Which new filing would actually change your thesis?

Before an earnings release lands, write down the observation that would cause you to revisit your thesis. Then compare like periods and like definitions. A good change log gives a portfolio manager enough context to challenge a conclusion without reconstructing the analyst's entire reading list.

Gross profit rose. Did the thesis hold?

In a fictional issuer's first quarter, gross profit is $30 million on revenue of $100 million. In the second quarter, gross profit rises to $33 million on $120 million. Gross profit dollars increased 10%, while gross margin fell from 30% to 27.5%. An AI summary that calls this “improving profitability” leaves the portfolio manager to guess which observation matters.

If your thesis is operating leverage, the missing link is the cost structure and mix behind the change. If your thesis is absolute gross profit growth, the observation may support it. Neither calculation tells you why the margin moved. Record management's explanation separately and specify what evidence would discriminate between mix, pricing and input costs.

Keep four lines beside the model

  1. Prior claim. State the specific expectation, the date you formed it and the metric definition.
  2. New disclosure. Save a link to the filing and the table or footnote. Record units, reporting period and whether a comparison was restated.
  3. Interpretation. Show the arithmetic and name at least one alternative explanation the filing leaves open.
  4. Decision to revisit. Write the observation that changes your view, who will check it and when the evidence should arrive.

Use a public filing to test the process. The SEC’s EDGAR search provides the issuer documents. A release, a slide and a quarterly report may describe overlapping periods on different bases. Record the precise source for each figure before comparing them.

A better research-assistant test

Give the assistant two short extracts with conflicting metric definitions. Ask it to explain what can be calculated, show its sources and identify the comparison it cannot yet make. An answer that supplies an unsupported causal story should fail the exercise even when its arithmetic is right.

This worksheet applies to public-source fundamental research. For a related failure mode in private-company materials, read how forecasts get mistaken for actuals.