The method, including its limits.

Meriax works from material that is already public and can be checked by anyone. Each input is recorded with its date and identifier so a readout can be traced back.

• Public company filings, including annual reports, quarterly reports and current reports.
• Company announcements and press releases published by the issuer.
• Historical price and volume series from a third-party market-data source.
• Public sector and industry classifications used to assemble a stated peer set.

We do not use paywalled research, leaked documents, non-public information or purchased ratings. If a source cannot be named, it does not enter a readout.

The model reads the source set and produces three things: a change detection pass, an anomaly flag and a short digest. It is a document and data tool, not an opinion engine.

• Compares the current filing with the prior version and marks changed passages.
• Flags unusual price or volume readings against a defined historical window.
• Summarises the changed passages while keeping the source reference attached.
• Formats the result as a dated readout with a visible source trail.

Inputs are public documents and historical series; outputs are descriptions of what changed. We do not promise an accuracy rate, a coverage rate or a particular result.

• It does not give price levels, ratings or transaction instructions.
• It does not tell you what to buy or sell, or how much to hold.
• It does not give personalised advice or manage money.
• It does not say where a price will go, and it does not call a market direction.
• It does not rank securities as better or worse investments.

A readout describes an event in a document and the context around it. What you do with that is your decision, made with your own licensed adviser if you have one.

Every output carries a date, a source trail and a model version. A sample of readouts is read by a person against the underlying filing, with attention to changed numbers, quoted language and the source references.

When an error is found, the readout is corrected and re-issued with the correction noted; the original version stays in the archive. To report an error, email [email protected] with the readout date, the ticker and the passage in question. We do not publish a validation score because a sampled review is not an accuracy guarantee.

• Filing scope can change between periods, so like-for-like comparison is not always possible.
• One-time items, restatements and changed accounting language can distort a simple trend.
• Halted, thinly traded or low-liquidity securities may have price and volume series with gaps.
• Third-party data can arrive late, be revised, or carry an error we have not yet caught.
• Document extraction can misread a table, a footnote or a cross-reference.
• The model summarises what is in the source; it cannot know facts that are not in the record.

These limits are part of the product, not an edge case. A readout should be checked against the source before anyone relies on it.

We do not publish investment performance, a track record, a success percentage or a comparison against a benchmark. We do not claim that using the product improves a result.

The sample on the Market page is an illustration of format. Its figures are examples, not reported results, and it is not a current view. Past price or volume behaviour does not guarantee a future result.