Get Your Tech from a Lab? Be Careful

Introduction

One area of #entrepreneurship that receives a lot of media attention is the launching of new products from university or federal laboratory research. While this can be done successfully, the annual new company spinoffs in this area is dwarfed by the number of commercial startups. This way of going forward with your idea can be very rewarding- if you know exactly what kind of technology you need. In later posts, I’ll talk about some of the university and lab technology commercialization I have been involved with. In advance, my limited experience is of not much statistical value but I hope to help you understand what can go right and what can go wrong with these efforts. In this post, I’ll talk in more general terms about tech commercialization and some products that were successful and others not so much.

Please keep in mind that you can do this and you don’t need to be a Ph.D. in the underlying science but you must be sharp enough to vet any prospective lab technology against your needs. Many of the federal labs in the US have tech transfer specialists that can guide you. A former acquaintance of mine, David McFeeters-Krone, founded a company to do this. See his company here- https://www.intelassets.com/About-us.php. And also see my later post on open innovation.

The Main Idea

Technology Commercialization- Proceed with Caution and Skepticism

Technology commercialization is the most challenging of all the ways to start a new company and I maintain that it is very risky for any startup. And, I (my opinion here) don’t think your degrees from elite universities or working at Fortune 500 companies makes a difference. It really should be attempted at the most pragmatic way possible. Why? Because tech commercialization is a triple threat- new manufacturing process that often involves new equipment or techniques or scale that has not been attempted before, an innovative new product that has unknown reliability, and displacing existing entrenched competitors. Risk management should be at the top of everyone’s list.

There are many unknowns that can combine in this space to create what are known as a “wicked problem” (abandon hope all ye that enter here according Dante’s Divine Comedy). And finally, there is the question of financial success- did the product recover its initial investment and lead and the firm to greater sales and profits? No, technology commercialization is not literally passing through the gates of Hell, but it can sure feel that way when your team is beset with technical demons in every imaginable area. Enough of the metaphors (there are many more), lets see what might be learned. From an engineer’s perspective, marketers are probably located in the 9th ring of Hell with the lawyers (ok, could not resist one more).

Example 1- The TI Digital Micromirror Device

Lets start with a qualified success- the digital micromirror. This was a micro-machined device, very elegant, that enabled the projector business to shift away from high powered halogen lamps and focusing lenses to a small device that could be precisely controlled electronically. Texas Instruments (TI) , a key competitor of Motorola during the time I was there, had great success with this device. But, TI was a very large company with very deep technical resources and financial pockets. There are many very detailed product histories written by experts about this device. Today it is a licensable product but still has some inherent problems with reliability , based on the material available to the public. But in general, I would argue that this was a market success but did it ever make a profit for TI? Hopefully some readers can answer this question.

A cursory review of the many publications about this amazing technology reveals many design and innovation platitudes, including a very nice writeup at the American Society of Mechanical Engineers (https://www.asme.org/wwwasmeorg/media/resourcefiles/aboutasme/who%20we%20are/engineering%20history/landmarks/dmd_brochure.pdf).

But profitability? Dr Hornbeck may have won a Oscar for his design and the amazing products that followed but it is not clear if the DMD device made money for TI. Was it worth it? What is the lesson for entrepreneurs? Perhaps we can gain some insight from the next example. Keep in mind that as an entrepreneur, betting your company on a single product success is a very risky bet. TI, unlike your startup, could afford to capture incremental or spinoff revenue from other product lines that were enabled by an unprofitable DMD.

Example 2- Analog Devices Accelerometer

I joined Motorola during the silicon micromachining (MEMS) “wars” (1994). During this time a number of smaller, venture backed companies were doing well with their pressure sensor technologies and a few other niche products. Since just about any semiconductor fab line could make these devices, it presented a very enticing new market to enter. MEMS devices offer extraordinary improves in speed of response, smaller size, and much lower cost than conventional devices. For example, the first airbags were triggered by a “ball in tube” device- no, I’m not kidding.

Analog Devices had beaten Motorola to the accelerometer market by licensing the UC Berkeley (UCB) design that used a very simple “lateral” or planar MEMS accelerometer. The UC Berkeley team had been innovating for years to make all kinds variations of this design. When Motorola entered the market late, with their own design, the competitive battlefield was already set. The Motorola device was elegant on its own but the UCB/Analog Devices device much simpler and easier to manufacture, complete with a patent picket line that protected the design from encroachment.

What’s the lesson for entrepreneurs? First to market wins? UCB is smarter than everyone else? Motorola was (at the time) too much in the camp of “not invented here”? My perspective is that there is probably a combination of things that went “right” for Analog Devices and yes, UC Berkeley was first out of the gate with a device that was designed for manufacturability. The other criteria is probably that the performance of the UCB design captured enough of the essential elements of a successful device (stability, sensitivity, selectivity, repeatability, reliability, and reproducibility) to capture many of the design slots available.

Perhaps one of the lessons is that the technology approach needs to be reconciled with the needs and the reality of the marketplace. Selling an accelerometer into the automotive market required extraordinary efforts to meet the low price, high volume, and reliability requirements.

As a senior with a startup idea, this is probably not the place to start.

Example 3- IBM’s non-contacting test technology

In my final year at Keithley Instruments, I was lucky enough to be involved with tech transfer of a very impressive technology developed by IBM’s Semiconductor Group. Up to this point, all testing of semiconductors was done at specific test sites scattered across a nearly finished silicon wafer. To test the wafer, a set of metal probes would be brought carefully into physical/electrical contact with the bonding pads. A set of instruments and a switching matrix would allow the programmer (me for example) to conduct just about any kind of electrical device characterization test one could dream up. The problem with this approach was that this testing did not occur until 90% of the water was complete and a layer of metal had been deposited. If something had gone wrong at an earlier stage, you probably would not be able to detect it. In semiconductor processing, each new layer interacts with the previous layers.

This new technology allowed the testing system to approach the wafer and position itself just over the surface. Another device created a small conductive layer of ions on top of the wafer without leaving a mark. Now, with an electrical connection so-to-speak, one could conduct many basic device physics oriented tests on a real product. Wow!

Did this product ever make a profit for Keithley? I think the answer is probably a resounding “no”. But the no answer does not suggest that the people were the problem but the way in which the company evaluated and managed risks might be one of the primary culprits. One of the challenging interplays for technology commercialization occurs between the three competing forces internally:

a. Marketers that develop the forecast for the new technology platform and its product spinoffs, based on capabilities provided by the development team

b. Financial controllers who use inputs from the development team and the marketing forecast to develop an actionable financial pro-forma analysis.

c. And the engineers charged with creating the saleable products.

You can get into this dreadful loop where marketers forecasts are too low to justify the spending needed by the engineers. Inflating the market forecasts to “make the spreadsheet work” needs to be critically scrutinized before you pull the trigger on committing millions of dollars to bring an emerging technology and product to the market.

Senior management will be sorely tested here. Nowadays, many projects of this scale receive incremental funding as they hit agreed upon milestones, not carte-blanche to spend until you get the product launched. But beware: strong willed or over aggressive senior managers that attempt to push projects through over the objections or concerns of the people doing the work is a clear red flag. In Keithley’s case ( I was not there when the product launched), better risk management by the senior staff might have give the new organization created to launch the task some time to gel into a first rate development team. This suggests that in these large scale commercialization efforts, organization dynamics and the rules they create for themselves may play a deciding role in the success or failure of the effort.

Example 4- Tin Oxides Sensors Take 1

I left Keithley during the ramp up for the IBM project and was recruited by Motorola to work in their micromachining group[1], a major interest of mine since its inception. At Motorola, our task was to commercialize a thin film gas sensor technology that had been developed by a research group based at a Swiss university. This material, tin-oxide, was well known to material scientists as having some ability to detect the presence of some gases such as carbon monoxide and methane, and a few others. It was unable to detect one most important gases- carbon dioxide. But, it was very non-selective: responding to many other nuisances in the air. A Japanese company had been making so-called “bulk tin-oxide” sensors for years. Bulk means they were not using the conventional planar semiconductor manufacturing process. Their sensors were much larger than our tiny thin film sensors, but they had found ways to coax the material to do what they wanted.

A research group in Switzerland had developed a thin film process, compatible with semiconductor processing, for manufacturing these new sensors. Many wonderful papers were presented at conferences. But one thing struck me- if this was such a great technology, why didn’t the large, existing sensor companies snap it up immediately? Motorola decided to jump in with a technology development effort In Motorola fashion at the time, they brought every possible resource to bear on this new project. I suppose the thinking was even though we were going to spend ourselves into a deep hole to bring this product to market, the size of the potential market suggested we would recover our investment and go on to make substantial profits for a while. But the market analysis I did not did not support this level of spending.

When I examined the Sensor Products Division’s financials for the first time- I was pretty horrified at what I saw. We had negative gross margins before we launched chemical sensors and they were not improving. But how could this be? Motorola had every possible resource and infinite patience (at the time) from the CEO and division managers- all very wonderful people to work with. See my later point about open innovation ( “not all the smart people work for you”).

I was lucky enough to be the strategic marketing person in the US with a very experienced engineer from France brought in to Motorola to run marketing in Europe. We spent a lot of money and good-will trying to get this technology off the ground and get the sensors to behave. But the technology fought back, frustrating us in many areas. The worst area we were hit with- reliability, selectivity, and repeatability. Lack of reliability was an anathema to any semiconductor company. Selectivity meant false negatives and positives and poor repeatability meant we could not measure the same thing multiple times and get the same answer. This effort was eventually shut down after I left Motorola. But what went wrong? Were there signs that could have helped us be more focused or careful in our approach? Perhaps the answers may lie in the formal and informal institutions (the “rules”) of how Motorola conducted its business.

Example 5- Tin Oxide Sensors Take 2

Apparently, my experience at Motorola with tin-oxide was not enough. When I realized that our family had to move again for better air quality, this time to the Portland area, my former colleague and the manager of the Motorola chemical sensor efforts recruited me to try again with the tin-oxide process but this time we would locate the business in Switzerland near the lab where the technology had been initially developed. I was a bit skeptical at this but jumped in as the VP marketing, determined to do my best. But- had anything changed in the underlying technology? Our little startup had only raised about $6M- Motorola could spill that much money on the floor every week and not notice.

At this new company, we did not have Motorola’s global reach, infrastructure, or other abilities. We did have a pretty good idea of the European markets. One of the key markets was the so-called “cabin air quality” sensors for automobiles. But selling to the automotive markets was dreadfully challenging- extreme reliability requirements, low average selling price (ASP), and long design-in lead times. Further, a new technology, like ours, would only appear in a few high end luxury models before they might move into the mainstream high volume models. As you can imagine, this meant that you had to stay in business for years, perform at the top of your game, hold on to your cash until the auto companies final started ordering volume. If this product is your only source of income- you are taking an enormous and probably unjustified risk. You can imagine what happened. This effort also eventually did not create major win for investors or the employees.

Example 6 Humidity Sensors

During my time at Motorola, I met a very interesting French engineer who joined the chemical sensor effort with Motorola France, mentioned earlier. This person had extremely detailed knowledge of humidity sensors and their testing (quite difficult). While we interacted somewhat, I think he probably understood more about sensors and their manufacturing and marketing than most of the people at Motorola. While he had spent time at a much smaller firm, he had a really good grasp of these ubiquitous sensors. We parted ways after I had to leave Motorola Phoenix for cleaner air. We did interact a few more times collegially but then as direct competitors. So, lets dig into this sensor technology and see if it is just a reprise of the 90’s MEMS wars or is there something more to learn.

Humidity sensors are critical to nearly every kind of gas measurement. They are also very important as uncontrolled humidity excursions can create all kinds of problems from mold (high humidity) to static electricity (low humidity). The Humirel company had developed an extremely small humidity sensor that also integrated some logic and signal conditioning on a single chip (so-called “smart sensor”). They were not a large vertically integrated company like Moto or TI or Intel but very skillful in how they carefully approached the applications for their device.

On the competing side was a very small company based in the San Diego area (Hygrometrix). Hydrometrix’s founder was a very clever and non-traditional fellow to run a sensor company. He had made a limited success with an extremely sensitive humidity sensor that used part of an African violet as the transducer (a transducer converts one kind of real world signal into an electrical signal). He grew the plants on his property, harvested the flowers, and through a closely held process, created this elegant strain gauge-based humidity sensor that was purchased by the US Department of Defense for use in the Navy. But this business eventually went away from Hygrometrix and they attempted to pivot (essentially a 1 person company) by developing a micromachined humidity sensor that combined an pressure sensor piezo electric strain gauge foundation with a chemical coating that expanded and contracted with the change in humidity. To build one of these, you would have to work through a semiconductor foundry, use some of their pressure sensor process and then coat the pressure sensor with the humidity sensitive polymer.

The great advantage of this design was that anyone with a MEMS pressure sensor could make this device. The major disadvantage of this device though, was not well known, including me. Motorola had taken an interest in this device but had passed on the licensing. No other firms lined up to license the device. Why? This unanswered question would prove to be the undoing of Hygrometrix, despite the heroic efforts of many people.

While I was between companies, I was approached by Hygrometrix to try and restart their sensor business. I called by best friends around the US and begged for their help. We raised a bit more money from the primary investor in the company. But inside, we were really struggling with the sensor’s tendancy to drift (not a good thing). My best friend Ed and I finally got to the heart of one of the problems of drift. The device, as I found it, was mounted to a chip package using a very brittle type of adhesive. The mounting process induced all kinds of unwanted stress and strain into the sensor element. Once we finally figured this out and tested the devices with the much softer adhesive, the device immediately responded much better. But by this time, my personal relationship with the late founder had deteriorated to the point where I left.

During my final days however, I also found out that all of the problems we had been experiencing with this device had already been seen and solved by many MEMS companies- a source of intense frustration. As a final kick in the shorts, I went to Brazil to represent the company at a humidity conference (yes they exist) and got publicly excoriated by some European engineer who interrupted my talk by shouting out that his company had tested this polymer and it had uncontrolled drift and that they had discarded this approach years ago. The chairman of the that session had to tell this guy to shut up. While this guy was first rate you-know-what, he was right. Wow, talk about a long trip home.

After I returned to the US, I remembered something that the Japanese company making the bulk tin oxide sensors had told me privately while I was at Motorola. They asked me why didn’t you ask us about tin oxide? We would have told you it was not the way to go. They had tried many times to make MEMS version of tin oxide but could not control the device’s tendancy to drift.

What’s the lesson here? Large company and small company make the same mistakes? If so, that would suggest that it is not a people problem per se but a process problem- how to thoroughly and honestly evaluate the capabilities of potential technology platform before committing?

Example 7. HP’s Nanowire and Sensors

In this last example, I want to talk briefly about my interactions with Hewlett Packard (HP). I had admired HP for many years and competed against their instruments division (now a separate company called Agilent) around the world while at Keithley. After making a short talk in Portland about sensors and sensor tech, an HP employee contacted me and asked if I could do some consulting for them in sensors. I agreed and began some work with this group. Talk about a dazzling array of scientists and engineers- I was quite intimidated by the R&D folks that I was interacting with.

HP, at the time, was engaged in a global battle to bring quantum effect technology to market. One of the potential markets that was identified was sensors. They were very clear at the onset- any market opportunity had to be in the $B range and needed to gain at least 20% market share or they were not interested. I explained repeatedly that those applications did not exist or were already dominated by Analog Devices and a few others. We worked together to try a define a market where the, as yet, undefined quantum effect sensors, would be used. I did several presentations outlining the state of the sensor market and some of the pitfalls and opportunities might lie. The net result is that HP did the right thing. They passed on the sensor market. They could have jumped in with 100% enthusiasm and 10% market knowledge but they did not. The management at HP seemed to have very well disciplined approach to bringing new products to market. A success? Yes, think so as they did not embark on “Mr. Toad’s Wild Ride”.

Summary

Some of the lessons that I think might be gained from this very cursory analysis are:

a. New tech needs to be very aggressively vetted against the key criteria for market success. This means it must be done by a multi-disciplinary team- not the inventors. Don’t be afraid to say “no”.

b. The market decides your success, not the “coolness” of the technology. But the market also defines the parameters or boundaries of where you need to operate.

c. Make the technology proponents justify their claims and show it with data, hopefully with third parties evaluating its efficacy- not the inventors and engineers. You already witnessed this first hand with the rollout of the COVID-19 vaccines that are based on a new technology pathway.

d. Know in advance- what the criteria is for launching a new product and have well established rule for engaging in an advanced technology commercialization.

e. Have checkpoints, meetings, or key milestones to guide the release of supplemental funding- no carte blanche.

Some of this is probably very familiar to you if you have had some format new product development training and experience. The problem with new tech vs a new product though are the large number of unknowns.

[1] The Sensor Products Division, my home at Motorola, made accelerometers, pressure sensors, and of course chemical sensors for a brief time.