Browse documentation

Ask a question. Get a defensible answer.

You do not need to learn FactIQ's tools before using them. Start with the decision you face, add a clear scope, and say what evidence would convince you.

On this page

Start here

A strong request needs only three things: the question, the scope, and the proof you expect.

Question

Ask what changed, how two things differ, or whether a claim holds up.

Scope

Name the companies or geography and the period that matters.

Proof

Request the quotes, data, citations, and caveats needed to trust the answer.

Run your first research question

Paste this into Claude, ChatGPT, Claude Code, or Codex after connecting FactIQ. It is specific enough to research and broad enough to surface a real disagreement in the data.
Consumer-health research
Has the US consumer weakened since the start of 2024? Give me the answer first, then show the strongest evidence for and against it. Use household spending, inflation, real earnings, and consumer credit. Keep the time periods and units comparable, cite every series, and tell me where the evidence is mixed or incomplete.

The prompt asks for a conclusion, tests it from several angles, and makes uncertainty visible. Swap in your own company, sector, market, or hypothesis.

Turn a topic into a researchable question

Most disappointing answers begin with a topic—“retail,” “AI,” “inflation”—instead of a question that can be resolved. Add these three pieces.
  1. 01

    Ask one thing

    Frame the work as a change, comparison, or test: “What changed?”, “How do these companies differ?”, or “Does the evidence support this claim?”
  2. 02

    Draw the boundary

    Name the companies, geography, and dates. “WMT, TGT, and COST from 2024–2026” is more useful than “large retailers recently.”
  3. 03

    Define proof

    Say what the answer must show: exact call quotes, a peer comparison, official data, a chart, or a short evidence table with citations.

From broad to useful

Instead of

“What are retailers saying about the consumer?”

Ask

“Has the consumer trade-down story strengthened or weakened across WMT, TGT, and COST calls since 2024? Show the conclusion first, then exact supporting and conflicting quotes with speakers, call dates, fiscal periods, and citations.”

Choose the answer shape

The same question can produce a sentence, a chart, or a full research note. Ask for the smallest output that will help you decide.

One fact

Ask for the value, date or period, unit, and source.

A trend or comparison

Ask for a chart with a finding-led title, comparable units, dates, and source line.

A company-research note

Ask for the conclusion, evidence for and against it, exact call evidence, and a watch list.

A deeper investigation

Ask for sections organized around findings, not sources, plus methodology and unresolved questions.

For company research, continue to earnings calls. To make an answer reviewable by someone else, use the evidence guide.

If the first answer is weak, change one thing

Do not respond with “try again.” Tell the assistant what failed, then tighten the question or evidence.

Too generic

Name the companies, dates, and decision. Ask for one conclusion rather than a tour of the topic.

Thin evidence

Require exact quotes or data points with speaker/source, period, units, and citation.

No useful matches

Check coverage, shorten the theme, and try the product, segment, or phrase the company itself uses.

Too many results

Split the work by company and year or fiscal period, preserving the evidence from each batch.

Too confident

Ask for the strongest contrary evidence, missing coverage, and what would change the conclusion.

Numbers conflict

Check units, time periods, seasonal adjustment, nominal versus real values, and fiscal versus calendar dates.

Use this template for your own question

Replace the brackets; leave the answer structure intact.
Research prompt template
Help me answer: [the question or decision]. Focus on [companies, sector, or geography] from [start date] to [end date]. Use [the evidence that would change my mind]. Start with the conclusion. Then show the evidence for it, the strongest evidence against it, and what remains uncertain. Include dates, units, exact quotes or source links where available, and note any important coverage gaps.

If a company, period, or dataset appears to be missing, ask the assistant to send FactIQ feedback with the exact ticker or source, date, expected result, and observed result—without personal information. To confirm the connector ran, see the client-specific invocation check.