# So What > Use this skill to turn evidence, analysis, estimates, and numbers into concise audience-facing conclusions that explain why the finding matters and what judgment, focus, metric, decision, action, or intended reaction it supports. Apply it when creating or reviewing slide decks, takeaway headlines, executive updates, talking points, readouts, decision briefs, or other narratives where facts need meaningful comparisons, audience relevance, implications, recommendations, or a coherent storyline. - Skill: `lifelonglazylearner/so-what` (Agent Skill, multi-file: 5 files) - Install (CLI): `npx skillmds@latest add lifelonglazylearner/so-what` - Raw SKILL.md: https://api.skillmd.com/api/skills/lifelonglazylearner/so-what/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: lifelonglazylearner (https://skillmd.com/u/lifelonglazylearner) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/lifelonglazylearner/so-what --- # So what Build a case the audience can understand and use. Ground the case in evidence, tailor it to the room, and never manufacture business context to make a finding sound important. ## Apply the core standard Make every material section help the primary audience answer: - What is the answer? - Compared with what, and for whom? - Why should I care? - What does this change? - What decision or intended reaction follows, if any? Keep this logic beneath the prose. Do not mechanically label every paragraph as a finding, implication, or action. Remove content that does not improve the audience's understanding, judgment, or ability to act. Move useful but nonessential detail to supporting material. Match the rhetorical job the audience asked for. If the request is for an estimate, present the estimate and its basis. Do not add an unsolicited recommendation, warning, approval threshold, or call to action. An informed audience may need a clear case rather than advice. ## Build the case ### 1. Ground the work Read the source material and inspect available context before drafting. Identify: - Supported facts, numbers, and comparisons - Interpretation supplied by the source or user - Decisions, recommendations, and unresolved questions - Limits that affect what the evidence can prove Research available sources when business context is missing. Do not turn correlation into causation, present judgment as fact, or invent a consequence. State an important unknown plainly. Keep limits that do not change the conclusion in supporting material rather than turning them into the headline. ### 2. Establish the audience contract Determine: - The primary audience and any secondary audience - What the primary audience cares about - What they already know - The decision, action, or reaction the content should produce Infer these from the request and available context first. If one critical element remains unknown and guessing would materially change the case, ask one focused question and wait. For mixed audiences, optimize the main narrative for the primary audience and add only the context secondary audiences need. ### 3. Form the answer first Write the direct answer in one plain sentence before organizing the supporting material. Give each material section one rhetorical job, such as establishing the opportunity, focus, stakes, success metric, or action. Reason upward through this ladder: 1. Evidence: what the sources establish 2. Audience consequence: why that evidence matters to this audience 3. Judgment or action: what focus, choice, metric, decision, action, or intended reaction follows Use the highest rung that the evidence supports and the audience contract requires. Do not stop at a chart fact when the section's job is to communicate a supported consequence or judgment. Do not climb to a consequence, causal claim, or recommendation that the evidence cannot support. Keep evidence, interpretation, and recommendation distinct in the audience-facing narrative without forcing those labels into the final prose. Include a recommendation only when the request or audience contract calls for one, and trace it back to evidence and audience needs. When a section's job is to select a metric, name the recommended metric directly only when the audience needs that choice and the evidence supports it. ### 4. Give numbers meaning For every important number, provide the comparison that makes it meaningful: a target, baseline, trend, benchmark, prior period, alternative, or affected population. Include the relevant unit, time period, and scope. At the first use of a percentage or modeled lift, make these elements recoverable without guesswork: - Denominator or comparison baseline - Affected cohort - Time period - Whether the value is observed, modeled, or assumed Translate modeled lifts into the absolute outcomes and timeframe the audience requested. For a scenario model, define each input in plain language and identify whether its value comes from observed evidence, a sensitivity adjustment, or a scenario assumption. Calculate and state simple absolute or relative differences when the source values support them. Prefer the transparent calculation to evaluative words such as "strong," "weak," "significant," "substantial," or "meaningful" unless the source defines that threshold or the text immediately states the audience consequence. Put a number in a headline when it is the requested answer, materially changes the judgment, or expresses the stakes. Otherwise, use it as supporting evidence. Preserve the requested magnitude and timeframe in estimate headlines. Explain the practical consequence supported by the evidence. If no valid comparison or consequence is available, state that limitation instead of implying significance. Do not dump numbers that do not change the audience's understanding or decision. Show a calculation once; do not repeat it in prose when a table or chart already communicates it. ### 5. Write for the room Use answer-first ordering adapted to the medium. A full brief may use: 1. Conclusion 2. Business or audience consequence 3. Supporting evidence 4. Decision, action, or intended reaction when applicable 5. Method and detail only where needed for trust Short formats may need only the direct answer and the support required to understand it. Choose the simplest presentation form that makes the audience's required judgment fastest and least error-prone. Use prose, a table, a visualization, a diagram, or a prototype only when that form communicates the relevant comparison, pattern, relationship, mechanism, or behavior more efficiently than a simpler alternative. Give each form one rhetorical job, preserve the evidence boundary, and remove it when it does not improve the audience's understanding or decision. Write slide and section titles as plain-language takeaways, not topic labels. State the audience-relevant outcome and include its magnitude and timeframe under the headline-number rule above. Keep cohort definitions and methodological qualifications outside the title once the scope is clear; do not turn the title into a scope dump. Define technical scope at its first meaningful appearance, then use concise shorthand consistently. When the medium has multiple layers, give each one a distinct job: - Headline: audience-relevant answer; magnitude and timeframe only when requested, judgment-changing, or needed to express the stakes - Definitions: model concepts - Table or chart: scenarios and calculations - Footnote, appendix, or expandable detail: baseline, provenance, assumptions, and limits Connect each material finding to its meaning, the audience impact, and, when relevant, the next step or intended reaction. Consider customer behavior, growth, cost, risk, operational load, strategic fit, and decision timing. Use only the dimensions that the evidence supports and the audience cares about. Do not force every finding into a revenue claim. An informative presentation may seek understanding or alignment rather than a decision. State that intended reaction clearly instead of inventing an action. ## Review existing content Match the requested review mode. If the user asks for critique, report findings without rewriting. If the user asks for revision, fix the supported issues while preserving source fidelity. Look for: - Naked numbers with no meaningful comparison - Topic-only or descriptive headlines - Facts followed by no interpretation - Generic implications that could apply to any business - Claims that exceed the evidence - Recommendations with no evidentiary path - Missing stakes, tradeoffs, owners, or timing when required by the decision - An unclear decision, action, or intended reaction when the audience contract calls for one - Methodology that obscures the conclusion - Percentages whose denominator or comparison baseline is unclear - Estimates without an explicit period or annualization - Observed inputs and scenario assumptions presented as equivalent evidence - Full cohort qualifications repeated after the scope is established - Prose that duplicates a table or chart - Unsolicited warnings, recommendations, or approval thresholds - Supporting evidence that is relevant but does not help answer the requested question Flag missing context that cannot be recovered instead of filling it with plausible language. ## Run the skeptical-audience review Before finalizing, apply the core standard once as a busy, skeptical member of the primary audience. Confirm that the evidence supports the conclusion, keep visible any validity caveat that could change it, and remove repetition or detail that does not improve understanding or judgment. For decks, run four checks: - Representation and redundancy: Is each element the simplest truthful form for the intended judgment? Does a chart, table, diagram, or prototype add understanding beyond a simpler alternative, and does the headline add consequence or judgment instead of merely restating it? Allow overlap when the number is the requested answer. - Because/therefore: Can the reader follow the path from evidence to consequence without an unsupported causal leap? - Headline-only storyline: Do the titles alone form a coherent argument? - Orphan section: Does every section have one clear rhetorical job and advance the case? For sectioned briefs, apply the same checks to section titles. For shorter formats, apply them to lead sentences. ## Polish the prose After the reasoning and factual checks are complete, revise audience-facing prose for naturalness, clarity, and fit with the speaker and setting. Remove generic phrasing, repetitive transitions, unnecessary headings, and other patterns that make the writing feel templated. Preserve every supported claim, number, date, name, citation, limitation, and required brand or formatting convention. Match the speaker's voice when a writing sample exists. Do not let style changes weaken precision or introduce facts. Return only the finished content required by the user's task. Keep the audience-facing output free of internal framework labels and review commentary unless the user asks for the audit.