What “with sources” should mean
A source-backed answer is not a paragraph followed by a list of links. The source must be connected to the claim it supports, and the reader should be able to reach it. For important claims, capture the relevant location, publication date, publisher, scope, and any limitation that changes interpretation.
Source quality depends on the question. Primary records, official statistics, filings, product documentation, and original research are usually stronger for factual claims than an unsourced summary. Expert analysis can be valuable for interpretation, but it should not silently replace the underlying evidence.[3] [4]
Authority
Does the source have direct access, responsibility, or demonstrated expertise for the claim?
Recency
Is the publication or data period current enough for the decision being made?
Scope
Do geography, segment, sample, definition, and time period match the claim?
Corroboration
Would another independent source support a consequential or surprising conclusion?
Turn a broad topic into a researchable decision
“Research the market” invites an encyclopedic summary. A useful brief names the decision, audience, alternatives, geography, time horizon, constraints, and evidence standard. It also separates questions that need facts from questions that require judgment.
- Decision: what will change if the research is convincing?
- Audience: who will review the evidence and what objections will they raise?
- Scope: which market, customer, geography, product, and period are included or excluded?
- Alternatives: which options or hypotheses should be compared?
- Evidence standard: which claims require primary or independent corroboration?
- Freshness: how recent must product, market, pricing, or regulatory information be?
- Output: memo, evidence table, presentation, spreadsheet, or another work result?
Use an evidence ledger while researching
An evidence ledger turns browsing into a reviewable process. Each row records the claim or question, source, publisher, publication date, relevant passage or table, scope, confidence, and how the evidence affects the decision. Conflicting evidence should receive its own rows rather than being averaged away in prose.
The ledger also prevents citation drift during revision. When a paragraph changes, the reviewer can check whether the original source still supports the new wording. The same ledger can feed a report, slide source notes, or assumptions tab in a model.
Claim
Write the smallest statement the source is expected to support.
Evidence location
Record the exact page, section, table, timestamp, or accessible passage.
Interpretation
Explain what the evidence does and does not imply for the decision.
Confidence
Label verified, supported, directional, assumption, or unresolved with a short reason.
Five research failure modes to catch
The most dangerous research errors often look polished. A verification pass should actively search for predictable failure modes instead of asking the system whether it is confident.
- Citation mismatch: the link is real but does not support the sentence.
- Scope substitution: evidence from another market, segment, period, or definition is presented as equivalent.
- Stale facts: prices, product capabilities, roles, laws, or market conditions may have changed.
- False precision: a weak estimate is reported as a precise fact without range or confidence.
- Missing counterevidence: the synthesis ignores credible evidence that weakens the preferred conclusion.
Convert evidence into a decision-ready work result
A research summary should answer the question, not replay the search process. Lead with the conclusion and confidence, show the evidence that matters, explain contradictions and limitations, and state what the team should do next. Put supporting detail in the evidence ledger or appendix.
Citation list vs evidence system
The difference is traceability: can a reviewer understand exactly how the source supports the decision?
| Research control | Citation list | Evidence system |
|---|---|---|
| Connection | Links appear at the end of the response | Each material claim maps to the source and relevant location |
| Quality | Sources are treated as interchangeable | Authority, recency, scope, and corroboration are assessed |
| Conflict | One answer is selected silently | Contradictions and unresolved questions remain visible |
| Reuse | The reader reconstructs evidence for each new format | The ledger supports documents, slides, spreadsheets, and later updates |
A six-stage sourced research workflow
Use this sequence for market, competitor, customer, policy, product, and strategic research where the result must survive review.
- 01
Frame the decision
Outcome: A bounded research brief with questions, scope, audience, and evidence standard.
- Write the decision and the hypotheses or alternatives being tested.
- Define freshness, geography, segment, and source requirements.
- 02
Map the source hierarchy
Outcome: A plan for primary evidence, official records, independent research, and interpretation.
- Start with sources closest to the underlying fact.
- Use commentary to explain evidence, not to replace it silently.
- 03
Collect evidence into a ledger
Outcome: A structured claim-to-source trail with scope and confidence.
- Record exact locations and dates as sources are read.
- Capture counterevidence and conflicts in separate entries.
- 04
Synthesize by question
Outcome: A conclusion that reflects the strongest evidence and visible uncertainty.
- Group findings by decision question rather than by source.
- Distinguish observed facts, calculations, assumptions, and recommendations.
- 05
Verify material claims
Outcome: A reviewed set of citations and calculations for high-impact conclusions.
- Open each source and test claim fit, scope, and recency.
- Recalculate material derived values independently.
- 06
Deliver and preserve the evidence
Outcome: A memo, deck, or model linked to its ledger and ready for revision.
- Lead with the answer, confidence, implications, and next actions.
- Keep the evidence base attached so future updates do not start over.
Prompts you can use
Replace the bracketed details, attach the relevant source material, and keep the review step in the same workspace.
Research brief
Prompt 01Research [decision] for [audience]. Scope: [market, geography, segment, period]. Compare [alternatives]. Prioritize primary and official sources, require independent corroboration for consequential claims, and state the freshness cutoff. Deliver a recommendation, confidence level, evidence ledger, conflicts, limitations, and open questions.
Why it works: It converts a topic request into a bounded decision with an explicit evidence standard.
Citation audit
Prompt 02Audit every material claim in this research. Open the cited source and record whether it supports the exact wording, matches the scope and date, and remains reachable. Rewrite overbroad claims, replace weak sources where possible, and mark unresolved items.
Why it works: It tests claim-to-source fit instead of accepting a bibliography as proof.
Counterevidence review
Prompt 03Act as a skeptical reviewer. Find the strongest credible evidence against the current conclusion, identify hidden assumptions and scope gaps, and explain which new evidence would change the recommendation. Update the confidence level after the review.
Why it works: It reduces confirmation bias and makes uncertainty decision-relevant.
Editorial method
How this guide was prepared
The Kona Team prepared this guide from source-backed research controls used across business planning workflows: decision framing, source hierarchy, evidence ledgers, claim-level citations, conflict handling, confidence labels, and downstream artifact traceability.
Read Kona’s editorial standardsSources
Sources and benchmarks
[1]
Using connectors in ChatGPTOpenAI Help Center
[2]
Introducing ChatGPT connectors betaOpenAI · 2025-08-11
[3]
Creating helpful, reliable, people-first contentGoogle Search Central
[4]
Write your business planU.S. Small Business Administration
[5]
Data connectors and analytics platformKona Business AI
[6]
AI business planning workspaceKona Business AI
[7]
AI data governance and metric opsKona Business AI