Overview
The Research feature in Harvey is designed to facilitate research across various domains, enabling users to search through extensive proprietary and public data sources. Within this feature, you’ll find EDGAR Research, specifically created to help users search, access, summarize, and analyze SEC filings, including forms 10-K, 10-Q, 8-K, and DEF14A.
EDGAR (Electronic Data Gathering, Analysis, and Retrieval) is the primary system used by the U.S. Securities and Exchange Commission (SEC) for submissions by companies and other entities required by law to file information with the SEC.
EDGAR
Select the 'EDGAR' feature from the Research section in Harvey.
In EDGAR, Harvey will answer questions or generate new content based on current reports, annual reports, quarterly reports, and annual proxy statements. The output includes linked references to highlighted sections of source documents for easy verification of Harvey’s work.
Key Considerations:
Strengths and Limitations: Harvey excels at understanding language but struggles with numerical operations and recognizing pictures and data tables.
Comparative Analysis: Harvey may not provide the best results when comparing entire filings. It may struggle with queries that require a thorough understanding of multiple documents.
Variability in Answers: Responses may vary based on subtle changes in query language and across multiple submissions. Refine your query to explore different possible answers.
Automatic References and Citations: References and citations are generated automatically, so there is no need to request them explicitly.
Source Volume: Harvey can take in many sources simultaneously and find the information that is relevant to your query. However, when a very broad query is asked and/or multiple filing types are selected, the volume of potential sources can slow down our systems and, in some cases, result in a processing error. When possible, consider framing your query in a tailored, specific way and selecting only the filing types that are most relevant.
How to Run a Query
Enter your research query in the open text box.
Select at least one knowledge source from the Filter menu, or load a Library prompt via the 'Load Prompt' option.
Click 'Ask Harvey' to get relevant findings.
You also have the ability to 'Save Prompt' if you want to use this exact prompt again.
Querying Tips
Before asking Harvey to run your query, please keep the following tips in mind:
Select Filing Types: Choose the specific filing types you want Harvey to search (e.g., 10-K, 10-Q, DEF 14A, 8-K).
Specify Filing Years: Indicate the filing year(s) you want Harvey to search. If not specified, Harvey will infer the relevant years based on your query. Harvey can review multiple years of filings simultaneously.
Identify Companies: Provide the name, stock ticker, or industry to identify the company or companies you’re interested in.
Include Exact Ticker Symbols: For companies that have changed their stock ticker after 2021, include the exact ticker symbol in your query to help Harvey locate the correct information on EDGAR.
Specify Format: Consider requesting information in specific formats, such as a table, email, memo, outline, or bullet points.
Specify Tone: Indicate the desired tone for the response, such as informal, professional, or firm.
Identify Audience: Specify the intended audience (e.g., “a client with little acquisition experience” or “a senior in-house attorney”).
Be Explicit and Specific: Make your queries explicit, descriptive, and specific. Providing extra direction can significantly improve results (e.g., use “holders of Company Common Stock” instead of “common holders” or “written legal opinion” instead of “legal opinion”).
Break Up Complex Queries: Divide complex queries into simpler parts, especially when they involve analytical or subjective tasks. We recommend one task per sentence in your query.
Example Queries
Find examples of technology companies that have filed 8-Ks regarding changes to management or the board of directors.
Provide examples of insurance companies that recently disclosed amendments to governing documents.
How are real estate companies disclosing risks relating to remote work?
What are credit card companies saying about product innovation in recent 10-Ks?
Frequently Asked Questions
Q: Should I use Assistant or EDGAR for research?
A: Our EDGAR Research feature is much more accurate than Assistant for queries that require underlying source material from these datasets, and also provides citations and links to the original sources.
Q: Does Research/EDGAR have risk of hallucinations?
Harvey tools, such as Research, inherently mitigate hallucination risk through a process of “grounding.” If you upload a document to Harvey, or search over a database (e.g., SEC EDGAR), Harvey’s response will be grounded in a source document and is therefore significantly less likely to hallucinate (as compared to an open ended workflow). Furthermore, our algorithms incorporate multiple models, and we monitor how different models perform at different tasks in a process chain and use the model that works the best for the particular job.
While hallucination risk is never zero, our investment in citations, examples, and proprietary algorithms significantly mitigates hallucination risk. Think of Harvey as a digital associate - i.e., a useful, very fast thought partner and generator of first drafts. The output requires verification, much like the work of a junior colleague, and saves significant time and enhances overall quality.
Q: What’s the earliest date supported in EDGAR for filings?
January 1st, 2001
Q: How quickly do newly published filings become available?
Within 5 to 15 minutes
For further assistance, please contact [email protected]