The quest for non-traditional data to understand and exploit opportunity has been around since the advent of markets. While the realization that alt data can be used to supplement existing analyses has caught fire the past few years, it has not spilled over to markets more broadly. Because of the unusual skill set of the proto-quantalist, it’s more than possible valuable domain experience gets lost in the mix, and this can lead to poor outcomes. Your author suspects post-hoc analyses aren’t common practice, but they could go a long way to improving process, something that is desperately needed in a nascent field.

In many cases, leveraging a new data source and/or approach can be an exercise in getting comfortable with new known unknowns. Is there sufficient historical source data for backtesting? Can ground truth be obtained to validate models? Are there other (similar) data sources that evidence like patterns? What about competitors already in-market? Or a startup trying looking to sell signal to anybody?

An example of WoolCo can help identify where alpha may lurk. The company is a manufacturer of winter jackets. Its primary input is sheep, which are transported to the shearing plant by road and rail. WoolCo also owns a own sheep farm to minimize volatility in the sheep spot market.

WoolCo employs workers, sources buttons, scissors, electricity and host of other inputs to make the jackets. [NB: your author is well aware of the vast simplification presented.] Once WoolCo’s jackets come off the factory line, they are packaged and shipped to distributors, retailers and customers. This likely means (in 2018) a complicated and interdependent supply chain that uses road, air, rail, freight brokers, e-commerce,
fax, EDI, 3PL, energy transmission and other tools/services to produce the end product.

Using a variety of techniques (eg, telematics, imagery, cargo/freight manifest, online panel, asset tracking, energy utilization) to monitor the physical and digital supply chain, the astute analyst will seek to interpret patterns of WoolCo’s inputs. The US military has a rich (and mostly classified!) history of excelling at this, largely because the stakes were so high. During the Cold War, the National Reconnaissance Office analyzed high altitude flyovers/satellite imagery and other techniques to determine Soviet troop levels based on factory production of tanks, jackets and who knows what else. In WWII, mathletes helped estimate German tank production.

The WoolCo analyst will hopefully hone in on one input that can be measured and correlated with performance, yielding far greater accuracy and less latency than the next best option. In theory, this will lead to an understanding of financial performance–enter the quest for alpha!

An Overview of WoolCo Inputs

However, just because a fund employs novel approaches, hires smart people with unique relationships and uses oversized compute does not necessarily mean the analyst is seeing what she thinks. Tech and bias can often blind one to the truth.

What’s an Obfuscated Supply Chain?

Historically, supply chains have been messy, opaque and expensive. Because of globalization, sensors, increased connectivity and a variety of other factors, they have become increasingly visible. This is great for investors and competitors, but not so exciting for WoolCo.

So what do can you do to prevent competitors et al from snooping? Erecting Potemkin villages is a reasonable idea and it’s a timeless strategy, through techniques have evolved. The Soviet doctrine of maskirovka is alive and well, with full-scale inflatable MIGs and blow-up missile systems. In WWII, Allied forces had their own tactical deception unit, aka the Ghost Army.

In the world of alpha hunting, estimating oil supply is nothing new, with experts measuring hull displacement, storage tank shadows and other techniques driven by data data data and machine learning. Without having on the ground (or some other non-correlated source) to verify, what level of confidence do you have in what is being measured? It’s not far fetched to think that oil producers, and more broadly, any party in a value chain exposing emergent signal may decide to plant false flags.Given the cost of somebody knowing, filling up oil tankers with seawater can be a pretty good idea.

While likely not a widespread ‘problem’ (or opportunity, depending where you sit) today, supply chain obfuscation will become one of the more pernicious and thorny problems when looking for source that can provide outsized returns. The astute analyst will develop techniques to understand non-statistical confidence based on non-correlated source source data.

[Ed: If you have any examples of supply chain obfuscation, I would love to learn more.]