How to Test Earnings Lead-Lag Signals Across a Supply Chain
September 11, 2026
Altsets
Research by Altsets Research
Use economically meaningful supplier-customer relationships and reporting calendars to test whether earlier supplier disclosures contain information before a customer reports without assuming causality.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- Micron and SK Hynix both reported operating results before Nvidia's August 26, 2026 fiscal Q2 2027 report, while Nvidia represents 17.62% of Micron revenue and 27.88% of SK Hynix revenue in the displayed relationship data.
- Reporting order creates a testable information sequence, but fiscal-period overlap, industry-wide pricing, other customers, and look-ahead bias must be controlled before describing the pattern as a lead-lag signal.
Potentially. Supplier earnings can contain useful information before a major customer reports when the relationship is economically material and the reported product or demand signal is relevant to that customer. The relationship identifies which lead-lag hypothesis to test; it does not prove causality or prediction by itself.
Nvidia is associated with 17.62% of Micron revenue and 27.88% of SK Hynix revenue. Both memory suppliers reported before Nvidia's fiscal second-quarter 2027 earnings call. That creates a testable lead-lag research setup.
The 2026 reporting sequence is observable
Micron reported fiscal Q3 2026 results on June 24, 2026 for a quarter that ended May 28. SK Hynix reported Q2 2026 results on July 29, 2026. Nvidia scheduled its fiscal Q2 2027 results for August 26, 2026, covering a quarter that ended July 26.
So two economically exposed suppliers disclosed operating information before the customer reported. That is useful for event research. It is not the same as saying the reporting periods line up perfectly.
The periods overlap imperfectly
Micron's fiscal calendar does not match Nvidia's. SK Hynix reports on a calendar-quarter basis. Nvidia's quarter ended later than either supplier's reported period. A trader who ignores those differences can create a false lead-lag relationship. The correct question is not: Did Micron report first?
The better question is: How much of the supplier's disclosed period overlaps the customer's operating period, and what information could plausibly carry forward?
Relationship data chooses the candidates
Without a supply-chain map, an earnings-calendar study can become a search across thousands of companies. The relationship data narrows the universe. Micron and SK Hynix are not random semiconductor names.
Nvidia is economically important to both suppliers in the displayed data. That makes their earnings more plausible research inputs for Nvidia than an unrelated memory company with no mapped customer relationship. The relationship does not guarantee predictive content. It justifies the test.
What can lead the customer report?
Several supplier disclosures can be relevant before a customer reports: shipment growth, pricing, product mix, capacity utilization, inventory, long-term agreements, customer concentration, management commentary, and new product ramps. The investor should ask whether the signal is customer-specific or industry-wide. That distinction is critical in memory markets.
Micron's June report contained several possible signals
Micron reported record fiscal Q3 2026 results and said HBM4 was in high-volume shipments for its lead customer's platform. It also described strong data-center demand and multi-year strategic customer agreements. Those facts may be relevant to the AI-memory environment surrounding Nvidia.
The public materials did not identify every lead customer or allocate the reported growth to Nvidia. Altsets supplies the named relationship. The filing and earnings materials supply the operating context. Those evidence layers should remain separate.
SK Hynix added a second pre-Nvidia observation
SK Hynix reported record Q2 2026 results on July 29. The company said AI infrastructure investment was driving strong demand for high-value memory products and that HBM4 mass shipments had begun. SK Hynix also has a publicly announced multi-year Nvidia memory partnership.
This creates a stronger named connection than generic industry commentary. It still does not mean SK Hynix earnings can be mechanically translated into Nvidia revenue.
Two suppliers can strengthen a signal without proving causality
If two economically exposed suppliers independently report similar demand conditions before the same customer reports, the observation becomes more interesting. But several explanations remain possible: broad memory pricing, industry-wide inventory rebuilding, demand from customers other than Nvidia, product-mix changes, capacity constraints, and different reporting periods. A multi-supplier signal is stronger than one anecdote. It is not proof of customer-specific causation.
A real lead-lag test needs repeated observations
One 2026 sequence is an example, not a statistical result. A serious test would repeat the process across many reporting periods using only information available at each historical date. For each observation, record:
- supplier report date;
- customer report date;
- supplier reporting period;
- relationship state at that date;
- supplier exposure metric;
- operating signal being tested;
- customer outcome measured later.
That turns a narrative into a research design.
Point-in-time data is mandatory
If a supplier relationship is discovered today and then projected backward into earlier earnings seasons, the test can contain look-ahead bias. The relationship needs to be known at the historical observation date. The point-in-time backtesting guide covers that requirement. This matters even more if the supplier-customer relationship changes over time.
Do not optimize the signal on the same history used to discover it
A researcher can easily search historical data until one supplier appears to "predict" a customer. That can create overfitting. A stronger design defines the signal first, tests it on one sample, and validates it on another.
Supply-chain data should narrow economically plausible pairs. It should not replace statistical discipline.
The interface depends on the stage of the research
A visual map can discover which suppliers report before a customer. An LLM connected through MCP can help retrieve and compare a handful of earnings disclosures. An API is better when the research becomes a repeated historical event study across hundreds of relationships. The map versus MCP versus API guide explains that division of labor.
A repeatable earnings lead-lag workflow
- Start with an economically meaningful relationship.
- Build the reporting calendar for both companies.
- Preserve each company's fiscal period.
- Identify supplier disclosures available before the customer reports.
- Separate customer-specific evidence from industry-wide evidence.
- Use point-in-time relationship data.
- Define the customer outcome before testing.
- Repeat across many periods.
- Validate out of sample.
- Treat the result as a statistical signal only if the evidence supports it.
The earnings-propagation guide focuses on how surprises can travel through a network. This article solves a different problem: whether reporting order can be tested as an information lead without assuming the relationship itself creates predictability. For relationship definitions and evidence limits, read the Altsets methodology. Browse Supply-Chain Data Use Cases for other event-driven research methods.