One of the key takeaways for me was that for open innovation to be effective in retail, cost of prototyping / small batch production needs to be very low. As such (and as alluded to in the post), access to 3D printing which enables a high level of customization and small batch production at reasonable costs seems like a requirement for open innovation to drive product development. I would be interested in looking at the minimum volume required to cost effectively transfer a new product from 3D printing for prototyping / small batch production to conventional manufacturing, which is still most cost effective to scale a product. In addition to SKU management, managing the manufacturing asset base through these transitions seems key to effectively leveraging open innovation to make an attractive financial return in addition to avoiding product stagnation / spur innovation.
Your questions about what will happen if standard ML applications become available to the masses for investing purposes is a crucial one. One of the issues that I grapple with ML in investing is to what extent will the penetration in ML/automated trading/passive investing erode alpha vs. enabling investors to capitalize market dislocations to drive alpha? Consider the following extreme scenario — if the overwhelming majority of the market is using the same ML platforms, then they will be guided to the same investment opportunities, thus resulting in more efficient price discovery and lack of differentiation in investment theses that reduces instances of alpha. In this scenario, having a differentiated (i.e., not machine/algorithmic) approach will be the only way to find alpha, as it will have been efficiently bid-away for all opportunities that ML has identified.
Howard Marks — the legendary investor — wrote a recent memo about the role of humans in investing going forward. I highly suggest reading it to help flesh out your open items, enjoy!
Interesting use cases for ML in logistics! In terms of your suggestion about using ML to better predict delays in customs clearance, I would be asking two follow-up questions to assess the viability of ML as an application before deciding whether or not to pursue it.
1) Given customs clearance involves a government agency as the counterparty (vs. private market actor such as suppliers when analyzing supply interruptions), will there be any confidentiality issues that would inhibit the gathering of a large data set, that includes data for more logistics providers than just DHL. If there isn’t reasonable confidence in the ability to a aggregate a sufficiently large data set, efforts should potentially be directed to use cases where data is more readily accessible.
2) ML relies on the notion that historical patterns can be used to build predictive models of future outcomes. Again, since customs clearance involves government agencies as the counterparties, I wonder if there will be less of a discernible relationship between observable driver variables and clearance time. Private market actors typically have motivations that are more easily unpacked and monitored, but I wonder if the same will be true with government agencies.
In response to your question about the level of customization that consumers will demand from car manufacturers in the future, it is important to also take into consideration the financial profile of car manufacturers. As an industry, car manufacturing has been plagued by very low return on investment (ROI), which in large part is driven by competition across manufacturers to constantly innovate and provide many options/models for consumers.
While innovation/options typically translates into higher prices, in the auto industry, consumers have come to take innovation for granted and manufacturers have found that they need to innovate out of necessity vs. to drive higher returns/margins. As such, there has been a lot of consolidation in the industry, both through M&A, and within organizations to simplify their product offerings to drive higher ROI. Given the trend toward increased standardization (e.g., small SUVs use same frames as larger sedans) to improve the returns profile, even if consumers demand customization, it is unclear whether the auto industry would meet that demand. As such, AM will may end up being more compelling in terms of simplifying the supply chain / inventory requirements for making spare parts on demand as opposed to enabling a structural increase in the level of customization consumers demand in the future.
In response to your question about whether 3D printing can succeed in the long-run without scale, there are two major additional cost item that greatly benefit from being spread across higher volumes of production, in addition to R&D and upfront capital investments that you highlight. The first is labor — specifically as it relates to the labor required for pre- and post-production processes, which can make up a large portion of total production costs . The second is the time and resources required to scan and upload the product/part design using CAD to create a printable image. Since this takes time for each unique product/part, but the model can be retrieved from 3D databases without any added cost for production for subsequent units once created, there are clear benefits to scale in developing the initial 3D model “set-up” costs to be spread across a larger volume of units . With these additional cost buckets taken into consideration, while it is still accurate to categorize AM as having lower requirements for minimum efficient scale of production, the benefits of scale should not be underestimated.
 J.B. Roca et al., Getting past the hype about 3-D printing. MIT Sloan Management Review 58, no. 3 (Spring 2017): 57–62.
 M. Holwef. The limits of 3D printing. Harvard Business Review Digital Articles (June 23, 2015).