How to Use Supply-Chain Data When Relationship Metrics Are Missing

Key findings

• Structural edges are useful for discovery and network topology but do not support economic ranking until directional metrics are available.
• In the supplied Nvidia graph, Quanta and Samsung are quantified while Amazon, Microsoft, and Super Micro Computer remain structural-only, illustrating why missing must not be treated as zero.

A supply-chain map can be useful even when some relationships do not have visible economic metrics. The mistake is treating a structural edge and a quantified edge as if they support the same conclusion. A supplied Nvidia network illustrates the difference clearly. The graph shows downstream relationships from Nvidia to Amazon, Microsoft, Super Micro Computer, Samsung Electronics, and Quanta Computer. Two edges carry visible percentage metrics:

• Nvidia to Quanta Computer: 0.53% of Nvidia revenue and 1.76% of Quanta's cost base
• Nvidia to Samsung Electronics: 0.04% of Nvidia revenue and 0.10% of Samsung's cost base

The Amazon, Microsoft, and Super Micro Computer relationships are visible structurally, but the supplied graph does not show comparable percentages for those edges. The right workflow is to preserve both kinds of information without pretending they are equivalent.

Structural data answers "who is connected?"

A mapped relationship can establish that two companies appear in the same supplier-customer network. That is already useful. It can support company discovery, customer or supplier lists, event watchlists, second-order network exploration, portfolio overlap analysis, and research prioritization.

But structure alone does not tell you the economic magnitude of the relationship. A connection to a famous customer may be small. A connection to an obscure customer may be economically important. The edge is a starting point, not the final conclusion.

Quantified data answers "how much does the edge appear to matter?"

Percentage metrics add economic context. Supplier revenue percentage asks how much of the supplier's revenue is associated with the customer. Customer cost percentage asks how much of the customer's cost base is associated with the supplier.

In the Nvidia example, the Quanta relationship is visibly larger than the Samsung relationship on both displayed percentage measures. That comparison is possible because both edges are quantified. The same comparison cannot be made against Amazon or Microsoft from this screenshot alone.

Missing is not zero

This rule matters enough to state directly: A missing metric is not a zero metric.

If Amazon's edge has no visible supplier-revenue percentage, it would be wrong to rank Amazon below Samsung's 0.04% by assigning Amazon a value of zero. The value is unknown in the supplied view. A research system should preserve that state. Otherwise, missing data quietly becomes a false economic conclusion.

Do not rank mixed-quality edges without labeling them

A practical table should separate quantified and structural-only relationships. For example:

This is much more honest than sorting all five relationships by a made-up numeric field. The same principle applies to portfolio screens and network scans.

Structural-only edges can still be high-priority

An unquantified relationship is not necessarily unimportant. It may deserve immediate investigation because the counterparty just reported earnings, a disruption occurred, a regulatory event affected the company, the relationship connects two portfolio holdings, it creates an interesting second-order path, and it appears in a strategically important end market. The edge can be important as a research lead even when its economic size remains unresolved.

Quantification should follow the question

Not every research task requires every metric. If the question is "Does Nvidia have a relationship with Microsoft in this network?" the structural edge may be sufficient. If the question is "How dependent is Nvidia on Microsoft?" a structural edge is insufficient.

If the question is "Which customer event is most economically relevant to Nvidia?" supplier-revenue percentages become important. The data requirement should follow the question.

This prevents false precision in AI research

This distinction is especially important when an LLM or automated research system summarizes a network. Without explicit missing-value handling, a model may describe unquantified edges as small, rank unknown relationships below quantified ones, fill gaps with generic internet knowledge, and collapse topology and economic magnitude into one confidence score. A better system labels what is known, unknown, and inferred. That produces less fluent but more defensible research.

A repeatable missing-metric workflow

1. Map the relevant relationships.
2. Mark which edges are structural-only.
3. Mark which edges have supplier-revenue or customer-cost metrics.
4. Never convert missing metrics to zero.
5. Use structural edges for discovery and topology.
6. Use quantified edges for economic ranking.
7. Retrieve more data only when the research question requires it.
8. State clearly when a conclusion is based on structure rather than magnitude.

The Nvidia customer-concentration analysis shows how this distinction changes a real company conclusion. The relationship-size versus relative-exposure guide explains how to interpret economic metrics once they are available. For definitions and coverage limits, read the Altsets supply-chain data methodology. Browse Supply-Chain Data Use Cases for other research workflows.

Methodology

Read the methodology for this research.

Related research

• Are Nvidia's Customer Relationships Concentrated? Quanta and Samsung Compared

• When a 10-K Says Customer A: How Supply-Chain Data Helps

• Why Relationship Size Alone Can Mislead Supply-Chain Research

• Does Micron's Customer Network Extend Beyond Nvidia Into AI and Server Hardware?