the same happened in software, right? On the other side, some of the differences that we noticed as these two workflows converge to being the same, the first one is around UIs and things evolving for text versus spatial information. So the biggest fundamental difference between hardware development and software development is the, the format of your output, right? As a software developer, you're working on text-based code. That's kind of your your day-to-day, your main format. That's the hardware developer. Most people operate in visual vector drawings, right? So think of like 2D images, 3D images, and those could be a circuit board. It could be a schematic. It could be a component. So adapting the workflows from one place to the
other, it's one of the biggest differences. And then the other big difference between the two disciplines is the, the cost of mistakes. So the, the cost of an error in software could be high. The cost of an error in hardware is exponentially higher because we're talking about physical and real world manufacturing. So compiling the wrong version of a board, it could mean anything from a couple thousand dollars on rework and faulty prototypes to millions of dollars in having to organize a new factory depending on how soon or late in the process you catch this mistake. So the, the hardware teams need to be as flexible as the software teams, but they also need to have higher checks and, and kind of, and, and guard rails to prevent some of
these mistakes from making it through because of the nature of how expensive they are. So, um, kind of going back to like the biggest difference in like, when we think about the formats between the two text versus images, um, we can talk about kind of like how things translate between the two spaces. So how do things look in one versus the other? So when, when you talk about software development and peer reviews, we talk about pull requests when we talk about it in hardware, we talk about design reviews, but, um, these processes may seem different in, um, high level, but they're actually a lot more similar than, than other people realize. So the equivalent of inline feedback for software, it could be, uh, snippets and hardware. So 2d kind of drawings around that, um, commits in software have an equivalent in
links to locations in hardware. There's CI/CD tests for software in hardware. We can run tests as well, uh, depth between the two, one in text-based data, the other one in visual vector and schematics, um, and then color coding things on one or to the other as well. And, um, um, if we follow the evolution of, of software, the, what we see as the future of hardware development as well, it's following a similar path. So where first came revision control, then came collaboration, and then came automation. Revision control has existed for some time in hardware, and it's getting a lot, a lot better. Collaboration, it's where the industry is at right now. And there's a lot of efforts going into making that process smoother for all of the stakeholders.
So what comes next to us in our mind is automation. What does automation mean in both of those? In software, we're talking about CI and continuous integration unit tests that you can run on your code on a regular basis. Our, our kind of thinking and what we, we work on is how does that translate to hardware, right? What does that mean into the hardware space and what is an automated test look like in that? So hardware today has a lot of testing built in place. The way that happens the majority of the times is at the end of a design process. Once when everything is completed, the shift that we see is from going from doing things when your design is completed, when it's really hard to go back and
implement those changes, shifting those tests to be set up at the beginning. And you set them up once, and then they run with you with the rest of the process. So that you can catch this information, not at the end when it's too late to make changes, but at this thing, as this thing's happening throughout the design process. So that's one of the big kind of trends and futures that we see in the hardware space. The other one is addressing cross-disciplinary needs the same way that hardware has a lot of disconnected tools that don't really work well with each other. The same happens at the stakeholder level where there is a lot of integration that needs to happen between the different teams and where information falls through the cracks or problems happen because things fall through
the cracks in the end. So think about electrical engineering to mechanical engineering interfaces, electrical engineering to firmware, firmware to operations, mechanical to operations, and all of this different when I was talking about the different people involved in hardware development. When these teams need to work with each other, there's a lot of error prone areas at those intersection points. So working beyond like your specific in field to more of like the broader product level, that's another area where hardware development is going to get better over time. And then kind of the third area to address is solving supply chain challenges that have been grappling, been distributed for the past several years. The other big difference between the software and the hardware side being that physical manufacturing
component where when you're done with your design, you're not done quite yet. In software, you may compile and ship your product and that's how his complexity in the hardware space, we enter an entire new process of physical manufacturing that comes with its own challenges along the way. So giving the teams the ability to have both visibility and flexibility to respond to changes in the supply chain side is kind of like the other big area for the future of hardware development. So with that, when I close for today, I want to thank everyone for your time and offer if you have any questions about the future of hardware and software and how these two disciplines are coming together, when are they coming together and how versus when they're not, feel free to
send me an email. You can email us at [email protected] or you can email me personally at [email protected]. I'll be happy to chat and answer any of your questions. So that's all for today. Thank you so much for joining. Appreciate your time and hope you enjoy the session and the rest of the conference. you
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
What hardware teams can borrow from software, and what they can't
Valentina Ratner explains why faster hardware cycles demand Git-style workflows, design reviews and early automated testing, and where hardware's physical stakes change the rules.
In short
Hardware engineering is following the same path software took a decade ago: first revision control, then collaboration, then automation. Valentina Ratner, co-founder and CEO of AllSpice.io, argues that hardware teams can borrow software's workflows, such as Git, pull-request-style reviews and continuous testing. She also argues that they must adapt these workflows for visual design data and for mistakes that cost far more than a bug in code.
Key takeaways
Gain a deeper understanding of the relationship between software and hardware engineering.
See how cross-disciplinary innovation is shaping the hardware industry.
Learn how applying software principles to hardware design will shape the future of engineering.
Faster cycles, same old tools
Ratner opens with a shift in how hardware gets built. As she describes it, a couple of decades ago taking a hardware product to market took about five years, with three or four prototype iterations of roughly a year each. Today, as the cost and time of prototyping have fallen, products reach market in 12 to 24 months, and teams run five to nine prototype iterations along the way.
The tooling has not kept pace. Many of those iterations still run through legacy data management systems designed for the slower era. "And nowadays they have to do things 3x as fast, but they have the same tools to do so than they did when things took longer." (2:45)
She contrasts this with software, which also has a large ecosystem of specialist tools. The difference is how well they connect. "So having a variety of tools isn't as much of a problem because they can talk to each other perfectly." (3:30) In hardware, information moving from one tool to the next gets reduced to the lowest common denominator: a PDF, an email thread or a spreadsheet. The result is manual, error-prone work. Her example is engineers comparing two design images side by side to list every change by hand, a task that can take hours and lets problems slip through.
Iteration speed and distributed teams
Ratner points to two more pressures. First, many hardware processes were built for waterfall methods. Qualifying a replacement component once took three weeks of meetings and checks, which was fine when manufacturing was slow. With 3D printing and fast PCB shops, an entire design cycle may now take that long, so teams have to react much faster.
Second, hardware teams are increasingly distributed. Specialist design work is often outsourced or handled by an expert who may live 3,000 miles away, so a team can become distributed even if the company was never set up that way. Hardware is also deeply collaborative, involving electrical and mechanical engineers, test, manufacturing, purchasing, planning, contract manufacturers and suppliers.
This is why, in her view, hardware is looking to software for best practices, starting with Git-based workflows. Hardware engineers increasingly work next to software teams, or write firmware themselves for anywhere from 5% to 50% of their time.
What the disciplines share, and where they split
Both fields feel pressure toward agile development, driven by technology and by consumer demand. Ratner uses a new phone as an example: if it took five years to launch, its technology would already be out of date. Supply chain disruption has made flexibility essential, with teams describing a game of whack-a-mole as parts disappear mid-design. "So you almost have to start designing with enough flexibility to be able to accommodate like three, four or five, seven different changes along the way." (9:00)
The differences matter just as much. The first is the format of the work. "As a software developer, you're working on text-based code." (9:45) Hardware engineers work mostly in visual 2D and 3D vector drawings: circuit boards, schematics and components. The second is the cost of mistakes. "The cost of an error in hardware is exponentially higher because we're talking about physical and real world manufacturing." (10:37) Building the wrong version of a board can cost anything from a few thousand dollars in rework to millions for setting up a new factory, depending on when the mistake is caught. Hardware teams therefore need software's flexibility combined with stronger guardrails.
Translating software practices into hardware
Ratner maps software concepts to their hardware equivalents. A pull request becomes a design review. Inline code comments become feedback on snippets of 2D drawings. Commits correspond to links to locations in a design. Diffs and color coding still apply, but to visual schematics rather than text, and CI/CD-style tests can run on hardware designs too.
She sees hardware moving through the same stages as software. "So where first came revision control, then came collaboration, and then came automation." (12:29) Revision control already exists and is improving, and collaboration is where the industry is today. "So what comes next to us in our mind is automation." (13:23)
The future: early tests, connected teams, resilient supply chains
Today most hardware testing happens at the end of the design process, when changes are hardest to make. Ratner describes a shift toward setting up tests once, at the start, and letting them run throughout the design so problems surface early.
She names two other areas for improvement. One is the handoffs between disciplines, such as electrical to mechanical, electrical to firmware and firmware to operations, where errors tend to collect. The other is the supply chain. Unlike software, finishing a hardware design starts a whole new manufacturing process, so teams need both visibility and flexibility to respond to supply chain changes.
Watch the talk
Software to Hardware: Two Sides of the Same Coin? by Valentina Ratner (17 minutes): https://www.womentech.net/video/software-hardware-two-sides-same-coin
About the speaker
Valentina Ratner is Co-Founder & CEO of AllSpice.io, an experienced engineer and former Amazon PM who managed infrastructure projects and internal productivity software. She holds a B.S. in Mechanical Engineering from Boston University, an M.S. in Engineering (Computer Science) and an MBA from Harvard, and co-founded AllSpice in 2019 alongside Kyle Dumont.
Questions this talk answers
Why do hardware teams need software-style workflows now?
Why do disconnected tools hurt hardware teams more than software teams?
What is the biggest difference between software and hardware workflows?
How does a software pull request translate to hardware?
What comes next for hardware development?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Git-based workflows
CI/CD
Agile
Waterfall
Cite this talk
Valentina Ratner. "Software to Hardware: Two Sides of the Same Coin?." WomenTech Network, 2023. https://www.womentech.net/video/software-hardware-two-sides-same-coin