Best practices for Chat bot

How to get the best answers, and improve productivity.

How to think about querying AI systems like Matterfact

1. Is the data contained in Matterfact’s index?

Matterfact can generate responses based on information that is explicitly stated or derived from the content in our index which is composed of 10Ks, 10Qs, earnings call transcripts and press releases of the top 2000 publicly traded US-issuers.

Scenario: If you ask for a piece of data that is more fine-grained or granular than what has been shared by the company. Tip: Broaden the question How many iPhones did $AAPL sell in 2021? → What did $AAPL report regarding iPhone in 2021? What is $PGR's target addressable market? → What has $PGR shared about its market size, market share/position?

2. How specific is your question?

Although Matterfact can handle broad questions, specific questions on concrete data will yield better responses.

Scenario: If you have a broad topic you are trying to answer. Tip: Add specific details to the question to help guide AI. Give me the financials for $AAPL → Generate a table of $AAPL’s total revenue, cogs and earnings for FY2024. What is the culture of $UNH? → What is $UNH’s leadership and employee development policy and initiatives? What company or corporate values has $UNH management shared?

3. How subjective is your question?

Matterfact’s index primarily reflects company management's perspective. Its ability to offer subjective opinions or speculative analysis is limited by our data corpus.

Scenario: If you are trying to answer a question where the response is subjective. Tip: Transform subjective questions into objective requests that are informed by your preferences. Will $MSFT artificial intelligence strategy win over $GOOGs → What has $MSFT shared about its AI strategy? What has $GOOG shared about its AI strategy? Compare the two approaches on XYZ dimensions. Generate a summary of $CMG’s last earnings call. → Generate a summary of $CMG’s last earnings call. In your summary include sections on last quarter’s performance and quantitative guidance for the upcoming quarter and year.

4. The smaller the question, the more accurate the answer

The scope of a data request can impact response quality. Requests for excessive data may lead to less precise answers.

Scenario: If you are trying to answer a question about multiple pieces of data and tickers. Tip: Break the question down into smaller requests.

What is $MSFT, $NVDA, $GOOG, and $AMD revenue by segment, revenue by geography, gross profit and net income for the last 10 years → Break down into smaller questions per ticker.

This week, expect more content/videos from us as we show you what kinds of questions you can our chatbot.

Examples of specific queries

Specific data (quantitative or qualitative) from one period or across periods

Matterfact can help you extract specific data points quickly.

Extract data as granular as reported in filings and also across time periods if reported

Find me the total number of $DPZ stores as of FY2024
Find me the total number of $DPZ stores in China from FY2022 to FY2024
How many US franchisees does $DPZ's network include?

Use natural language shorthand for dates; your query also does not have to include a period

What was $ROST's revenue for the last four quarters?
When and what price did $V acquire Pismo for?

Query multiple tickers for specific data simultaneously

What was $AAPL and $GOOG revenue by segment for the last three years?
What was $BA and $LMT backlog as of FY2024?

Categorical Questions

Matterfact can help you learn standard and company-specific categorical questions. These questions are typically based on a specific category of content typically disclosed in the company’s filings. The below is not an exhaustive list because each company may disclose company specific categories (i.e., $BA for aircraft, $PFE for therapeutic areas)

Merger and Acquisitions

What companies did $V acquire between 2022 and 2024 and for what price?

Customer Analysis and Concentration

What customer concentration if any does $BA have?
Who are $BA’s major customers?

Segments

What segments does $PGR operate in?
What countries does $DPZ operate in?

Risks

What risks has $DPZ disclosed in their latest 10k?
Does $UNH have any outstanding litigation?

Management Future Plans

What were the strategic priorities of $UNH reported in 2024?

Time-Series Trend Analysis

Matterfact can help you quickly identify trends across time, including calculating YoY or QoQ rates of change.

Trend analysis

What have been $PFE's revenue trends by therapeutic segment from 2021 to 2024?

Add a simple time-series calculation

What are the key revenue streams for $MCO? Get the last four years and calculate year-over-year change for each revenue stream. Put the results into a table.

Trend qualitative context

What drove $DPZ’s revenue growth in Q4 2024? What reasons did management provide in the earnings call

Simple Derived Metrics

Matterfact can calculate simple derived metrics (i.e., numerator/denominator) - margin analysis for example. It is best practice to include the definition of the calculation.

Provide a table of $BA's revenue, total costs and gross profit from 2021 to 2024. Calculate gross margin (gross profit / total revenue)
Provide a table of $BA's revenue, total costs and gross profit from 2021 to 2024. Calculate gross margin (gross profit / total revenue)