Ensuring Research Results in Manufacturable and Scalable Products

(related sections in the author's DFM book are noted in parenthesis)

Most research starts out just trying to prove an idea will work. And then what?

Once it “works” most managers and venture capitalists usually try to rush it into production. And how well does that go?

The “Valley of Death” between concepts and viable products

The official page of the Breakthrough Energy Coalition (led by the most famous high-tech leaders) is quoted as saying

“Experience indicates that even the most promising ideas face daunting commercialization challenges and a nearly impassable Valley of Death between promising concept and viable product. Neither government funding nor conventional private investment can bridge this gap.”

The "Unsurpassable Mountain" after the Valley of Death

And then, if the product does make it into production, it becomes obvious that it costs too much. So then it is time for “cost reduction.” But this site teaches all the reasons why cost reduction after design is so hard to do especially if the cost metrics are primarily based on part cost, which is typical.  However, just trying consumes a lot or resources and time while all the changes attempted (usually cheap parts) cause quality problems, make it hard to scale up to stable production, and may also degrade functionality, all of which causes more firedrills to do even more changes.

Why does this happen so much?  

One reason is that some of the most popular “Phase/Gate” processes acrually skips from “Concept Testing” phase to “Prototype Testing” phase without any product design phase in between!   Other "project management" oriented processes are just so rushed that there is not time to do good design.

This tells engineers that once a proof-of-principle or concept experiment “works,” they should skip directly to building a prototype, which will cause all the problems cited in the following article when products are designed strictly for functionality.  The white paper then shows solutions: Concurrent Engineering for Challenging Products.  

If your company has such a dysfunctional "process," don't compromise your research or wait for the process to be changed.  Instead, do everything on this page -- and this site -- in your project is your own "micro-climate" as recommended  at the end of this page

The following methodologies show how to avoid this “valley of death,”

Research determines product architecture

The definitive book on DFM (in Figure 1.1, right) says that 60% of cost is
determined by the product architecture. Proofs-of-principles or
even experiments can determine or imply the product architecture or
limit its options.

    As the graph shows, the only way to achieve ambitious cost goals is through concept
 breakthroughs. The article, Designing Low Cost Products shows examples of cost
 breakthroughs for two types of products:.

Electronics costs can be greatly reduced by specifying higher levels of silicon integration, eliminating manual wiring,, modularization (see below), combining circuit boards, and, if not possible, replacing all circuit board connectors with flex layers between boards. Most of these will require working early with vendor/partners, as discussed below.

Large Structures
costs can be greatly reduced by replacing expensive welded structures by more manufacturable assemblies of automatically machined parts
that are assembled rigidly and precisely by various DFM techniques.

Similarly, other important research, like concentrated solar power (CSP), would need breakthrough concepts to solve the biggest cost challenges, like the current need for hundreds of thousands of two-axis servo mechanisms to reflect predictable sun rays to a stationary tower.


When to do Innovate

   The time to do innovate is in the Concept/Architecture
phase shown in the recommend time-line, which is described at:

What to do in the Research Stage

Concept selection. Don’t just jump at the first idea that comes to mind or whatever is easiest to demonstrate, like Edison’s cylindrical phonograph, which he did launched into production despite the fact that cylindrical records could not be stamped out like manufacturable records.

Verify feasibility digitally through simulations and analyses of:

- all aspects of functionality
- in all anticipated environments
- with specs that are all needed and achievable
- for the anticipated market variety
- with potential risks mitigated with tools like Failure Modes and Effects Analysis, which can be used to compare competing concepts
- that can be built with readily available parts on widely available processes with production skill levels (as discussed below)
- using readily available tolerances.

  • If necessary, scale back non-essential functionality, specs, and product variety using product platform design .

  • If necessary, re-formulate the product concept to make the feasibility more viable.

  • If the above consideration narrows the market, re-evaluate the business plan now before trying to build a non-manufacturable product.

  • ensure that the final product will be manufacturable and scalable as recommended below.  Always remember:

There are many ways to make something work;
There is only one that is the lowest cost.

Prioritization. (Section 2.2.1 in the book) Focus design efforts on what is most important to customers and get the rest off-the-shelf (Section 5.18).

For Electronics  For instance, customers don't buy your products for the power-supply.  But they expect it to work all the time. So, instead of wasting valuable time designing anything that is hard to design and has nasty failure modes (smoke, fire, system failure), specify proven off-the-shelf power supplies that have proven "track records" in your industry.  

This needs to be done first, because off-the-shelf power supplies come with voltage values
that are prevalent in your industry and you will need to design your product for those voltages.

For Electronics.  Similarly, don't consume valuable resources -- and jeopardize quality -- designing routine electronic functions that are readily quickly available off-the-shelf as modules, sometimes called "single board computers," that have standard interfaces, for instance, for:

  • Processing: off-the-shelf processing boards are available for every industry
  • Memory, which can be increased with plug-in modules, even in the field.
  • Input/Output and communications, possibly based on USB, Ethernet, HDMI, or VGA ports
  • Motion control for actuators, motor axes, sensor inputs, etc.
  • Data acquisition and number crunching.

Some off-the-shelf modules come with thoroughly debugged software, which will free SW engineers to focus on your product's unique software.

For Industrial Equipment.  Similarly, don't waste valuable resources designing  anything that is readily available off-the-shelf such as:

  • guards, doors, and shields
  • stairs, railings, and platforms
  • material handing devices, dispensers, conveyors,
    even palletizes
  • mechanism controllers, actuators, and sensors
  • cable assemblies using standard connectors and cables
  • cabinets, enclosures, partitions, fans, doors, latches,
    and locks, which are available from catalogs or can
    quickly be built to-order in many standard sizes.
    Two companies that have vast selections of electronic
    cabinetry that can build them to-order:
        Above right: image from  Hoffman brand of Pentair:
        Below, right: image from Cooper Industries:



Off-the-Shelf parts will actually cost less because the parts and their tooling is already designed.  If this is not immediately apparent, then "cost" must be defined as total cost (Chapter 7) See total cost web page. 

The paradox of product development is that off-the-shelf parts must be chosen first 
and then the product is literally designed around them.

For everything else that really needs to be designed, project teams need to do the following:

Part Availability:  The very first experiments or proofs-of-principle must be based on readily available parts and materials. If not done early, there may not be enough parts available for production. Do not count on changing parts for availability when going into production because these difficult changes take time and resources to do, and, even worse, introduce many new variables that  degrade quality and even compromise consistent performance

Parts from Bins or Lists.   Do not pick parts from just any part bin in the factory or any entry on an "approved parts list" because most of those are still there for legacy parts, which have obsolescence challenges. And even if you pick a part value that is used on current products, you may pick a duplicate part number, which may not be as available as the most common part number.

Proven parts. An important way to minimize risk and assure success is to specify only proven parts, materials, and processes that have their own “track records” that can be the basis for reliability calculations

 Inherently scarce parts.  Do not  base research on “rare Earth” elements or single-source suppliers or availability only from one country. And to avoid resistance later, avoid materials that have regulatory challenges or  could be toxic to people or the environment. 

Rather, select parts from standard parts lists (Section 5.8) that have been approved for new product designs. 
 If not already done, you will have to create standard parts lists for your research hardware,
especially for any that may have availability problems.

Rescue Parts.  Do not tune a design or rescue something that is not working with unusual parts with many increment values. like shims (often done in .001" increments), resistors (from racks of 1 ohm increments), crystals (with whatever frequency that makes it work), coils (with whatever number of windings that makes it work), obsolete component packaging (e.g. lead-through parts when products are now manufactured on Surface Mount Technology equipment), or any other unusual parts or obsolete technologies.

Rather, make your design robust enough to work with readily available standard parts
that can be automatically assembled and soldered. 

Questionable Sources.  Do not pick parts from hobby shops, hobby part sites, lab equipment catalogs (that have vast inventories of non-production parts), or surplus warehouses (which is the basis of many silicon valley legends).

Rather, work with your Purchasing Agents to find good parts from their best suppliers

If you can’t make it work with easily available parts from good suppliers, the research may not be feasible.

Achievable Tolerances. The very first experiments or proofs-of-principle must be based on tolerances that are routinely achievable in production environments.  If not done right, products will always be unnecessarily too costly and hard to build because tolerances are so hard to loosen later.   The root cause of this is the common temptation to do whatever it takes to make  one  proof-of principle “work,” so  there is pressure to specify tight-tolerance parts. This brings designers 4 immediate acclaim, but will doom the product’s chances for cost effective production.

Optics and Systems Needing Precise Alignment.  Most research efforts for optics and lasers prove it will work on a precision ground marble slab with precision mounting blocks for all the mirrors, lenses, and sensors, which are tediously adjusted by skilled technicians until it "works."   If that is just thrown over the wall to manufacturing, the "product" would have (a) many tight tolerances specified for all key dimensions and (b) onerous, slow, difficult, and costly alignment procedures with high skill demands (next point). 

Rather, the architecture  should be optimized  to provide the necessary tolerances at the least cost with the least skill demands.  This can be accomplished with Guideline P14 (in Section 9.2 in the DFM book), which shows how to fabricate many dimensions at tight tolerances by laying out the architecture so that all dimensions can be machined on one part in one setup (one chucking) on a multi-axis machine tool, like the first illustration on the Flexible Manufacturing article at: .  However, this must be optimized in the architecture stage to ensure all the critical dimensions are on one part and dimensioned properly.

If you can’t make it work with achievable tolerances,, the research may not be feasible.

Skill Demands. Similarly, the other half of doing "whatever it takes” is to use highly skilled technicians, who pride themselves at being able to make anything work, even unmanufacturable designs. The result is that such designs will have low quality, slow throughput, and may not work consistently without the original skill levels, which would keep cost high. 

For example, instead of designing the usual hard-to-build welded machine frames that have high skill demands, design assemblies of parts that are automatically machined on CNC machine tools and then assembled precisely and rigidly by various DFM techniques, as described at the following page.  An added bonus is that such a frame can be designed to be a back-ward compatible "drop in" replacement for current products.  If this is done first, it can actually help fund the research and then become the basis for the proof-of-principle.  All of this is described at: , which has many illustrated examples.

If you can not make it work with ordinary skills,, the research may not be feasible.

Widely Available Processing. The very first experiments or proofs-of-principle must be based on widely available processing equipment, like ordinary CNC machine tools. Needing unusual, extra precise, or custom-made machine tools or processing equipment will raise costs, show deliveries, and even hamper scalability.

If you can only build it on ordinary machine tools, the research may not be feasible.

Concurrent R&D. Concurrent Engineering should start early to ensure that research will be manufacturable and scalable (Chapter 2). The  web-page on Scalability shows how to use Concurrent Engineering to design scalable products.  The essence of Concurrent Engineering is working early with Manufacturing and Purchasing people early to avoid the all the problems that come from designing in isolation and "throwing it over the wall" to the factory, which is discouraged at best-practice companies:

Paul Horn, who oversees research at IBM says: “Everything we do is aimed at avoiding a ‘handoff’ -- there is no ‘technology transfer.’ It is a bad phrase at IBM.” Research teams stay with their ideas all the way through to manufacturing.".

Independent research labs in the government or universities should pre-select the best manufacturing company for the same reasons that the best of the best practice companies use pre-selecting vendor/partnerships when a manufacturer needs to outsource custom parts.  See Chapter  10, "Fully Integrate Suppliers into the Product Development System," in the book,  The Toyota Product Development System (Morgan and Liker,  Productivity Press, 2006).  Both the linked article and this book prove that vendor/partnerships result in less total cost and scale up faster than starting the design work alone and then sending that out to the low-bidder

Don’t offshore production ecause, in addition to not saving total cost , offshoring will thwart concurrent engineering if factories are not even working at the same time. This means that your engineers must design everything in isolation, know how to design for manufacturability, and throw perfect and complete drawings over the ocean and then the Contract Manufacturer will “build-to-print.”

Conclusions on Time and Resources:   Doing all of the above  may take a little more time and effort up-front, but will avoid many more times the months, and resource-hours later trying to fix the design with changes after so much is "cast in concrete" and boxed-into-many-corners.  

Conclusions on Cost.   If research is done "on the cheap" with inadequate "seed money" or trying to getting venture capital funding with anything that "works," the result will most likely be lost investments at  the  “Valley of Death” quoted above.


Implementation at manufacturing companies:  Even before this is incorporated into the company product development process, all of this can be implemented  immediately by a project in its own microclimate, (introduced in Section 11.7.2) in project "obeya" rooms (Section 2.7.2  in the DFM book).  Manufacturing companies working early with research labs can work early together in a project room at either the research lab or the manufacturing company.  In fact, Toyota uses the traveling Obeya to "move downstream to the plant as the program moves downstream" (page 263 in The Toyota Product Development System).

Consequence of not doing this:.

If all the above is not done proactively in the research phase, then un-manufacturable proofs-of-principle or prototypes will then have to be commercialized which preserves the research "crown jewels"
and then re-designs everything around them for manufacturability, as shown at

See a  list of Commercialization clients listed near the bottom of:

n customized seminars and webinars, these principles are presented in the context of your company amongst designers implementers, and managers, who can all discuss feasibility and, at least, explore possible implementation steps

In customized workshops, brainstorming sessions apply these methodologies to your most relevant products, operations, and supply chains.

For questions or comments about any of this contact:

  Dr. David M. Anderson, P.E., CMC 1-805-924-0100; e-mail:


copyright © 2019  by David M. Anderson

Book-length web-site on Half Cost Products:

[DFM Consulting]    [DFM Seminars]    [DFM Books]    [Credentials]    [Clients]   [Site Map

[DFM article]     [Half Cost Products site]   [Standardization article]   [Mass Customization article]   [BTO article]   [Rationalization article]