Hardware Product Development Process

This post is the prequel to my prototyping and manufacturing series. It is also the foundation for my book, “Bringing a Hardware Product to Market: Navigating the Wild Ride from Concept to Mass Production”.

Here is a quick whiteboard sketch from one of those conversations.  It illustrates key stages of the hardware product development process for a typical, medium to high volume consumer electronics device, to be sold in the US.  Let’s assume the product involves custom PCBs inside custom plastic housings, has a cost of goods below $50 USD, and is targeting an annual volume of 10,000 to 50,000 units per year for the first 12 months of production. Let’s further assume the product will be assembled, tested and shipped by a third-party contract manufacturer (CM), with whom you have developed a partnership relationship.

Hardware product development process
Hardware development process

If you are a student of the Lean Startup movement, your first reaction is probably outrage.  This looks exactly like the much maligned stage-gate process from the Jurassic Age. Didn’t we do away with all this waterfall business with the advent of Agile Software Development?

My response: Alas! Hardware development is different from software development. Consumer electronics, while often including integrated firmware and cloud based and/or mobile software, is especially not like a pure software product.  There are several key differences between hardware and software product development.

  • Long lead times in designing a subassembly.
    • A subassembly with multiple parts that fit together in intricate ways need to be designed together – the more “designed” the product, the greater the packaging challenge  from a physical standpoint and the longer it takes to complete the detailed design.
    • Therefore it is very hard to “complete” the detailed design of any one part until the entire subassembly is designed. Even the most carefully designed part could need to change at the last minute to accommodate  assembly considerations as the subassembly is being finalized.
  • Long lead times in procuring custom parts.
    • Agile methodology, practiced properly, can yield working software at all times, and key milestones in 1 or 2 week sprints (definitely no more than 3 weeks).
    • For hardware, the elapsed times to procure custom parts can run much longer than that.  Rapid prototyping techniques can print or grow a part in 2-3 days.  But if you need a complex 3D metal part made (such a custom gearbox housing), the machining lead times can run from 4-6 weeks.  Cable harness lead times are generally in the 2-4 week range.  PCB procurement run 1-3 weeks at prototype quantities.  Production lead times are usually even longer.
  • Sequential phases resulting in longer elapsed times from start to finish.
    • While a lot of engineering tasks can be parallelized, at the end of the day, each prototyping phase is complete when the entire product is assembled, debugged, and “brought up” to yield a functional prototype.
    • The functional prototype supports both engineering testing and user testing, neither of which can be done until the prototype is complete – resulting in a sequential schedule and timeline.
    • The results of engineering and user testing provide input into the next phase, causing another cross dependency.
  • System integration takes much longer than pure software projects.
    • With pure software products, as long as the software design is rigorously conducted prior to writing code, once the design is done and APIs are defined and documented, developers can write code and test code in parallel based on the agreed upon APIs between modules.  Integration will still take time, but a lot of the legwork is done prior to that point, so the elapsed time can be shorter.
    • For hardware products with software running on board, no matter how much care is put into system architecture and design, unexpected side effects come up when each prototype is assembled and tested for the first time. Unfortunately, these side effects won’t show up when you test modules in isolation – only when the entire product is fully assembled.
    • For this reason, system integration for hardware/software products is usually measured in weeks (for simple products) or months (for highly complex products).  Teams ignore this task at their own peril – if you don’t do this right during the early phases, you will simply pay for it in project delays later on.
  •  High level of capital investment required in tooling, requiring careful due diligence.
    • Unlike software projects, there is usually a high level of non-recurring engineering cost (NRE) in designing and developing the hardware product itself, and the cost and lead time for procuring prototypes and pre-production builds at each stage of the game. You can use rapid prototyping techniques to iterate your design quickly in the early prototyping phases, up to a point. But at some point you have to lock and load and invest in tooling.
    • Tooling can cost $5k to $500k (or, in the case of industrial automation products, up to a couple million dollars).  The tooling cost is a function of the complexity of the product, the number of custom parts that require custom molds, the processes used to make these custom parts, and the projected volume.

When you are looking at a multi-month, $1-2M investment when all’s said and done, you had better put in some formal check points (or gate reviews), where you assess what you have learned with each phase, and decide whether you will go forward and spend the next $500k.  It’s the only fiscally responsible thing to do.

Now let’s look at these phases in detail.

  • Ideation: 
    • Every program starts with a lightbulb moment.
    • The idea then goes through some sort of vetting process to determine whether it has adequate economic promise to be worth pursuing or not. Hopefully, you will have done some market analysis, identified a beachhead market, performed a bunch of discovery research, developed buyer and user personas, and have come up with a very good idea of a solution that solves some unmet need in this market. (If none of this makes sense: read Disciplined Entrepreneurship: 24 steps to a successful startup by Bill Aulet.)
    • Some people insert a gate (one of those diamonds depicting a milestone) called “Basis for Interest”.
      • This is a time to get back to 50,000 feet and think deep thoughts about whether the idea has enough merit to warrant going forward to the next step.
      • Too often, technical founders fall in love with their invention and push forward without thinking this through – the end result is often a beautifully engineered product targeted for the wrong market, that does not succeed from a business perspective.  A little strategic thinking and market research at this stage goes a very long way towards improving the odds of success.
  • Development Phase 0: feasibility and early architecture – occurs in your own facility
    • In this phase, the technical team constructs a series of quick engineering breadboards, a.k.a. duct tape prototypes,
      • The objective of these prototypes is to prove the efficacy of the science and technology behind the idea.
      • Prototypes generated at this phase works somewhat like the final product but looks nothing like the final product.
      • Firmware is dependent on having a custom embedded platform to develop on – one way to accelerate this process is to use development kits from companies selling a key component (e.g. a development kit with a Coretex M3 chip in the same family as your final selection, or a Bluetooth development kit)
    • The team should also be working on early stage user experience and form exploration for industrial design to help drive system architecture decisions if and when they move on to the next stage. Sometimes the ID exploration might generate foamcore, foam, or printed physical models.
    • Primary market research should be strongly featured in this and all future phases – both discovery research (to determine who the customers are and what their needs and wants look like) and product research (to determine whether the proposed solution is usable and useful to end users, and saleable to economic buyers).
    • There should be an explicit gate review at the end of Phase 0 to decide whether to move on to the next engineering phase. This is often called the “Basis for Development” gate.
      • Two criteria are necessary to move forward: First, technical feasibility must be proven – you don’t want a science project in the middle of a development schedule. Second, market research must support the business case for moving forward with this product.
      • The big engineering spend starts in Phase 1 – a bit of planning and retrospective is well worth the effort to make sure the money is not spent on the wrong product.
  • Development Phase 1: Engineering Prototype – occurs in your own facility
    • In this phase, there is a coordinated effort for industrial design, mechanical, embedded engineering (PCB, cable harness development, firmware development) and software engineering to work together to create a fully integrated, looks like, works like prototype with production intent.
    • All the key questions should be answered by now – market receptiveness, technical feasibility, key system architecture decisions and more.
    • It is critical to incorporate design for manufacturing (DFM) thinking into the engineering development work at this phase.
      • The best way I know is to design every custom part with the target manufacturing technique in mind.  Then, start interviewing suppliers for each part, and engage top candidates in design reviews early and often.
      • Getting suppliers involved from the get go makes sure the parts that are designed are compatible with the manufacturing processes and the capabilities of the manufacturing partner in making those parts.  It is the single most effective method to minimize the difference between Phase 1 and Phase 2, accelerating time to market and reducing non-recurring engineering cost. It also starts building a relationship between your company and the supplier, who will be a key part of the process in the years to come.
      • On COTS components (aka purchased components): although this is “only” an engineering development phase and not “real manufacturing”, it is still a good idea to work with a supply chain expert to source all key components from OEM suppliers. You may be able to get the job done faster using Digikey as your supplier, but you will be guaranteed a massive redesign when you need to bring down the cost and end up having to switch out some components.  It’s much cheaper and less painful long term to get your cost down and DFM principles incorporated in this engineering phase.
    • Typically people create 1-5 prototypes at this phase. It is a first generation design and there are bound to be issues, whether with design errors, fit/finish problems, or outright functionality problems. It simply does not pay to make 50 copies when you might have to trash 49 of them to fix a critical issue.
    • Prototyping techniques are usually used to create custom parts for this phase.
      • For plastics, it may involve rapid prototyped parts (e.g. SLA, SLS, or 3D printed parts)
      • For metals, it may involve machining (even if the target manufacturing process involves casting or extruding the final part).
      • For PCBs there are a lot of quick turn PCB fabrication services who can create a small run SMT board for a fee.
      • Cable harnesses are usually produced using prototyping techniques as well. (What? There aren’t any cable harnesses? You are hand soldering wires? Stop and hire an embedded engineer with production experience immediately!)
    • Final assembly is typically done in house by the engineering staff – because:
      • Nothing comes together correctly the first time around, and the Dremmel tool is often needed
      • The assembly process is extremely instructional in highlighting design issues that need to be addressed, particularly design for manufacturing issues, and you would not want to outsource this learning to somebody external to your team.
    • It is good practice to have a design review at the conclusion of Phase 1.
      • The review should include a demo of the integrated prototype, running an early version of the production software.
      • Any findings from the assembly process, as well as product research in the field and in the lab (E.g. usability research)should be integrated to help generate a short list of things to tweak / change for Phase 2.
  • Development Phase 2: Engineering Verification Test (EVT) a.k.a. the “Pre-Production Prototype” build
    • The EV build is a design iteration of the engineering prototype that addresses a targeted short list of issues raised during the assembly, system integration and testing of the first generation engineering prototype.
    • The best possible outcome is an EV build that looks very much like the Phase 1 design.
      • This is because the larger the scope of planned changes, the larger the probability of new issues created by those changes.  In the worst case you could end up inserting a Phase 2.5 to fix those issues (prolonging time to market and burning up more of your cash runway than what you can afford – resulting in an unfavorable fund raising situation to bridge the gap).
    • Some people make up to 50 copies of the product at this stage to facilitate effective field testing.
    • Prototyping techniques continue to be appropriate. At higher volumes (above 10 copies), crossover low volume manufacturing techniques like urethane casting for plastic parts become very appropriate.
    • Final assembly is still done in house.
    • At the end of Phase 2, there should be a clear gate to show a demonstration of the EV build, review findings from field tests and early durability tests, and make a go/no-go decision to release the design to manufacturing. This is sometimes called the “Basis for Production” gate.
      • This is a particularly important gate that needs careful consideration, since this is the crossover point where the program shifts from an engineering focus to a manufacturing/production focus. Ownership of the program often shifts from the engineering department to the manufacturing and operations department.
      • The supply chain, including the contract manufacturer, has been assembled and will become very active in the program after this point.
      • Also, prior to this stage, the engineering expenditure is limited to the engineering development and prototyping costs.  After this point, if a decision is made to proceed, one will need to order tooling – which is a significant capital expenditure that deserves due respect.
  • Manufacturing Phase 1: Design Verification Test (DVT) – occurs at the CM’s site
    • The parts used in the DV build should nominally be physically identical to the EV build. The only difference is that the EV build is typically assembled by in house staff, whereas the DV build is assembled by the contract manufacturer’s staff.
    • Prototyping techniques are still appropriate for this phase, because the tooling cycle takes months and fully tooled parts are typically not all ready by the time the DV build needs to be assembled. That said, some early parts might come in on time for this build.
    • The primary purpose for the DVT is to teach the contract manufacturer how to build the product, and in so doing, identify any remaining DFM issues to be addressed. There should be little to no differences from EV to DV.
    • There are several activities unique to the DVT phase:
      • The CM uses this experience to work with you on process development – inventing how to make your product repeatably and with a high quality of output. This is why it is paramount the design should stop changing at this time. Any changes need to be minor; all changes must be tightly controlled and documented via an Engineering Change Order process.
      • The DVT phase is also the  time to develop assembly fixtures, as well as any test fixtures and manufacturing test and calibration software to be used during the build process.
      • DV prototypes are often appropriate for use in regulatory testing.  For instance, if your product plugs into a wall outlet, you will need a safety mark from a Nationally Recognized Testing Laboratory (NRTL) which will test your product to the appropriate safety standards.  If your product has a radio inside, you will need emissions testing to meet FCC requirements and get an FCC number, which you are required to display on the product. If you are selling your product in Europe, you must have a CE mark.  There are exceptions to this rule, but this is generally non negotiable for consumer electronics products sold on sites like Amazon or stores like Best Buy and Apple.
    • At the conclusion of the DV phase, there is typically a PV readiness gate review.
      • The engineering design, manufacturing test systems and manufacturing processes are examined for readiness to move on to the next production phase.
  • Manufacturing Phase 2: Process Verification Test (PVT) – occurs at the CM’s site
    • The PV build tests out the manufacturing processes invented in the course of the EVT and DVT phases.  In other words, there should be no engineering design changes from EVT to PVT – it’s all about process from here on in.
    • The PV build will involve parts made via the final manufacturing process, since those processes are necessary to create high quality parts with good fit and finish to support a saleable product.
    • Depending on how the preceding phases went as well as the lead times for custom tooling, this phase could be as short as a few weeks or as long as a few months.
    • Once PVT is complete, there is a final gate review where you and your CM make the decision to move forward into the ramp up to mass production.
  • Manufacturing Phase 3: Mass Production (MP)
    • If you get here, congratulations! You made it!
    • Now expect to spend the next 9 months working on sustaining engineering to stabilize your product on the manufacturing line and get to that high quality of output you expect to see.
      • This sounds depressing, but it’s simply a numbers game. You will have found and fixed a lot of “infant mortality” issues in both the design and the process early on, where the finished goods simply does not work out of the box. After that, it’s low incidence defects that will hold you enthralled for months.
      • If there is a defect that occurs with a rate of 1 in 1000, and you make only 200 of your product a week in the first 3 months, it will take you months to properly characterize the defect and come up with a mitigation measure.
      • This is normal and perfectly to be expected – every hardware manufacturer goes through this, even Apple. (Remember Antennagate?) So plan and budget for it, and don’t get annoyed. It’s a normal part of manufacturing bring-up.

If I haven’t convinced you that hardware development is different from software development early on, hopefully you have no further illusions by now. Hardware is a lot more work, takes a lot more time, and uses up a lot more capital. But that’s ok!

At the end of the day, we interact with physical objects and it will always take innovative people to continue to create great products to serve our needs.  Knowing about the phases and work content should help you better plan and budget for all the activities that will take your innovative hardware product to market.

 

9 Responses

  1. […] 5. Hardware Product Development Process @chenelaine […]

  2. James Poli
    | Reply

    Excellent overview. Having worked for a couple small brand manufacturers, this is exactly how things work.

    • elaine
      | Reply

      Thanks!

  3. Star Simpson
    | Reply

    This is a really great and thorough post, thank you for writing this!

    • elaine
      | Reply

      Thanks!

  4. Jon Nordby
    | Reply

    Errata: there is a duplicated paragraph starting with ‘When you are looking at a…’ (just before the details)

    • elaine
      | Reply

      THANK YOU for pointing that out! Fixed! Appreciate the heads up.

  5. Alan Jones
    | Reply

    Really good summary. I highly recommend in the Ideation phase, sketch the first ad campaign and get sign-off from everyone. Plaster the walls with posters selling the future product.

    The purpose is to really clarify what important new feature(s) or benefit(s) are needed to get the win. I’ve seen too many products fail because sales, marketing, designers and engineers weren’t on the same page about what was really critical for success.

    Lead Engineer: “We did it! The anti-gravity widget works!”
    Product Manager: “You idiot, it has to be blue!”

    • elaine
      | Reply

      That’s funny! I haven’t done the first ad campaign but I have done the first web copy to describe the product. That is especially handy when testing value propositions with landing pages or FB ads or Google ad word experiments.

Leave a Reply