It is almost the end of 2024. This article is the final article in the CADDi's product team Advent Calendar 2024. While the vast majority of this blog is written in Japanese, I write this in English as CADDi has offices and customers in four countries, and has employees from around the world.
- Looking back
- The beginnings of CADDi Drawer
- The manufacturing industry
- About software engineering
- Building engineers, to build for the future
Looking back
While every year in a startup generally feels faster than the last, this year in particular felt richer and more diverse than in years past. We have had phenomenal growth in our enterprise SaaS products, which has been quite the journey. But with exponential growth came exponential challenges, and this year we put significant effort in building out the management team, merged two lines of businesses into one, involved the engineering team to improve management accounting, developed a dedicated recruiting team within the VPoE office, ran governance programs for new geographies, and began hiring from overseas again.
The beginnings of CADDi Drawer
Early feedback on our prototypes
We launched CADDi Drawer a few years ago. It started as just an idea, then came a prototype whipped up in just a few days. You could hardly call it a product, but it was enough to garner interest and get valuable feedback through customer interviews. This allowed us to better understand the potential purchasers and users of this would-be product. Often in the startup world, new products struggle to get quality feedback from industry professionals, but we were fortunate enough to have an existing customer base of manufacturing professionals, thanks to our customers of our main line of business at the time, CADDi Manufacturing.
In that business, we operated as a virtual factory, building out and operating complex supply chains to deliver custom ordered mechanical parts to our customers. At its peak, we had inspection and warehousing facilities throughout Japan and Vietnam, where we conducted final inspection of parts before they got shipped off to our customers. While we have since merged that line of business into CADDi Drawer, the sweat and tears from operating a physical supply chain gave us deep insight into the issues facing manufacturers, from the difficulties of a worker on the line, all the way to challenges of business planning.
Back to the prototype. It was essentially a user interface akin to Google image search, but for mechanical drawings. These drawings, sometimes referred to as orthographic drawings, are a graphical representation of the mechanical part to be created, its various dimensions, and other engineering details. A typical automobile would have tens of thousands of such drawings, one for each of the parts that get assembled together.
Our first demos were very limited. You would type in some keywords, such as “bracket SS304”, and you would see a grid of thumbnails of drawings that contained those keywords. What surprised us was that on every demo, we consistently heard positive feedback about how being able to see a the thumbnails was instrumental in visually “scanning through” large volumes of data. Why were customers so interested in seeing thumbnails, an experience most of us take for granted? There are plenty of software packages and services out there that give us that kind of experience. The good old file explorer in Windows has both list views grid views. Sharepoint and Google Drive allow us to search by text.
The importance of domain expertise
The difference, we realized, was in the details of the user experience. It was the speed of the search and rendering, and also the thumbnails being just the right size–big enough to identify the overall shape and other important details, but small enough to enable efficient visual scanning. B2B enterprise software is not known for greater user experiences, but on the ground it makes all the difference. After all, the purchasing team is different from the end user. Nonetheless, we believe in the power of scratching an itch in just the right way, but we recognize that simply scratching itches will not change the world--we must identify just the right set of itches to scratch, that will build up to make a big difference. We even have a concept that we internally call the “double loop” where small incremental gains on the ground (small loops), can be used to incentivize building up and executing large corporate strategy shifts (big loops).
Looking back, after a significant amount of growth, it became clear that there is a lack of domain specific software that address issues felt on the ground in industry verticals. It is difficult for the typical software professional to understand and empathize with the pressures of being in an unrelated industry such as manufacturing. We believe it takes a combination of technical prowess and domain expertise to identify problems that can feasibly be solved with software. Thanks to our heritage in physically producing and delivering machined and sheet metal products, we are fortunate to have had the opportunity to develop a team that has both. Many of our customer facing teams include members from manufacturing powerhouses such as Toyota, Honda, and Mitsubishi, while our software teams have engineers with backgrounds in payment systems, e-commerce, mobile apps, and IoT to name a few.
What started out for us as just a simple demo, implemented as a grid of thumbnails powered by an off the shelf search tool, has now morphed into a powerful knowledge engine for some of the largest corporations in Japan and abroad. Despite expanding far beyond just drawings to other valuable data assets such as 3D CAD and documents to be a Manufacturing Intelligence platform, the CADDi Drawer product still retains an important user experience lesson from its infancy–the grid of thumbnails.
The manufacturing industry
Physics complicates things
Manufacturing is the manifestation of ideas into concrete physical reality. It underpins everything from our smartphones, to cars, to the data centers that power our apps. Even in the digital age, we humans still live in a physical world, and we saw during the pandemic just how much of our lives depend on the global supply chain that powers the flow of goods. Perhaps things will be different when we are all hooked up to the matrix, but I’m not sure if that’s a world I would like to live in anyhow.
Compared to digital goods, the laws of physics complicate everything. If I drill a hole in the wrong spot, I cannot hit Ctrl-Z and undo. If I drop and damage a part, I cannot recreate it with a copy and paste operation. If I make a design change and modify some dimensions, I cannot migrate existing physical parts to the new dimensions. If the required nuts and bolts do not arrive at the assembly line on time, all I can do is call up the supplier and inquire. All that to say, it is astronomically cheaper to correct for mistakes early in the process, ideally in the digital world. There is very little we can do to economically “fix things up” once we are in the physical world.
Retiring employees take knowledge with them
Businesses in manufacturing, from small job shops to the Fortune 500s, face a shortage of skilled labor, exacerbated by their aging workforce approaching retirement. The loss of knowledge and experience that has been built up over the decades have heavy consequences for everybody. No longer can you just ask that guy who's seen it all. The veteran walking encyclopedia will no longer be there to answer questions on a whim.
The advent of digital technologies has allowed us to have huge advances in robotics and automation, much of that critical experience still lives in the reams of paper stored in a warehouse, and cached in the minds of the veterans. Despite all the media coverage of digital technologies and AI, we were surprised to hear that even some of the largest industrial manufacturers in the world still have countless historical drawings in boxes. It turns out, it's not meaningful to just scan them all into a file server, if you don't have a good way to use them.
In the software space, all of our software and design docs tend to be in the digital space, readily accessible at the touch of a key. We benefit from the accessibility of information, the ability to reuse existing components, and build on top of them. In contrast, manufacturing has a much longer history, requiring us software professionals to step up to the plate to help leverage the past, and push the industry forward.
The pressures facing manufacturers today
The last decade has shown an increasingly volatile world, from chaos in the supply chain during the pandemic, to the multitude of geopolitical escalations across the globe. We consumers are demanding ever more variety, delivered faster, and straight to our doorsteps. We demand increased corporate responsibility for CO2 emissions and environmental impact. Manufacturing is inextricably intertwined with logistics, borders, economics, and politics. It supports our every day lives, and it must continue to change to adapt to the new realities.
We believe that the best way forward is to leverage the knowledge built up over the past decades, by learning from historical data, and leveraging it to build on the shoulders of giants. Much of our current efforts are focused on making sense of existing data, through investments into data engineering, domain specific AI models, and building applications that allow for data acquisition without getting in the way of everyday operations, and provide highly refined data to feed use cases critical to our customers.
About software engineering
Personally, as a technologist, working with web-related technologies is source of entertainment. Compared to the vendor-coupled embedded systems I used to work on, the open nature of web development is a breath of fresh air. It is a testament to human collaboration, with major corporations from across the globe collaborating on open standards, to build the foundation for what’s to come. While the high volatility of the ecosystem can be distracting, the same freedoms that allow for volatility have also pushed the boundaries of what the internet and the browser could accomplish.
To name a few highlights of 2024 that come to mind, we saw the release of React 19, building upon Suspense and React Server Components. The xz utils backdoor was a strong reminder on the challenges of securing our software supply chain. ClickHouse added Iceberg support, and DuckDB hit the 1.0 milestone. We saw the next major set of changes to the Rust language in the 2024 edition, to be released as stable in just a few months. Kubernetes is everywhere and turned 10 years old. OpenAI acquired real time vector database Rockset, and major cloud providers are hard at work developing their own LLMs, while Meta released Llama3 as an openly available model.
It was an eventful year in the world of web development, data engineering, and AI, and I could nerd out all day about this. However, as a professional in an industry vertical, none of that is important. We have problems to solve, and technology is a means to an end. Programming languages, frameworks and methods are all just tools in a toolbox. As much as I would like to go on about the phenomenal engineering behind the scenes, a customer could care less. What makes a great engineer different from a good engineer, is the ability to acquire, select, and leverage the best tool for the job. Not a tool that works, but the best tool. That best tool is not necessarily the one that leads to the most elegant technical architecture, but it is the one that fits the organization and the problems facing customers. And therein, I think, lies one of the challenges for myself, and the greater software community, in the coming years. We must simultaneously work towards two somewhat orthogonal objectives, the pursuit of technological advancement, and the effective application of tools to solve problems.
Building engineers, to build for the future
As the adage goes, when all you have is a hammer, everything looks like a nail. It is our responsibility to ourselves as software professionals to continue to grow our individual toolboxes, so that we know what tools exist in the world, and to share that knowledge with our communities. But it is our responsibility as employees to ensure that we are acting in the best interest of our employers, by leveraging the right tools for the right problem. Not because it tickles our curiosity, but because it leads to the best results for customers.
At CADDi, we believe that having a large toolbox of both hard and soft skills is what allows us to build just the right solution. We want our software engineers to be have T-shaped skills–well-versed in their respective fields, but also possessing a holistic view of what it takes to build a product. We want our engineers to not just keep up with the fast paced change of technology, but also continue to develop breadth within and beyond technology. Thanks to our customers, we are facing unprecedented growth in our business and product, but with that comes the need to understand the economics of the business, the complexities of human thought, and the ability to roll with the punches as the environment changes.
While much of the developed world outsourced their manufacturing overseas in the search for ever cheaper goods, Japan has remained a bastion of manufacturing. We leverage this as an engineering team in Japan, taking the best of both worlds--the manufacturing expertise in Japan, and an ever globalizing software engineering workforce. But with this advantage comes the challenges of having a highly diverse workforce in a relatively homogenous culture in Japan. The most obvious is language but differences in business customs and culture also come to mind.
In the last year, we have made big strides towards this challenge. We have had a number of great folks join our management team, many of whom have experience both in Japan and abroad. Some ran businesses in the US, while others brought up development centers in Asia. We have also worked to enable a more diverse working environment by developing engineering specific HR functions to address the needs of a multi-national team with mixed languages--something a Japan-centric HR team would not be able to do. The technology division all-hands meetings are run twice, once in English and again Japanese. We have regional all-hands to account for geographic differences.
We are not perfect, but we strive to continue empower our customers by leveraging the world class manufacturing expertise in Japan, combined with world class software engineering.
As we try to answer our customers' appetite for better software to democratize their historical data, we are accelerating our hiring from around the world to keep up. I hope this blog post gives a bit of insight into the potential impact of our work. CADDi's mission is to "unleash the potential of manufacturing", and this next year will be a time of transformation as we rapidly scale out our team, business, and customer base around the world.
It’s an exciting time, and if you’re interested in taking part, please check out our open positions in Japan and in the US, Vietnam, and Thailand.
