# Findings to Slides > Paste a finished analysis - findings in prose or bullets and the tables behind them (Markdown, CSV or tab-separated) - choose the audience and say what you want decided. The browser numbers every data point, recomputes each table's period-over-period change and checks any stated change against it, re-totals weighted scoring matrices, and flags a data dump, methodology first, causal wording, a small sample, jargon for the audience, a missing ask and a topic-label headline, for free. Then a story lane writes the headline, three points, the six-beat arc, cuts, caveats and the ask, and a deck lane writes the seven-slide storyboard. URL: https://findings-to-slides.skillsafe.ai/ API tutorial: https://findings-to-slides.skillsafe.ai/api.html ## Who it is for Analysts, founders, product and growth leads, and anyone who has finished an analysis and must now present it to executives, a board, investors, a cross-functional team or a non-technical audience - and wants the "so what" first, three points the data supports, and a deck that ends with a decision. ## The free reading (in the browser, no model) - The analysis: prose and bullets (every sentence with a figure becomes a data point, D1..), and tables with a header row in Markdown, CSV or TSV. - Period tables (columns such as Q3 and Q4, before and after, baseline and current): every row's change from the first to the last period, in points and relative terms for rates, and any stated change column checked against it. - Weighted scoring matrices (a Weight column and option columns, with a Total row): every option's weighted total recomputed and the weights' sum checked. - Findings (DS-01..): a stated change or total the table's own columns contradict, or weights that do not sum to 100% (high); more than 15 data points (a data dump); the paste opening with the method; causal wording (caused, drove, due to, led to); a small sample (n under 100); statistical or technical jargon for the chosen audience; no ask; a draft headline with no number or that is a topic label (medium). - Exports: the reading as Markdown, every data point as CSV. ## The paid lanes (gpt-terra, metered) - story: one answer per page finding, in order (arithmetic findings confirmed); the framework - problem-solution, trend or comparison; the headline by the formula (a specific number, the business impact, actionable context) and two more; exactly three points, each citing one to three data points, with the so-what; the six beats - hook, context, rising action, climax, resolution, call to action; the data points to cut; up to three caveats; the ask. - deck: seven slides in the skill's data story flow - the headline, the context, the discovery, the deep dive, the recommendation, the impact, the ask - each with an action title, one to three bullets, the data it shows, the visual and what it highlights, and a speaker note; plus an appendix. After a story (the button under it), the storyboard follows that story's headline, points and ask. ## How replies are held to the input Every reply is reconciled in the browser against the page's reading: a number in the headline; every data and finding id one of the page's; no figure that the analysis or the page's reading does not contain (rounding allowed); no figure the page found mis-stated in the story or on a slide; for a story, the findings answered in order, three points with data, the six beats in order, cuts never used in a point, a caveat after causal wording or a small sample and no causal wording in the headline or the points, and an ask; for a storyboard, seven slides in order with action titles, slide 1 the headline, one to three bullets each, data on the discovery and deep dive slides, and with a story handed over its headline kept and every point's data shown. ## Limits - The paid lanes see up to 14,000 characters of the analysis (the middle cut on whole lines beyond that) and up to 60 data points. - The page checks arithmetic and wording, not whether the analysis is right; a story can only be as sound as the work under it. - Period columns are recognised by their headers (quarters, months, years, before/after, baseline/current, plan/actual); other numeric columns are read as figures by row. ## Source Derived from the agent skill @wshobson/data-storytelling (https://skillsafe.ai/skill/@wshobson/data-storytelling), from the repository wshobson/agents (MIT): story structure and the narrative arc, the three pillars, the problem-solution, trend and comparison frameworks, the headline formula, the do's and don'ts, handling uncertainty, visualization techniques and the data story flow. ## Pages - https://findings-to-slides.skillsafe.ai/ - the app - https://findings-to-slides.skillsafe.ai/api.html - API tutorial (cURL, Python, JavaScript, Go, Java, Ruby, PHP, C#)