2021-02-08 7 min read


Notes, 2021-02-08.

Last week I got a tip that Dragon Innovation, Avnet’s supplier vetting & manufacturing oversight business, shut down last October. Founded in 2009 by former iRobot employees Scott N. Miller and Herman Pang, Dragon (along with Bolt, the venture fund with which it was closely associated) was a mainstay of the hardware scene and did a great deal to anchor Boston’s status in it.

(They were also a sponsor of this newsletter, a vendor to my side business, and had raised venture funding from my former employer, Undercurrent. And, I consider Scott a friend.)

Avnet, a publicly traded company with a market cap of $3.7B and annual revenue in the neighborhood of $20B, confirmed by email that they are “making some strategic changes to streamline our ecosystem offerings.” And streamline they have: Dragon’s website and the years of content they had produced (much of which has appeared in this newsletter) has been summarily delisted.

This news came to me as a bit of a surprise, and I reached out to Scott N. Miller (who’s now working under the name Dragon Ventures) to hear how he viewed more than a decade’s worth of hardware development history. An edited version of our conversation is below (with weekly links, as usual, ⏬ at the bottom).

-Spencer Wright

The most clicked link from last week's issue (~16% of opens) was on Bolt’s (now closed) Boston offices.

An interview with Scott Miller.

SW: What did Dragon Innovation do? What was your business model? SNM: Well, one of the challenges I had with Dragon Innovation was that I had never quite figured out the right business model. I was most passionate about being transparent to our customers and not being a middle man; I wanted our customers to issue the PO straight to the factory, so that if they worked with Dragon it was because we were doing something useful and not because we were holding them hostage. I’m not sure the exact number, but we worked on between $100M and $500M in gross purchases over the years and we never ever took one penny from a factory - hidden or obvious.

SW: So, how did engagements with customers typically work? SNM: We generally worked on a retainer model, billing between $5K and $30K a month depending on the level of effort. We would provide feet on the ground in China, embedded in the factory - frequently a whole team, structured so that it was in effect an extension of our customer’s team. We would manage and solve all of their manufacturing challenges. There was always a need for a project manager, but we could easily cover the additional needs that arose - quality, sourcing, translation, etc. I would say that our services were equivalent to hiring a VP of engineering, and we billed accordingly - maybe an average of $15K a month.

SW: Oh, that’s interesting. So you were eliminating this key hire that they’d otherwise need to make, and charging roughly what that person’s salary would be - but you’d save the customer the equity they’d spend on hiring that VP level role? SNM: Yeah, I guess I never thought of it like that. I think in our early days we couldn’t take equity because we needed to feed our employees. In the later days we may have wanted to, but at that point we were part of Avnet and it was all more complicated. In retrospect the equity model probably would have made better sense.

SW: Why’s that? SNM: Well, equity creates a really good long term alignment. Consulting is good for solving problems quickly, but if the customer’s goal is to succeed over the long term then the best way to incentivize is through equity. What I’m doing now with Dragon Ventures is combining three approaches: Consulting, advising, and investing, and I think that the combination of these business models is really good. Consulting brings in cash, and advising and investing bring in equity. It creates really good alignment, and it cements that alignment into a long-term relationship.

SW: Switching gears: One of the things that Dragon Innovation worked a lot on was de-risking product launches. You did this first through Dragon Certified, then (in partnership with Avnet and Kickstarter) through Hardware Studio. How did these programs work - and more broadly, why do so many early stage companies fail to deliver crowdfunding projects on time? SNM: So many of these companies just don’t do the necessary planning work. Instead of doing the math - 6-8 weeks lead times for tooling, 6 weeks or whatever for FCC certification, all the per-part costs for their injection moldings - they baseline their crowdfunding goals and timelines based on what other companies have done. They set their goals from the top down, but in order to ship on time you’ve got to build your models from the bottom up. I do think that the level of play has gotten better over the years, but still the focus generally is on “how much money can I raise” rather than “how much money do I need to raise.”

SW: So your role was basically just to be the grownup in the room, forcing them to sit down and do all of that math? SNM: Yeah, I think many of these creators are very very smart people - but sometimes you need to be more than smart. Or maybe smart is just the wrong thing. Like, you could be less smart but if you’ve gotten the right experience by just going down the path more times then you’ll be in a better place. It’s really hard to anticipate what’s going to happen without that experience. So, our role was often like being a parent or a coach - just asking questions and pressing them and through that process they were able to gain more wisdom and have better outcomes.

SW: A lot has changed in the ~13 years since you founded Dragon Innovation. Many of the companies that seemed destined for success are now shut down, and the dynamics around hardware startups are completely different. How do you see the industry today? SNM: Well, in ‘09 it was sort of the heyday of hardware. We had Kickstarter, Pebble, Makerbot, Dropcam, Formlabs all just starting… and then sort of from 2017 until now it’s been on the down. But I think it’s coming back again - it feels like the level of quality and sophistication has increased dramatically. Gone are the days of just putting a board in a clamshell, and the products I’m working on now are as exciting to me as anything I’ve seen in my career. It’s an exciting time.

Planning & Strategy.

  • On the Wayback machine, an old blog post from Dragon Innovation on prepping for Chinese New Year.
  • Astra, then Alameda-based rocket startup, will go public via a SPAC.
  • A remarkably thorough explanation of why cat food cans are not designed with 1:1 height:diameter ratios, a strategy which would minimize surface area per unit volume:
  • Shapes that minimize surface area are also very difficult to heat evenly, whereas short and flat (or long and skinny) cans can be thermally processed quickly.
  • The tops and bottoms of cans tend to bulge under the pressure from thermal processing, and are therefore often made from thicker stock. Therefore, surface area minimization is not the same as material minimization.
  • Tooling needs to be changed out whenever you change can diameter or height, so can manufacturers have an incentive to lock one of those variables in and then play with the other one. And, the lid stamping tooling is harder to change out than the side cutting tooling is, so you tend to see a lot of cans designed with just one of a few diameters.
  • Scrap rates increase when can diameter increases. My sense is that this would be due to an increase in defective parts, as the larger diameter shouldn’t affect how the parts nest in a sheet.
  • Smaller diameter cans may be easier to package and warehouse, presumably because they’re more likely to divide nicely into an arbitrarily sized box.

    ...and in case you were hoping for it, here’s a fairly descriptive (and still totally flabberghasting) video of food cans being made.

Making & Manufacturing.

Maintenance, Repair & Operations.

Distribution & Logistics.

Inspection, Testing & Analysis.

  • A nice, explanatory image and academic paper on soil characterization. “Soil texture can be determined in the lab via particle‐size analysis using the hydrometer method; this method determines particle‐size distribution by measuring the time it takes for soil particles in water to settle out of suspension. The hydrometer method can be time consuming, requires certain equipment, and is not optimal for most people. However, anyone can determine textural class using the ‘Feel’ Method. This involves forming a moist soil sample into a ball and squeezing it between the thumb and index finger to make a ribbon. Texture is determined by ribbon length (if one can be formed at all) and grittiness or smoothness of the sample...With practice, the ‘Feel’ Method provides an immediate and accurate way to determine textural class, and even percent sand, silt, or clay, while in the field.”
  • A free 2D linkage simulator for Windows.


Cattle prod selfie stick.

Read the full story

The rest of this post is for SOW Subscribers (free or paid) only. Sign up now to read the full story and get access to all subscriber-only posts.

Sign up now
Already have an account? Sign in
Great! You’ve successfully signed up.
Welcome back! You've successfully signed in.
You've successfully subscribed to Scope of Work.
Your link has expired.
Success! Check your email for magic link to sign-in.
Success! Your billing info has been updated.
Your billing was not updated.