Matt Leaverton – technical program manager at Velentium, addresses a dilemma that many business will come across with automated test stations.
Consider the following hypothetical situation: It is midnight the night before the bid for the next test station development project is due. You say to yourself, "I know a thing or two about computer vision" and proceed to mark this line item down as a cursory engineering effort, thinking you are saving money. Too many months later you find yourself drowning in cheap cameras, overcomplicated software, and despair while serenading yourself with the sound of deadlines whizzing past.
The eternal quandary
Perhaps everyone reading this has already solved for their company how to efficiently design complex systems. Congratulations! You are ahead of the curve and can stop reading.
During my time developing test-system automation software at a major software company, I observed rampant "not-invented-here" syndrome. This seductive concept lures even the best-intentioned engineers to design everything from scratch because "our version" will clearly be harder/better/faster/stronger than the off-the-shelf solution, simply because it was us that created it. The allure of "DIY" for its own sake never fades for engineers, whether at a small engineering startup or a multi-national conglomerate. When you next find yourself quoting, designing, or managing such an engineering effort, consider the following heuristics when deciding whether to custom build or buy off-the-shelf components:
- NUD – (New, Unique, Difficult) - Focus your design on the most critical portions of the system. If the system would fail if this component is not exactly fit to your needs, build it. If what you need is truly unique and not available for purchase, also build it.
- Unique offering - If a portion of the design is part of your value proposition as a company and you believe retaining control and customisation of the design could impact revenue flow, then consider building.
- Market advantage - If the solution you consider is worth the investment for the market advantage, build.
From the tone and content thus far, clearly, I favour the "buy" option. In the medical device industry, every design decision has a long tail. Equipment in the system will require calibration, periodic maintenance, traceability, capability justification, and verification activities at a minimum. Custom designed equipment then adds the additional burdens of prototyping, PCB re-spins, machining, manufacturing lead-times, design documentation, configuration management, custom software/firmware, and long-term support--to name a few. In a market where the supply chain is completely unpredictable and that special chip that the custom board your team lovingly created suddenly has a 52-week lead time, then finding an off-the-shelf solution starts to look especially attractive. In a manufacturing line-down situation, nothing beats being able to call a service department instead of flying a treasured employee across the country (or world) at your own cost to save the day.
Sometimes “buy” is plainly not the best option. At Velentium, we chose to invest in creating our own software test automation framework. We distilled decades of experience in testing medical devices into a powerful, easy-to-use framework tailored specifically to our development workflow, including creation of operator friendly, touchscreen optimised GUIs. We knew it was the right move because designing and building automated test systems is something we do often, for many clients, on many projects. For us, “build” made sense because we were able to create a reusable solution that saves us and our clients time and money again and again.
I'll take door number three
Build vs. Buy is the eternal discussion in design products. I contend that especially in the medical device industry that there is a third option that we should strive to choose as often as possible. Option number three is "neither". The most efficient, effective, and efficacious design element is "nothing". There is no capital cost, no calibration, no maintenance cycle, no V&V, and no risk in “nothing”. You too can use “nothing” in your projects. Unlocking this ability comes from deliberate design and clever engineering, which stems from thorough requirements. Before making any decisions about build vs buy, put on your system design hat. Seek to remove as much of the design as possible--push requirements upstream or downstream, merge tests and requirements where possible, test only what is truly necessary, and simplify, simplify, simplify.
Step back and ask yourself:
- What problem am I truly working to solve? Am I working to solve the correct problem?
- Are the requirements defined? Is the problem space defined and understood?
- What is the risk associated with this design?
- What is the project budget and schedule?
- Does my team have the skillset to integrate and support this choice?
- What milestones ride on these decisions? Are there studies or submission deadlines that could be delayed by this choice?
Ultimately in the design medical devices, patient safety and quality of life is the primary focus. Always consider how you can reduce risk and reduce complexity (which in turn reduces risk) to yield a focused, effective, and useful system.
