Anurag Gupta is an innovation leader in technology and product development. Anurag has led global teams through dozens of successful New Product Introductions (NPIs) for companies including Google (Glass, many Nest generations), View, Inc., Sense Photonics, and Velodyne Lidar.

If you want to dramatically increase success downstream of NPI, you must remember, “Slow is fast.” Your choices during NPI will have drastic repercussions when you make it to mass production (MP), so every decision must be made carefully. But what does this look like in practice? Below, I’ve outlined the best ways I’ve found to implement this approach to maximize yield with as few NPI headaches as possible. 

Framework for NPI success
Avoid these shortcuts that become pitfalls
Streamline NPI with Manufacturing AI technology

Framework for NPI Success

1. Take a Phase-Gated Approach with One Design at a Time

A phase-gated path is crucial for structured progress. This involves well-defined checkpoints and stages, such as OK2Concept and OK2Design, ensuring clarity and focus. Entrance and exit criteria from EVT, DVT, and PVT are absolutely critical to define and hold the team to.

It's also vital to only have one prototype design at each stage – handling multiple prototypes simultaneously dilutes resources and hinders collaboration. I once came into a smaller company where people were simultaneously working on multiple design approaches. As the new leader, I got pushback when I suggested reducing to a single design and I hesitated to shut everything down. Everyone wanted to build things and felt that the thinking and design legwork, reviews, and documentation took too long. However, when the prototypes weren’t meeting requirements, we were stretched so thin across so many designs that collaboration fell apart. Only then did the engineers agree we needed to consolidate our efforts on one design.

2. Avoid the 'Fix It Later' Mentality

In NPI, a proactive approach is essential. Engineers often rush to prototype, overlooking opportunities for improvement and cost savings. Finalizing designs too early can escalate costs and complicate manufacturability. Costs never come down later, so if you dial in expensive choices from the start, you’re locked in. This principle applies equally to hardware, firmware, and software. The key is integrating cost considerations and manufacturability into the design process, not as an afterthought.

3. Include Hardware Testing Strategy in Your Design

When you design your product, you should always consider how you will test it. While this philosophy prevails in ASIC design, where test and design teams design together, it applies to all hardware and software development. In devices, sometimes product design, mechanical, or electrical engineers feel like figuring out hardware testing is someone else’s job – like reliability engineers or (the aptly named) test engineers. However, a design is not done if it doesn’t meet the requirements – which includes testing. Ensuring your teams collaborate on test design and specifications early can prevent fire drills before or after development.

Avoid These Shortcuts That Become Pitfalls

1. Don’t Skip Second Sources

A significant pitfall in NPI is the reliance on a single source for materials. From an engineer’s perspective, reducing the variable of two vendors enables them to really test the design – however, from a product and business perspective, the design of a part isn’t complete if only one vendor in the world can make it. Single-sourcing can lead to supply constraints and higher costs – and adding a second source after EVT is harder because, over time, there’s less flexibility in the design. That can lead to less competition, which is worse for cost negotiation. Always ensure that you have multiple sources for all key parts in the plan from Day 1.

2. Don’t Skip Design for Manufacturability (DFM)

I once came in as an engineering leader at a company that was seemingly doing everything right – they were a smart team with regimented phase gates and multiple suppliers. But, they failed to consider manufacturability from the start. The result was a prolonged, painful process with seven EVTs. The team fully turned over two times and it took three years to get the product out. Contract manufacturers (CMs) don’t want to be responsible for fixing design issues in DVT – but involving them in DFM from the beginning can be a huge accelerant. Whether you have in-house experts for every manufacturing process or not – your vendors have a wealth of DFM knowledge, and leveraging it may only take one or two extra weeks in your process, but save you years later on. I highly recommend doing DFM in pre-proto, proto, and pre-EVT.

3. Don’t Punt Issues to End-of-Line Testing

Excessive reliance on end-of-line testing can lead to significant yield losses in early production. A robust testing strategy should involve thorough testing of sub-systems throughout the production line, rather than waiting until the final stage. Six Sigma techniques are designed for high-volume manufacturing, but still have a place in low volumes. Anything you load onto the production line—sub-systems, software, mechanical, and electrical—should all be tested. More importantly, your design should be amenable to testing at the subsystem level. Your line should be optimized to ensure yield from the get-go.

4. Don’t Skimp on Firmware: You Need Three Builds

Firmware must have three builds: diagnostics, manufacturing, and product release. It should be modular so it’s easy to maintain these three builds – but they need to be separate.  The mistake I’ve seen repeatedly is to focus only on the product release firmware and to use that same build for manufacturing, while ignoring a diagnostics build altogether. 

When there’s only one build, every time the firmware build is updated (which is weekly), the entire manufacturing line needs to be revalidated for production! This could take weeks or months – and sometimes that due diligence is skipped, causing huge problems later on. It’s critical to keep these builds separate.  

A diagnostics build is critical to support automated test infrastructure for both validation and manufacturing.

Streamline NPI with Manufacturing AI Technology

During my time at View, we used Instrumental to turbocharge our NPI process. We had an ambitious schedule and a very lean team.

We leveraged it as a shortcut for hardware testing – instead of building out an expensive testing system that would need to be changed every time we made a change, we used Instrumental’s AI technology to do visual inspection and discover assembly and design issues throughout multiple steps of our assembly process. As the design changed, Instrumental’s algorithms were easy to update and adapt. Instrumental helped push through the “move fast and fail fast” stage of our design so we could quickly identify what needed to be fixed so we could ship on time.

I’ve worked in larger companies with more resources in time and team – and would love to have had Instrumental there too. Accelerating the process of issue discovery and failure analysis is a win for any team: there’s no such thing as an NPI completed “too early” or a PVT with yields that are “too high”!

Conclusion

Accelerating NPI in manufacturing is a challenge that requires methodical planning, leadership buy-in, and the integration of innovative technologies like machine learning in manufacturing. By adopting a phase-gated approach, avoiding short-sighted shortcuts, and leveraging AI manufacturing tools like those offered by Instrumental, companies can enhance their NPI processes and bridge gaps in testing and analysis. Designing for manufacturing from the start and keeping the lines of communication clear are both key to successfully achieving NPI with as few bumps and scrapes as possible.