> For the complete documentation index, see [llms.txt](https://matterhorn-doc.mometic.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://matterhorn-doc.mometic.com/reference/faq.md).

# FAQ

Answers to common Matterhorn questions about AI scoring, ML beta, data coverage, demand evidence, refresh behavior, watchlists, and research confidence.

## Does the language model choose the scores?

No. Deterministic code computes standard scores, and a frozen tabular model supplies the ML signal. Ask and analyst commentary write over retrieved records and contribute no scoring weight.

## Is 70/30 just the old Matterhorn score plus AI?

No. The 70% component side uses the ML release's financial/evidence family aggregate and separate durability estimate. It is not the original five-component composite. See [the blend](/understand-the-scores/ml-enhanced.md).

## Does turning on ML alter every panel?

No. The supported rank and radar surfaces change. Standard company research, the checklist, Compounding, History, Movers, theme scenarios, and standard analytics retain their own basis. See [the mode-change table](/understand-the-scores/ml-enhanced.md).

## Why can a high-ranked company have weak demand or missing valuation?

The rank aggregates several areas. Other strengths can outweigh a weakness, and missing inputs follow explicit neutral handling rather than becoming observed favorable data. Inspect coverage and the thesis-critical gaps before forming a conclusion.

## Are earnings transcripts incorporated?

The pipeline can ingest transcripts and extract supported demand or operating observations with provenance. Coverage varies by company, period, and provider. A new transcript in the ordinary evidence record does not automatically mean it has entered the frozen ML bundle.

## Does the model consider debt and dilution?

Yes, supported leverage and share-dilution features are part of the quality family, and the ordinary financial checklist also considers related measures. Missing evidence and industry applicability affect coverage.

## Why is there no backlog for some companies?

Some businesses do not report backlog, do not use a comparable definition, or do not have that economic model. A blank is not automatically a demand failure. See [demand measures](/investigate-the-business/forward-demand.md).

## Why did Refresh leave the ML ranking unchanged?

It reloads a frozen publication. Updating source data, rebuilding features, scoring with an explicit model, and publishing a validated bundle are separate steps.

## Does a verified citation make the conclusion correct?

It checks the relationship between a numeric claim and its cited record under implemented rules. It does not prove the author's interpretation, causal claim, or investment thesis.

## Is a high rank an instruction to buy?

It is a research priority. Examine valuation, current conditions, evidence gaps, and your own decision criteria. Matterhorn does not execute orders or manage a brokerage account.

## Does a low standard emergence score rule out a good investment?

No. Remaining scale, the measured rate, and durability conditions influence that score. A mature compounder can offer a different opportunity from an early growth company. Do not force every thesis to require a 100× outcome.

## Can I save research notes or export all results from this interface?

The documented interface supports browser-local pins and comparisons. It does not provide a general thesis notebook or a general CSV export button. Keep dated notes externally and include links to the company evidence you used.

Related: [troubleshooting](/keep-research-current/troubleshooting.md), [glossary](/reference/glossary.md).