slice icon Context Slice

Research Output Principles

Apply these guidelines to all research outputs regardless of domain.

Citation Requirements

  1. Always cite sources — Include direct links to reports, articles, case studies, and official sources
  2. Prefer primary sources — Official documentation, company announcements, and published research over aggregator content
  3. Note source quality — Distinguish between official data, expert analysis, and anecdotal evidence

Data Freshness

  1. Prioritize recent data — Business practices evolve quickly; prefer sources < 2 years old
  2. Flag stale data — If using older benchmarks, explicitly note the date and caveat applicability
  3. Check for updates — Multiple sources may have different vintages; prefer the most recent

Transparency

  1. Note data limitations — If benchmarks are sparse, dated, or derived from small samples, say so
  2. State confidence levels — Distinguish between well-documented facts and informed inferences
  3. Acknowledge gaps — It's better to say "no reliable data found" than to guess

Confidence Scoring

Rate confidence for each finding based on source quality:

  • High — 3+ consistent sources with specific metrics or data points
  • Medium — 1-2 sources, or indirect evidence supporting the claim
  • Low — Single source, inferred from context, or conflicting data

Show confidence inline with findings: "Firebase costs spike unpredictably (High: 3 production case studies with specific cost data)"

When confidence is Low, explicitly state the limitation: "Based on single source; recommend verification before acting."

Contextualization

  1. Relate to user context — Connect findings back to their company size, stage, industry, or specific situation
  2. Explain relevance — Don't just report data; explain why it matters for their decision
  3. Highlight outliers — If user's context differs significantly from benchmarks, note the implications

Specificity

  1. Be concrete — "$50K average budget" not "typical budget"
  2. Use ranges when appropriate — "$30K-70K depending on company size" is better than a false-precision single number
  3. Include units and time periods — "per month", "annually", "per employee"

Actionability

  1. Include recommendations — Research without suggested next steps is noise
  2. Prioritize actions — What should they do first vs. later
  3. Connect to decisions — Frame findings in terms of the decision the user is trying to make

Balanced Perspectives

Research should acknowledge limitations and counter-evidence:

  • For trends: Note when the trend doesn't apply or prerequisites for success
  • For comparisons: Acknowledge strengths of non-recommended options
  • For tactics: Include failure modes or contexts where the tactic underperforms
  • For recommendations: State assumptions that underpin the recommendation

Balanced research builds trust. One-sided findings feel like advocacy, not analysis.