Sales Analysis Dashboard — Instant Insights from Any CSV
· Apps
Drop a CSV, get 100+ time-period analyses, trend comparisons, and seasonal patterns instantly. Zero setup, no signups, 100% local processing.
Drop any CSV file and get instant sales analysis with 100+ time-period breakdowns, trend comparisons, and seasonal pattern detection. No Excel formulas, no setup — just upload and explore.
How It Works
- Drop your CSV file onto the dashboard
- The app automatically detects date and value columns
- View 100+ automated analyses instantly
- Filter, drill down, and explore patterns
- Export visualizations or data
Automated Analyses Include
- Time Periods — Daily, weekly, monthly, quarterly, yearly
- Year-over-Year — Compare any period to the same period last year
- Moving Averages — 7, 30, 90, 365 day trends
- Seasonal Analysis — Detect repeating patterns by day of week, month, quarter
- CAGR & Growth Rates — Compound annual growth calculations
- Percent of Total — Contribution analysis by any dimension
- Rolling Sums — Cumulative totals over any period
CSV Format Requirements
The app works best with CSVs containing:
- A date column (any common format: YYYY-MM-DD, DD/MM/YYYY, etc.)
- A value column (sales, revenue, units, etc.)
- Optional dimension columns (product, region, customer, etc.)
Privacy & Security
All processing happens in your browser. Your data is never uploaded to any server. This makes it ideal for sensitive sales data that can't go to cloud analytics tools.
Technology Stack
- React + TypeScript
- D3.js for visualizations
- Papa Parse for CSV parsing
- 100% client-side processing
The Problem It Replaces
Most sales analysis starts the same way: a CSV export from an ERP or POS system, a pivot table, a handful of helper columns for month and quarter, and then twenty minutes of dragging fields around before anyone learns anything. That work is identical every time, on every dataset, which makes it a good candidate for automation. This dashboard runs the whole standard battery of period analysis the moment a file lands, so the first thing you see is the answer rather than a blank pivot.
How Column Detection Works
On upload, each column is sampled and scored. A column is treated as a date if a high proportion of sampled values parse cleanly under common formats, as numeric if values coerce to finite numbers after stripping currency symbols and thousands separators, and as a dimension otherwise. Cardinality decides what to do with dimensions: a column with a handful of distinct values becomes a filter, while a high-cardinality column such as invoice number is ignored for grouping. If detection guesses wrong, every mapping is editable before analysis runs.
What the Metrics Actually Mean
Year-over-year comparisons align calendar periods, not row positions, so a month with missing days still compares against the full equivalent month last year rather than quietly under-reporting. Moving averages are trailing, not centred, which keeps the latest point meaningful for monitoring instead of shifting with future data. CAGR is calculated from the first and last complete periods only, since partial periods at either end inflate or deflate growth badly on short histories.
Seasonality is measured by comparing each period bucket to the overall mean and reporting the index, so a December index of 1.4 means December typically runs forty percent above the average month. This is deliberately simple: with two or three years of data, a full decomposition would overfit, and an index is easier to defend in a meeting.
Preparing Your CSV
Keep one row per transaction or per period, one header row, and avoid merged summary rows inside the data — subtotal rows are the single most common cause of doubled figures. Dates should be consistent within a column, negative values should represent refunds rather than being stored in a separate credit column, and blank value cells are treated as zero only when the surrounding rows suggest a continuous series.
Where It Falls Short
A sample file to test before using your own data
Use this small example to check date and value mapping. Its four rows total 1,200 in sales; each month totals 600. These are illustrative values, not a forecast.
| Date | Region | Product | Sales |
|---|---|---|---|
| 2026-01-05 | North | Service A | 400 |
| 2026-01-12 | South | Service A | 200 |
| 2026-02-05 | North | Service B | 500 |
| 2026-02-12 | South | Service B | 100 |
Export these columns to CSV with one header row. After import, confirm the date field, the sales field and the two dimensions before interpreting any chart. A correct total with an incorrect date mapping still produces misleading time analysis.
Reconcile the first upload in five checks
- Compare the imported row count with the source export, excluding the header.
- Compare total sales with the source, accounting for any filters you applied.
- Check the earliest and latest dates, plus a few ambiguous day/month values.
- Inspect negative values and returns; confirm they reduce sales once rather than appearing twice.
- Filter one known region or product and reconcile its subtotal independently.
Prefer ISO-style dates such as 2026-02-05. A value like 03/04/2026 is ambiguous without a locale convention. Check decimal separators too: 1,234.50 and 1.234,50 can be interpreted differently. Remove currency symbols and use one numeric convention across the column if detection is uncertain.
Choose a comparison that answers the question
Compare a complete month with a complete month, or month-to-date with the same elapsed portion of the comparison month. February revenue below January may reflect fewer trading days rather than lower daily demand. Where weekends or public holidays matter, a daily average can help, but it does not by itself control for the trading calendar.
A missing day is not always a zero-sales day: it may mean the export omitted records. Confirm completeness before relying on moving averages. Likewise, a percentage change from a zero prior value is undefined; explain it as a new activity or use the absolute change instead of presenting an enormous growth rate.
Two years can suggest a seasonal pattern, but one promotion or supply interruption can distort it. Inspect the raw history around unusual peaks before treating a month index as an expected future outcome. The dashboard describes the data you supply; it does not establish why sales changed.
When to move beyond a browser analysis
Use this approach for a quick, repeatable investigation of a clean export. If you need scheduled refresh, a shared source of truth, controlled access or reconciled multi-currency reporting, build a maintained reporting model instead. See the Excel dashboard examples for layout choices and the Excel to Power BI migration guide for the next step.
- Files above roughly 200,000 rows start to feel slow in browser memory
- No forecasting — it describes what happened, it does not predict
- Multi-currency data must be converted before upload
- No saved state between sessions by design, since nothing is stored server-side
Need a dashboard built on your own data?
The browser tool is a quick start. For a dashboard connected to your systems with refresh and agreed definitions, see the Excel dashboard service or send a brief.