Direct Sourcing

Shoe Dog is not a book about running shoes. It is a book about a structural arbitrage — the gap between what a factory in Kobe charges for a pair of shoes and what an American runner will pay for them. Phil Knight’s insight in 1962 was not a product insight; it was a distribution insight. Japanese manufacturing capacity was producing goods at a cost structure the American market hadn’t priced in. He was the mechanism that moved the product across that gap.

Between the factory and the customer were two intermediary layers, each extracting rent: the trading company that managed the factory relationship, and the importer who handled customs, warehousing, and distribution. Knight needed both. Neither was optional. The capital requirements for the first order, the relationships required to manage the trading company, the warehouse needed to receive the goods — none of this was accessible to a single person acting alone. The minimum viable team for the structural play was a small organization with institutional support.

That constraint has changed.

The jobs to be done are identical to 1962: find excess manufacturing capacity, secure a price the target market hasn’t seen, move the product to the buyer. What has changed is the infrastructure underneath each of those jobs.

Third-party logistics providers in Shenzhen now integrate directly into Shopify, handling inbound scanning, pick-and-pack, and global fulfilment with local tracking numbers — the customer never sees Chinese origin. Sample production is three days for most garment categories; ten days for custom embroidery. Canton Fair access requires no import history and no minimum prior orders. Private-label packaging is available from 3,000-unit runs at a few cents per unit. The trading company — Knight’s indispensable rent-extractor — is no longer a requirement. The importer is optional. The warehouse is optional.

What remains unchanged: the factory still wants to know it is dealing with a buyer who can move volume over time. Trust is still the bottleneck at the factory relationship. Capital is still required to cover minimum order quantities. The product still has to find its market. These are not trivial constraints. But they are a materially smaller version of the same constraints Knight faced — and they are now, in principle, solvable by a single operator with adequate preparation and modest capital.

The experiment is a test of a specific, falsifiable claim: a single operator, without institutional capital or an existing import business, can run a margin-positive factory-direct DTC physical goods operation at meaningful scale — starting from zero.

“Meaningful scale” means a business that generates income materially above what the operator’s time would earn otherwise. “Margin-positive” means after all costs — sampling, MOQ capital, 3PL fees, returns, customer acquisition, platform fees — the per-unit economics justify the enterprise. “Single operator” means no warehouse team, no trading company, no partners whose sole function is to manage intermediary relationships.

The experiment holds those three constraints fixed and varies the category, the factory relationship strategy, and the demand architecture. The category selection matters: physical goods businesses where the differentiation is attribute-expressible — where the difference between this product and a commodity substitute can be stated in a sentence — are structurally better candidates than categories where differentiation is purely brand-emotional.

Four outcomes would falsify the hypothesis, in decreasing severity:

The most decisive falsifier: the factory relationship cannot be established without a trading company or agent, because factories in the relevant category will not deal with buyers below a trust threshold that requires an intermediary to certify. If the rent extraction is structural rather than legacy, the experiment ends there.

The second falsifier: per-unit economics are viable in the spreadsheet but not in practice, because the cost of customer acquisition at realistic volume eliminates the factory-to-retail margin that makes the play interesting. The arbitrage exists in manufacturing; it does not survive contact with paid media at scale.

The third falsifier: the operation is viable at small scale — a hundred units, a few customers — but does not compound. The limiting factor is attention, not capital: managing factory relationships, logistics exceptions, returns, and customer acquisition simultaneously exceeds the cognitive budget of a single operator before the business reaches a self-sustaining size.

The fourth, weakest falsifier: the operation works, but only in narrow conditions — specific categories, specific factory relationships, specific markets — that do not generalize. The experiment succeeds as a personal income source but fails as a generalizable blueprint.

For the hypothesis to hold, three things have to be simultaneously true:

First, there must be a category where factory-direct margin is real and the category’s differentiation is attribute-expressible. Not every category has this. Commodity goods have no margin; highly differentiated goods require brand-building that exceeds solo-operator capacity. The viable categories sit in between: goods with genuine material or process differentiation that can be described precisely enough that the right buyer finds them without a brand reputation doing the work.

Second, the factory relationship must be negotiable on trust-building terms rather than volume-first terms. Some factories will not deal below a minimum annual commitment that requires institutional capital. Others will deal on milestone pricing — pay more per unit initially, earn lower pricing as volume milestones are hit. The latter is a viable structure for a solo operator; the former is not.

Third, the demand architecture must be buildable without continuous paid media spend. A DTC physical goods business whose customer acquisition is primarily paid media is structurally vulnerable to the same closed legibility problem that SPF diagnoses in marketplaces — the platform controls the channel, the operator pays rent for access. A viable solo-operator model requires a demand architecture built on something the operator owns: a product that generates word-of-mouth, a distribution channel that doesn’t require continuous spend, or a category where organic search or AI assistant referral is a real acquisition path.

This experiment is in design, not in market. The observations above derive from documented cases — in particular the approaches visible in Jack Bloomfield’s factory-floor sourcing documentation and Justin Lavoie-Hébert’s Canton Fair sourcing practice — not from a live test. The experiment is being set up as a future run, with the design informed by watching operators who are already running versions of it.

What the documented cases show: the infrastructure hypothesis is correct. The 3PL integrations work. Canton Fair access is real for small buyers. Factory relationships are negotiable on milestone terms. The cases also show where the documented operators struggle: customer acquisition is the bottleneck in every case, not sourcing. Every operator who figured out the factory side still faced a demand architecture that required either continuous paid media or a prior audience. This is the constraint the experiment will be designed to address directly.