Do You Actually Need 20 Years of Supply-Chain History?
July 23, 2026
Altsets
Research by Altsets Research
Current relationship data is enough for many portfolio decisions, while historical snapshots become necessary when the question depends on durability, change, backtesting, or what was actually knowable at the time.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- Current data can answer today's overlap, diversification, and catalyst questions without requiring a long historical reconstruction.
- History becomes necessary for durability analysis, valid backtests, relationship-change studies, and avoiding hindsight when the investor needs to know what the network looked like on an earlier date.
Twenty years of supply-chain history sounds valuable, but most investors do not need twenty years of data for every decision. The current network can answer many of the questions that matter today: what a company depends on, where a portfolio overlaps, which external companies deserve monitoring, and what known catalyst could reach several holdings.
History becomes valuable when the question itself is historical. The right amount of history depends on what you are trying to prove.
Current data is enough for many first decisions
If the investor wants to know whether today's portfolio shares a major customer, the current network is the relevant starting point. The same is true for checking whether a candidate stock repeats an existing supplier dependency or whether an upcoming earnings report matters to several holdings.
Those decisions do not become automatically better because twenty years of snapshots are available. Current relationship structure can already reveal the concentration that exists now.
This matters for cost and effort because an investor does not need to turn every research task into a historical study.
History matters when durability is part of the thesis
A current relationship tells the investor that two companies are connected now. It does not reveal whether that relationship appeared last quarter, persisted for a decade, disappeared and returned, or changed dramatically in economic importance.
Those are different investment questions. A long-lived customer relationship can support a different interpretation from a newly formed relationship, while a supplier dependency that has steadily weakened may deserve less weight than the current snapshot suggests.
Historical data is useful when the investor needs to distinguish a durable economic structure from a temporary configuration.
Backtesting requires the network that existed at the time
Historical supply-chain data becomes essential when the investor wants to test a strategy. A backtest that uses today's relationships to explain stock behavior five years ago leaks future knowledge into the past.
That problem is easy to miss because relationship data feels structural. In reality, customers, suppliers, securities, acquisitions, product cycles, and economic importance all change over time.
Point-in-time snapshots let the researcher ask what the strategy could actually have known on each historical date. That is a much stricter requirement than simply having a long archive.
Public filings show why history can be fragmented
TSMC's investor site provides annual reports going back across many years, while its 2025 annual report describes a business that produced 12,682 products using 305 process technologies for 534 customers. Public documents can contain enormous amounts of useful historical context, but reconstructing a normalized relationship network across years from those documents is a different task from reading one current filing.
This is where historical relationship data saves research time. The value is not only the number of years. It is the ability to compare the same kind of relationship state across dates without manually rebuilding the network from changing disclosures.
History is also useful for checking whether a story is actually new
Investors frequently react to announcements as though a relationship appeared overnight. Historical data can show whether the commercial connection is new, whether the companies have been linked for years, or whether only the product context changed.
That distinction can matter for valuation and catalysts. A genuinely new customer can change the thesis differently from a new press release describing an old relationship.
Historical context can therefore prevent investors from paying for novelty that is mostly narrative.
More history does not automatically mean more signal
A twenty-year series can still be useless if the metric is poorly defined, the company changed dramatically, or the investor is testing a question that only became economically relevant recently. Older data should not be included merely because it exists.
The research window should match the economic question. A semiconductor relationship spanning several architecture cycles may need a different historical window from a short-lived customer contract or a recent AI product transition.
This keeps history from becoming another form of false precision.
A practical investor can use history selectively
The easiest workflow is to begin with the current network. If the current relationship changes the decision, then ask whether durability, trend, or historical timing matters enough to justify looking backward.
For a portfolio-overlap check, the answer may be no. For a backtest, relationship-change study, or thesis about customer durability, the answer may be absolutely yes.
That is a much lower-effort way to use historical data than assuming every stock requires a twenty-year reconstruction.
The conclusion is that history is optional until the question needs it
Long history is powerful because it unlocks questions current data cannot answer. It is not a prerequisite for benefiting from supply-chain data.
An investor can use the current network for today's portfolio decisions and reach for historical snapshots when the thesis depends on durability, change, or a valid point-in-time test. The expensive mistake is not using too little history. It is using the wrong time horizon for the question.
The point-in-time backtesting guide explains why current relationships cannot simply be projected backward. The thesis-change-log guide shows how historical state can preserve what was actually known when an investment decision was made.
For relationship definitions and evidence limits, read the Altsets methodology.