--- title: "Going From Supply Chains to Trading Signals with a Knowledge Graph You Can Own" description: "Use NOSIBLE Atlas for supply-chain research, customer momentum and historical backtests. Work with a point-in-time company graph in your infrastructure." url: "https://nosible.com/blog/from-supply-chains-to-trading-signals-a-knowledge-graph-you-can-own" --- [Trading Signals](https://nosible.com/blog/tag/trading-signals) [Information Retrieval](https://nosible.com/blog/tag/information-retrieval) [Web Search](https://nosible.com/blog/tag/web-search) # Going From Supply Chains to Trading Signals with a Knowledge Graph You Can Own [Stuart Reid](https://www.linkedin.com/in/stuartgordonreid/) 2026-10-19 7 min read Copy as Markdown ![Atlas Graph Harness workflow: entity and relationship grounding, graph investigation, execution checks, reviewer feedback and delivery of accepted results.](https://nosible.com/images/2026/10/hero-graph-harness.png) [NOSIBLE Atlas](https://nosible.com/blog/announcing-atlas-the-point-in-time-knowledge-graph) is a point-in-time knowledge graph for investors. It connects suppliers and customers with owners, products and places. Atlas does a lot more than that, but this post focuses on supply-chain use cases for investors. For a broader introduction, check out our [launch post](https://nosible.com/blog/announcing-atlas-the-point-in-time-knowledge-graph). We deliver the entire graph into your infrastructure, where your team can add private entities and infer proprietary facts from your proprietary data. Each fact includes its supporting sources and dates, the number of times it's been seen, the extraction model's confidence score and the reranker model's confidence score. That gives your team the data to test supply-chain signals. One example is [customer momentum](https://www.aqr.com/Insights/Research/Journal-Article/Economic-Links-and-Predictable-Returns), where customer returns help predict subsequent supplier returns. Researchers have examined signals spanning [several customer tiers](https://insight.factset.com/propagating-momentum-information-through-global-supply-chain-networks) and how predictability changes around [supplier earnings announcements](https://onlinelibrary.wiley.com/doi/abs/10.1111/1475-679X.12158). ## Why Sovereignty Matters Your analysts spend years researching the companies you invest in. Their supplier checks, customer revenue estimates and investment theses should become data your firm can keep using. Atlas lets you connect that work to public evidence, so the research you're paying for can feed your own models and strategies. You can deploy the ENTIRE Atlas graph on-prem and add your proprietary data inside your own infrastructure. Your firm controls access, storage and backups. That matters when the graph contains confidential research, portfolio exposures and the relationships you're using to develop a trading signal. We built Atlas on an open-source stack. We fine-tuned [GLiNER](https://huggingface.co/gliner-community/gliner_large-v2.5) for named entity recognition and [GLiNER-Relex](https://huggingface.co/knowledgator/gliner-relex-large-v0.5) for relationship extraction, and built our own reranker to check the evidence. We can run extraction ourselves and adapt the models to the entities and relationships investors care about. That gives us control over what Atlas learns to extract as your research needs change. Your private Atlas can keep growing as your team does research. Add a supplier your analysts have found, connect it to your holdings and check which positions share that exposure. Add customer revenue estimates and use them to weight a signal. You decide what goes into the graph and how your team uses it. ## The LLM <> Graph Harness The Atlas Graph Harness lets you ask questions in natural language. An LLM plans an investigation over the graph's entity types, relationship types and dated evidence. It chooses graph traversals and SQL operations, inspects the results and revises its approach as evidence comes back. Supplier, customer, ownership and product relationships can be combined to connect operating businesses to listed companies and examine shared exposures across holdings. You can extract a fragment around a company or portfolio, choose relationship types and filter links by observation count, extraction confidence and reranker confidence. Use those fragments in your own models and extend them with private research. ![Atlas Graph Harness architecture: entity and relationship grounding, iterative graph investigation, execution checks, reviewer feedback and delivery of accepted results.](https://nosible.com/images/2026/10/hero-graph-harness.png) 1. **Ground the question.** Entity search resolves names and aliases. Embeddings retrieve candidate relationship types, a cross-encoder reranks them, and Jev or an LLM helps resolve uncertain selections. Atlas retains selection provenance. 2. **Run the investigation.** Rust traverses a memory-mapped compressed sparse row (CSR) graph. Read-only SQLite joins, intersects, excludes and aggregates materialized results. Native tools handle paths, rankings and analyses. 3. **Check and refine.** Execution checks enforce supported operations, valid identifiers, scope, resource limits and provenance. A separate reviewer call assesses whether the executed results, explanation and graph answer the original question. The planner can revise its investigation using feedback. 4. **Present the evidence.** Answer tables stay separate from size-limited drawings; visualization limits cannot change results. Drawing recorded supporting paths tied to final answer rows preserves intermediaries. Completeness describes indexed data and declared scope; real-world relationships can be missing. 5. **Stream and save the session.** Next.js invokes Rust through N-API workers, streaming updates. Reviewed graphs and explanations can precede saving. PostgreSQL atomically commits accepted sessions and terminal responses. ## Three questions for Atlas I ran three prompts through the harness, combining supply-chain and other Atlas relationships. Each used data from January 1, 2010 through September 26, 2026. Counts describe indexed records, not exhaustive company or product lists. ### Example 1: Products and Brands > **Show me every product or brand sold by The Coca-Cola Company or any of their subsidiaries worldwide** Atlas returned **156 distinct product records across 190 seller–product links**, including Fanta, Sprite, Coca-Cola Zero Sugar, Diet Coke, Thums Up and Fuze Tea. Following recorded subsidiary links lets a researcher collect product names across the group and monitor them for demand, pricing or product recalls. ![Coca-Cola product and subsidiary graph with relationship filters and entity types.](https://nosible.com/images/2026/10/atlas-supply-chain-coca-cola-overview.png) ![Supporting source evidence and confidence scores for the Coca-Cola HBC Italia relationship with Coca-Cola.](https://nosible.com/images/2026/10/atlas-supply-chain-coca-cola-evidence.png) ![Coca-Cola India entity details, evidence dates, aliases and connections.](https://nosible.com/images/2026/10/atlas-supply-chain-coca-cola-entity.png) ### Example 2: Supplier Geography > **Show me every organization located in Taiwan that supplies to either Nvidia, AMD, Intel, Qualcomm, or Apple** Atlas returned **50 organization records** with both a location link to Taiwan and a direct supplier link to at least one of the five customers. Examples include MediaTek, TSMC, Quanta Computer, Foxconn, Pegatron and Delta Electronics. Some companies have multiple name records, so these are not 50 unique businesses. A fund can use the fragment to investigate supplier exposure shared by its technology holdings. ![Taiwan supplier graph connecting organization records with Nvidia, AMD, Intel, Qualcomm and Apple.](https://nosible.com/images/2026/10/atlas-supply-chain-taiwan-overview.png) ![TSMC entity details, evidence dates, aliases and customer connections in the Taiwan supplier graph.](https://nosible.com/images/2026/10/atlas-supply-chain-taiwan-tsmc.png) ![Supporting source evidence and confidence scores for the recorded UMC supplier relationship with Qualcomm.](https://nosible.com/images/2026/10/atlas-supply-chain-taiwan-evidence.png) ### Example 3: Investors and Customers > **Show me every company backed by Founders Fund or Peter Thiel that is also a supplier to the US Army, Navy, or Airforce** Atlas returned **eight matching company records across 23 backer–company–military combinations**, including Anduril, Impulse Space, Palantir and SpaceX. The investigation combined recorded `invests_in` and `supplier_to` relationships. A researcher can use the result to screen for military customer exposure and inspect the investment and supplier evidence behind each match. ![Recorded investment and military supplier relationships connecting Founders Fund and Peter Thiel with company records.](https://nosible.com/images/2026/10/atlas-supply-chain-founders-fund-overview.png) ![SpaceX entity details, evidence dates, aliases and recorded connections to Founders Fund and U.S. military branches.](https://nosible.com/images/2026/10/atlas-supply-chain-founders-fund-spacex.png) ![Supporting source evidence and confidence scores for the recorded Maximus supplier relationship with the U.S. Air Force.](https://nosible.com/images/2026/10/atlas-supply-chain-founders-fund-evidence.png) Those relationships can then be combined with returns and earnings dates to test trading signals. ## Your Next Quant Project ### Customer Momentum Signals: Predicting Supplier Returns from Customer Returns Using Supply-Chain Graphs Cohen and Frazzini's [2008 Journal of Finance study](https://www.aqr.com/Insights/Research/Journal-Article/Economic-Links-and-Predictable-Returns) found that customer returns predicted subsequent supplier returns. A fund can test this by ranking suppliers using their customers' recent returns, then measuring subsequent supplier returns over several months. When revenue shares are disclosed, weight each customer's return by the share of the supplier's revenue it accounts for. ### Multi-Tier Momentum Signals: Testing Customer Returns Across Several Tiers of the Supply-Chain Graph Yamamoto, Kawadai and Miyahara's [2021 Journal of Portfolio Management study](https://insight.factset.com/propagating-momentum-information-through-global-supply-chain-networks) tested return signals from direct customers and customers further down the supply chain. They compared one-month and 12-month return windows. Signals using up to four tiers produced the highest annualized returns in their tests. A fund can compare the results from direct customer links with those from several tiers. ### Timing Customer Momentum Signals with Supply-Chain Graphs and Supplier Earnings Announcement Dates Madsen's [2017 Journal of Accounting Research paper](https://onlinelibrary.wiley.com/doi/abs/10.1111/1475-679X.12158) examined how supplier earnings dates affect customer momentum. Attention to customers increased before suppliers reported, and customer returns predicted supplier returns before the announcement, but not afterwards in his tests. A fund can compare holding periods ending before, at and after supplier earnings. ## The Only Backtest Friendly Graph Atlas lets researchers select relationships supported by evidence available at a chosen date. Use that view when constructing signals in historical backtests to avoid including information discovered later. A relationship's first evidence date may differ from the date the companies started doing business. Atlas delivers a source-backed, point-in-time company graph into your infrastructure. Your team can filter its relationships by evidence date, extract the fragments needed for a strategy and add proprietary research to its private copy. [Talk to us about using Atlas with your investment universe](https://nosible.com/start-trial). [All Research](https://nosible.com/blog) Related Research ![The NOSIBLE Relevance Lab showing what Atlas links to Eli Lilly, with selpercatinib traced to Eli Lilly through Loxo Oncology.](https://nosible.com/images/2026/10/atlas-tickers-relevance-lab-2000x857.png) [NOSIBLE World](https://nosible.com/blog/tag/nosible-world) [Information Retrieval](https://nosible.com/blog/tag/information-retrieval) [Web Search](https://nosible.com/blog/tag/web-search) [Trading Signals](https://nosible.com/blog/tag/trading-signals) ### [The Hidden Path to Every Ticker: 15% to 22% Coverage, Zero Look-Ahead](https://nosible.com/blog/the-hidden-path-to-every-ticker) 2026-10-06 11 min read ![Magnifying-glass illustration representing vector search across company news](https://nosible.com/blog/illustrations/inspect.png) [Web Search](https://nosible.com/blog/tag/web-search) [Information Retrieval](https://nosible.com/blog/tag/information-retrieval) [Trading Signals](https://nosible.com/blog/tag/trading-signals) ### [Using Vector Search to See Signals in Company News](https://nosible.com/blog/using-vector-search-to-see-signals-in-company-news) 2024-01-21 21 min read ![Dotted NOSIBLE map showing seven oil chokepoint polygons and 5,261 matching World events for September 2026.](https://nosible.com/images/2026/10/world-geography-chokepoints.png) [NOSIBLE World](https://nosible.com/blog/tag/nosible-world) [Information Retrieval](https://nosible.com/blog/tag/information-retrieval) [Web Search](https://nosible.com/blog/tag/web-search) ### [Polygon Queries: Intelligence Across Space and Time](https://nosible.com/blog/polygon-queries-geospatial-temporal-intelligence) 2026-10-16 8 min read > Use NOSIBLE Atlas for supply-chain research, customer momentum and historical backtests. Work with a point-in-time company graph in your infrastructure. **URL:** https://nosible.com/blog/from-supply-chains-to-trading-signals-a-knowledge-graph-you-can-own