Can Supply-Chain Network Motifs Become Quant Features?
September 14, 2026
Altsets
Research by Altsets Research
Yes. Repeating local graph shapes such as shared-customer wedges, shared-supplier wedges, fan-in hubs, and fan-out customer structures can encode economic patterns that degree and centrality scores collapse into one number.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- The supplied Shin-Etsu customer fan-out and Nvidia shared-customer structures demonstrate that equal edge counts can encode very different economic shapes and therefore different quant hypotheses.
- Network-motif research uses recurring subgraphs to represent higher-order structure, providing a basis for testing supply-chain motif formation, persistence, and dissolution as features beyond node-level centrality.
Yes. Repeating small network structures can become quant features when each pattern has a clear economic interpretation. A one-customer-many-suppliers pattern, a one-supplier-many-customers pattern, and a two-hop dependency chain describe different competitive and risk structures even when the individual companies have similar degree or centrality scores.
The same number of edges can form very different economic shapes
The supplied Shin-Etsu Chemical view forms a simple fan-out pattern. Shin-Etsu maps to Samsung Electronics, TSMC, and Intel, with displayed supplier revenue shares of 2.43%, 4.02%, and 1.79% respectively. One supplier is connected to several major semiconductor customers. That motif can represent customer diversification while still leaving the supplier exposed to a common industry cycle.
The Nvidia-centered data produces the opposite type of pattern. Micron and SK Hynix both map into Nvidia as a major customer, creating a many-suppliers-to-one-customer structure. That motif can represent common demand exposure and potential competition for the same customer's spending. A simple degree count can say that several edges exist. The motif tells the quant how those edges are arranged.
Motifs can encode hypotheses that centrality misses
Centrality compresses a company's global network position into one score. Motifs preserve local structure. A supplier with three unrelated customers and a supplier with three customers that all depend on the same downstream platform can have the same degree but very different economic concentration. Likewise, two companies can each have five suppliers while one sits behind a single shared bottleneck and the other has genuinely independent upstream paths.
Network science uses motifs precisely because recurring small subgraphs can capture structure that node-level statistics miss. Recent work also uses evolving motifs for link prediction in supply-chain-like trade networks, while financial network research has used triadic motifs to distinguish market states. A stock model can test whether particular supply-chain motifs contain incremental information about risk, revisions, return dispersion, or relationship formation.
Direction and edge type define the motif
A quant should not treat every triangle or three-node pattern as identical. Supplier-to-customer direction matters, and a direct relationship is different from two companies merely sharing the same node. One useful motif library can include shared-customer wedges, shared-supplier wedges, supplier-customer chains, reciprocal commercial structures where supported, and fan-in or fan-out hubs.
Quantified metrics can add a second layer. A shared-customer motif where the customer represents 20% of each supplier's revenue is economically different from one where both relationships are tiny. Structural-only motifs can remain binary, while quantified versions can be weighted only when the directional metrics are comparable. The feature family should not fabricate weights simply to make every motif numeric.
Motif changes can become event features
The network pattern can change over time. A supplier with three dispersed customers can become increasingly dominated by one. Two portfolio companies can begin sharing a new customer. A previously independent supplier chain can converge on one upstream bottleneck. Those are motif transitions rather than simple edge counts.
A point-in-time dataset allows the quant to test whether motif formation or dissolution precedes changes in fundamentals, volatility, correlation, or returns. That is different from relationship churn at the individual edge level because the feature is the change in a multi-company structure. The researcher can ask whether the emergence of a shared-customer motif predicts greater conditional covariance even if none of the underlying relationships is individually new.
The conclusion is that local graph shape can be a feature family of its own
Yes, supply-chain motifs can become quantitative features. The advantage is not mathematical novelty by itself. The advantage is that a motif can encode an economically interpretable pattern such as common demand, common bottleneck exposure, or diversified customer reach that a single centrality or degree statistic cannot distinguish. The strongest motifs are the ones whose direction and economic mechanism are defined before returns are examined.
The network-centrality guide explains what global graph-position measures can and cannot capture. The network peer-group guide explains how shared customers and suppliers can define economic similarity across firms.
For relationship definitions and evidence limits, read the Altsets methodology.