Build a causal map for an investment thesis
A causal map turns an investment narrative into a sequence of testable links. Begin with an operating change, connect it to customer behavior and unit economics, then identify how it could affect cash. A map is useful because a correct observation at the beginning does not guarantee the outcome at the end.
Write one arrow at a time
For each arrow, ask what must remain true, what could interrupt it, and what would be observable. “A new factory creates value” hides several questions: whether it opens, whether output meets specifications, whether customers buy, whether selling prices cover costs, and whether payment arrives before financing is exhausted.
Annual reports separate business description, risks and financial statements. Use those sections to check different links. The map itself is your inference, so do not attribute its entire conclusion to a source that supports only one input.
Hypothetical map: opening a production line
Assume a manufacturer plans an extra line with capacity of 50,000 units. Existing customer orders are not disclosed. The attractive headline is capacity growth, but unused capacity cannot create operating receipts on its own.
| Link | Necessary condition | Candidate check |
|---|---|---|
| Installed line → saleable output | Acceptable yield and certification | Commissioning disclosure, scrap rate |
| Output → sales | Demand at viable prices | Comparable shipments and realized price |
| Sales → operating profit | Incremental costs remain controlled | Gross profit and startup costs |
| Profit → cash | Customers pay; inventory does not accumulate | Receivables, inventory and operating cash |
If only 20,000 units can be sold initially, applying full capacity to a revenue estimate assumes away the bottleneck. Record that assumption rather than quietly treating installed capacity as utilization. Likewise, the first shipment can establish technical progress while providing little evidence of repeat demand.
Check for competing paths
A map should include at least one alternative route. Sales might rise because a competitor temporarily stopped shipping, rather than because the new line improved product quality. Gross margin might rise because raw materials became cheaper, rather than because the line is efficient. Both developments can be financially helpful, but they imply different persistence.
Look for feedback loops as well. Excess output can increase inventory, pressure discounting and weaken cash collection, which then limits the funding available to improve yield. Showing only positive arrows produces a promotional diagram rather than a research tool.
Verify the date order: the alleged cause must precede the proposed effect, allowing for production and disclosure lag. Check whether each observation covers the same product and plant. Group-level results can hide an unsuccessful project inside a larger profitable business.
A causal map does not establish causation statistically and is unsuitable for proving an effect from correlation alone. Its practical output is a prioritized list of links to investigate. Start with the weakest necessary link, not the most visible event.
A further check on the example
For the production-line map, add a source date and expected lag to every arrow. Certification may precede shipment, shipment may precede customer acceptance, and acceptance may precede cash collection. If those observations come from different product generations, the chain is not continuous. Mark the break explicitly. A useful review starts by asking which arrow received new evidence, then which downstream conclusions remain assumptions; it does not mark the entire map successful because one engineering milestone was achieved.
FAQ
How many nodes should the map contain?
Enough to show the decisive operating steps. If every corporate activity appears, split the map around the question that could change the conclusion.
What if two drivers move together?
Keep both explanations open. Seek a natural comparison, disclosed bridge or later observation that can separate them; do not assign precise causal shares without evidence.