Data Report Analysis

You receive a business data Excel file but don’t know how to analyze it, chart it, or report on it. This case shows how to use LanMate to turn messy data into an analysis report with conclusions, charts, and ready-to-present material.

Skills Used

Built-in skills (installed with LanMate, ready out of the box): Excel spreadsheets (lanmate-xlsx), data charts (lanmate-data-viz), HTML reports (lanmate-html-report)
Skill Store: Data Visualization Analysis — choose charts based on the business question and draw conclusions

Pain Points

Chaotic data definitions: inconsistent date formats, merged cells, null values and outliers mixed together
Not knowing what to analyze: you have data but no business questions
Charts without conclusions: lots of charts made, but no idea what they show
Presentation material is hard to produce: after analysis, you still need to reformat it into a report

Recommended Workflow

Stage	Prompt Focus	Output
Read the sheet	Describe the workbook structure, field meanings, and dirty data first	Data dictionary, issue list
Define metrics	State the business questions to answer (not “analyze this”)	Metric definitions table
Clean	Explain how to handle nulls, duplicates, and outliers	Cleaned xlsx
Visualize	Choose chart types based on the business questions	Chart PNGs
Report	Generate a self-contained HTML analysis report	Single-file HTML report

Prompt Examples

Sales Data Analysis

Please read the "e-commerce-sales-data.xlsx" file I uploaded — do not modify the original file.
Business question: Analyze this month's sales performance and profitability by product line,
determine which product lines contribute the most and which have weaker profit performance,
and identify abnormal fluctuations in this month's sales.

Please output:
1. Explain the meaning of data fields; check for missing values, duplicate records, outliers, and field format issues
2. By product line, calculate sales, gross profit, gross margin, sales share, and gross profit contribution share, and rank them
3. Summarize daily sales and gross margin, and analyze the daily trend of this month's sales performance
4. Generate bar charts comparing sales and gross profit by product line, and a line chart showing daily sales changes
5. Identify dates with clearly abnormal sales or gross margin, and explain based on the data — do not speculate on causes without evidence
6. Summarize this month's best-performing product line, product lines that need attention, and 3 conclusions usable for business review

Output a self-contained HTML analysis report with charts embedded.
Requirements: preserve the original data; the statistical process must be traceable; chart titles should directly express the main conclusion;
mark causes that cannot be confirmed from the data as "unverified" — do not fabricate.

Multi-Table Merge and Reconciliation

Merge the weekly sales tables for 6 regions in the input directory.
First check column names, data types, date ranges, currencies, and primary keys — stop and list differences if inconsistent.
Deduplicate by order number, but keep records of duplicate sources.
Before summarizing, output total row count, nulls, outliers, and duplicate count.
Generate clean-sales.xlsx, exception-list.xlsx, and reconciliation.md.
The amount total must reconcile with the sum of each source table — do not generate management conclusions when the difference is not 0.

Acceptance Criteria

Numbers can be traced back to the source file; formulas can be recalculated
Outliers are not silently deleted; the cleaning process is documented
Chart titles directly express business conclusions, not generic titles like “Sales Chart”
Causes that cannot be confirmed from the data are marked “unverified”
The HTML report is self-contained (no external dependencies) and can be sent directly to others for viewing

Common Mistakes

Common Mistake	Why It Happens	Better Practice
“Analyze this Excel”	No business question, so the model can only give a generic summary	State what questions to answer, which metrics to calculate, and which dimensions to compare
Data definitions not specified	Multiple definitions mixed together, making results unreliable	Specify a single data source and statistical definition
Charts without conclusions	Making one of every chart	Choose charts based on business questions; every chart needs a conclusive title
Original data not preserved	Modifying the original file directly	State explicitly “do not modify the original file”; save cleaned data separately
Letting the AI speculate on business causes	The model tends to give explanations	State explicitly “mark causes lacking evidence as unverified — do not fabricate”