Data Evidence Report

Verify the numbers. Then explain the change.

Provide an order or channel export, reporting month and definitions. The Agent validates data, calculates metrics and builds charts with traceable sources. Refresh with next month’s file; missing columns, conflicting records and zero denominators stay explicit.

Currency-separated order metrics, or channel totals with weighted click and conversion rates.
A readable HTML report, standalone SVG charts and inspectable JSON / Markdown.
Metric definitions, exclusion and duplicate records, and a brief you can reuse for the next month.

Available now · v1.0.0

USD 9.99 · Single purchase, or get access through Pro.

What would you like to make?

Monthly orders

Review the month’s order amounts with refunds and duplicate records.

Currency-separated amounts, definitions, source records and exclusions.

Channel review

Compare channels without averaging daily percentages.

Rates from summed counts, with explicit zero denominators and limits.

Next-month refresh

Reuse the report definitions and detect changed columns next month.

A new report and brief, with previous results preserved and conclusions rechecked.

Refunds and two currencies in one export

Actual output · original fictional example

The input

An original fictional order export with refunds, duplicate records and an explicit reporting period.

The work

Check duplicate order identities, select creation dates in the report timezone, apply status rules and aggregate each currency separately.

Browse all output examples

Actual order report — Actual output from original fictional package data: validate records, then calculate per currency.
Actual channel report — Sum counts before rates; undefined ratios are not shown as fabricated zero.
Actual evidence and definitions — Inspect source record references, reporting boundaries, exclusions and duplicate handling.

From your material to the finished files

Define and validate
Establish the row grain, month boundary and included statuses. Missing values are not coerced to zero, and conflicting records are not silently chosen.
Calculate with evidence
Keep currencies separate and sum numerators and denominators before calculating rates. Preserve filters, duplicate handling and source records for review.
Refresh and review again
Keep the previous run, update the source and month, then regenerate. Handle field and definition changes explicitly; old conclusions do not establish the new month’s result.

Bring this task into your Agent

Prepare these materials

Provide CSV or a JSON object array, your question, reporting period and timezone. Orders need gross, refund, currency and status definitions; channels need comparable view, click and conversion counts. The Agent reads the columns and prepares the brief; you do not write formulas or JSON.

Sign in to acquire the package.
Extract the entire folder into your project: .agents/skills/data-evidence-report/SKILL.md (Codex) or .claude/skills/data-evidence-report/SKILL.md (Claude Code).
Open a new chat in that project, attach your materials and replace the bracketed fields in the brief below. Ask the Agent to check package instructions and dependencies first.

Use data-evidence-report and read SKILL.md first. Analyze [order CSV/JSON path] for [inclusive start] to [exclusive end] in [timezone]. Confirm row grain, included/excluded statuses and refund definitions, then prepare the column mapping and brief. Calculate gross, refunds and net separately by currency. Deliver HTML/SVG, source records and reusable configuration. Report missing or conflicting data without fabrication; do not present the order snapshot as cash flow or profit.

Installation & troubleshooting guide

The first result can be revised

Analyze next month’s file under the same definitions and preserve this run. Explain changed column mappings first. Compare amounts by currency and label rate differences in percentage points.

What the package includes

Currency-separated order metrics, or channel totals with weighted click and conversion rates.
A readable HTML report, standalone SVG charts and inspectable JSON / Markdown.
Metric definitions, exclusion and duplicate records, and a brief you can reuse for the next month.

What you provide separately

Python 3.11+ and system IANA timezone data; core computation uses the standard library with no analytics-service account.
Your own Codex/model access, plus a browser for viewing and PDF printing. Associated fees are separate.

Before you start

How is this different from a weekly deck?

It starts from raw exports, validates and calculates, then preserves evidence and refresh configuration. A weekly deck organizes existing findings. Outputs are HTML/SVG, not native PPTX.

Does the helper upload data?

The calculation helper reads local files without network access. Your Agent/model may process selected input under its own service settings; this package does not promise offline AI inference.

Can it analyze any spreadsheet?

This version supports order snapshots and aggregate channel CSV/JSON exports. Refund event ledgers, joins, attribution, profit and currency conversion are outside scope. Ambiguous definitions need clarification.

What about PDF and Excel?

Print the HTML to PDF in a browser and review pagination. Export Excel data as CSV first; workbook formulas are not read and native XLSX is not produced.

How do I purchase and use it?

Purchase individually or download with an active Pro membership. Extract the complete folder to .agents/skills/data-evidence-report/ in your project and ask Codex to read SKILL.md. Provide CSV/JSON, period, timezone and your business question. The Agent maps columns and prepares the brief, clarifying ambiguous definitions with you.

Not audited accounting, profit computation, investment advice or a general data platform. Currencies stay separate; unsupported causes are not inferred. Internal examples are not customer results.

More to make

Waitlist Landing Page — Create a responsive prelaunch page with a product story, FAQ and clearly labeled local demo form. Includes editable source and desktop/mobile previews.
SaaS Landing Page — Create a full SaaS marketing prototype with positioning, features, workflow, billing-period controls, FAQ and an implementation handoff.
Editorial SaaS Design — For writing and knowledge products. Upright editorial serif, generous paper margins, deep green actions, asymmetric manuscript composition. Give your Agent a brief to create a visual system and editable responsive page, or apply the direction to an existing project.
Industrial Catalog Design — For hardware, furniture and industrial products. Ruled catalogue grid, square geometry, product-first imagery and compact model annotations. Give your Agent a brief to create a visual system and editable responsive page, or apply the direction to an existing project.