Reviewing proxy statement filings becomes repeatable when each tool passes identifiable evidence to the next stage. Filing search produces candidate records; retrieval preserves documents; extraction organizes disclosures; comparison creates questions for reviewers. The goal is a review trail that explains where each value came from and why a change was flagged.
SEC Filing Data is an independent third-party provider, not affiliated with or endorsed by the SEC. Its documented discovery and retrieval capabilities provide the foundation below. Parsing, AI analysis, comparisons, dashboards, and review queues are proposed customer-built or separately sourced steps.
How the tools fit together
Use a shared source record to connect the stages summarized below.
| Stage | Tool’s role | Output | Validation checkpoint |
|---|---|---|---|
| Identify | Match watchlist names and identifiers | Confirmed issuer and meeting context | Correct legal entity and event |
| Discover | Search website or query API | Candidate filing records | Form, date, and completeness |
| Preserve | Retrieve and catalog documents | Source files and version links | Main document, duplicates, revisions |
| Extract | Parse text and tables separately | Normalized values with passages | Headers, footnotes, units, years |
| Compare | Apply customer-defined comparison rules | Source-linked exceptions | Like-for-like values and missing data |
| Validate | Route questions and record decisions | Reviewed findings and error log | Evidence supports each resolution |
1. Identify the issuer and the right proxy statement filings
Start with a watchlist containing each company’s legal name, ticker where relevant, and Central Index Key, or CIK. Confirm that the company matches the entity you intend to research. A familiar brand name may differ from the legal filer, and a ticker alone should not settle an uncertain match.
Define the meeting or transaction before selecting a form. The SEC’s filing research guide distinguishes DEF 14A, the definitive proxy statement used in routine proxy contexts, from PRE 14A, a preliminary proxy statement where applicable. It identifies DEFM14A and PREM14A as definitive and preliminary merger-related proxy statements. These are relevant distinctions, not an exhaustive list of proxy forms.
SEC Filing Data’s proxy statement tracker lists definitive filings with form, filer, filing agent, date, and viewing actions. Use it to locate candidates, then open the selected filing and confirm its company name, meeting date, and purpose. Do not assume the newest document is the annual proxy you need.
Save a selection reason with the confirmed issuer and event. An annual director-election review needs different information from a merger review; do not apply one disclosure template indiscriminately.
2. Choose website review or API discovery
For a small watchlist, begin with the website. Search by company name, ticker, or CIK, select the matching company, and inspect its filing list. Check the form and date, open the relevant result, and use an available PDF download option when a readable saved copy is useful. Follow any sign-in prompts shown by the site.
When collection becomes repetitive, evaluate the SEC Filings API for connecting filing discovery to an internal application. The useful distinction is operational: website browsing suits individual checks, while API requests can supply filing records to a repeatable collection process.
The SEC filings search api lists company searches by CIK or ticker, form-type searches, and pagination controls. A company query can return several filing types; inspect the records and select the intended proxy statement filing. Conversely, a form-type query alone does not restrict results to your watchlist.
Your application would submit searches, inspect responses, request further pages, and select candidates. Record search inputs and collection time so omissions can be investigated. A failed request is an incomplete search, not evidence of no filing. This workflow does not depend on an undocumented feed or trigger; direct SEC.gov search remains a factual cross-check.
3. Retrieve the document and retain its identity
Discovery identifies a submission; retrieval obtains the disclosure you will analyze. Select the main proxy document rather than an unrelated attachment, and retain its original SEC source link.
The SEC EDGAR downloader describes original-document access and PDF generation. Confirm the current operation requirements before implementing requests. A generated PDF is a reading copy, not proof that the filing has been parsed into reliable governance data.
Create one source record per selected submission. Save the issuer name, CIK, form, filing date, accession number, original document URL, local filename, and retrieval date. These are proposed fields for your own records, not promised API response keys. Fill gaps from the source filing rather than guessing.
Preserve revised proxy materials (DEFR14A) and additional definitive proxy soliciting materials (DEFA14A) as separate records linked to the relevant meeting or transaction. Read each filing to determine whether it revises, supplements, or clarifies the original. Do not assume every additional-material filing changes substantive disclosures or that a later filing replaces the earlier one. Record additions and replaced disclosures based on the text. Identify repeat downloads using accession numbers and document URLs; retain earlier versions and exclude confirmed duplicates.
4. Organize disclosures with a separate extraction step
After retrieval, configure your own parser, document-processing tool, or AI extraction workflow. This is a proposed customer implementation, separate from SEC Filing Data’s filing-search and retrieval capabilities. Begin with a defined set of fields rather than asking a model for an unrestricted summary.
Feed the selected source document and a context-specific field template into that step. For an annual meeting, useful fields include director nominees, disclosed independence, committee membership, executive compensation, shareholder proposals, and board recommendations. A merger review requires a different template reflecting the transaction’s disclosures.
Use headings to locate sections, then reconstruct tables with their row labels and column headers. Carry headers across page breaks, attach footnote markers, and check that amounts remain paired with the correct executive and fiscal year. OCR may help when text is embedded in images, but its output needs checking against the image.
The output should pair each normalized field with its original wording, supporting passage, source-record identifier, and page or heading location. Normalize names, currency, scale, and fiscal years without discarding the originals. Distinguish “not found,” “not applicable,” and extraction failure; none means zero. Match recommendations to proposal text, since item numbers can change between years.
The SEC’s Investor Bulletin: Say-on-Pay and Golden Parachute Votes illustrates how a say-on-pay resolution may refer to executive compensation disclosed in the CD&A, compensation tables, and narrative discussion. Retain these connections so a copied total remains interpretable.
5. Compare corresponding disclosures and build a review queue
Match the current filing with the previous year’s corresponding meeting materials. Separate filing dates from the fiscal periods described inside compensation tables. When the same fiscal year appears in both documents, use that overlap to check extraction consistency before calculating a new annual change.
Apply explicit rules to comparable fields: changed amount, added nominee, altered recommendation, or missing disclosure. Check currency, scale, executive identity, and measurement basis before calculating differences. A missing value should trigger an extraction or coverage question; it does not establish that a policy or proposal disappeared.
A hypothetical example
Suppose a fictional issuer’s parser extracts the same executive’s Summary Compensation Table total as $6 million for fiscal 2024 and $7 million for fiscal 2025. All figures here are hypothetical. The normalizer records dollars, fiscal years, and matching table definitions. The comparison rule calculates a 16.7% increase and opens an exception.
The exception carries both source URLs, table locations, quoted rows, and footnotes. A reviewer opens them and finds a hypothetical $1 million one-time award in the newer disclosure. They verify its treatment and record “change confirmed; award component explains increase,” with their name and review date. If the parser missed a footnote, they also log an extraction defect. The flag prompted research; it did not establish excessive pay, misconduct, or a voting recommendation.
A customer-built spreadsheet or dashboard can assign exceptions to owners, track unresolved cases, and distinguish extraction corrections from substantive changes. Governance teams can use reviewed results in meeting briefs while retaining the underlying evidence.
6. Validate findings before expanding the process
Before scaling, manually review a small sample containing similar issuer names, dense tables, missing fields, and later materials. Create a checked reference set and compare automated output against it. Count missed disclosures and incorrect flags, identify recurring causes, and rerun affected cases after fixes.
Before accepting a finding, the reviewer should verify:
- Confirm issuer matches, document identity, and duplicate handling; mark superseded content without deleting its history.
- Inspect table-row alignment, continued headers, footnotes, fiscal periods, and units against the original.
- Check each proposal/recommendation pairing and whether later materials changed it.
- Record the reviewer’s decision, supporting evidence, error category, and any correction needed.
Repeat these checks on new batches and after parser or template changes. A falling error rate in one sample does not establish accuracy across every filing type.
Automate the repetitive work while preserving a clear path from every finding to the original disclosure. An extraction flag does not establish a legal violation, determine a vote, or replace investment judgment.
Start with a defined watchlist and a small set of review questions. Use SEC Filing Data for documented discovery and retrieval, validate your separate analysis steps, and expand only when reviewers can explain and reproduce the results.
