WORK THROUGH A REAL TASK

How to Turn a CSV into a Chart in InriGPT

Reproduce a known result from a downloadable CSV, check its arithmetic, and understand the chart’s limits.

Prepared for InriGPT by Inri Inc · Updated 27 October 2026 · Fictional practice material

A chart is useful only when you can explain the rows and calculation behind it. This guide uses a four-row practice CSV with known totals. You can reproduce the result locally in the InriGPT chat, then spot the common data problems that would make a visually attractive chart misleading.

Download a small dataset with a known answer

Month,Sales_USD
January,120
February,180
March,90
April,210

The expected total is 120 + 180 + 90 + 210 = 600. The arithmetic mean is 600 ÷ 4 = 150. The minimum is 90 in March, and the maximum is 210 in April. These are practice amounts, not Inri Inc revenue or a customer result.

MetricExpected resultHow to check
Valid data rows4Exclude the header; count the four numeric records.
Sum600Add the four Sales_USD values.
Average150Divide the sum by four.
Minimum / maximum90 / 210Compare all four values rather than the chart order.

Create the chart in the public workspace

  1. Open CSV chart and attach the sample file.
  2. Choose Month as the label column and Sales_USD as the value column.
  3. Enter a short request such as “Chart these monthly sales” and press Send.
  4. Compare the displayed statistics with the known answers above before sharing anything.

The CSV chart tool calculates locally in your browser and does not consume an AI generation request. Selecting the separate data-analysis AI tool changes that boundary: a bounded text excerpt can be sent to the model. Local chart calculation does not mean every feature in the chat stays on your device.

Check what the current implementation plots

The local chart accepts up to 1,000 data rows and displays at most 30 bars. Statistics cover the valid numeric records included by the parser, while the visual can show fewer bars. A 30-bar picture is therefore not necessarily a picture of every row used in the total. Check the displayed row counts and retain your original file.

Negative values are labeled and use a distinct color. Bar length represents magnitude, so read the sign rather than interpreting every long bar as positive growth. This simple bar chart does not automatically perform seasonal adjustment, infer missing dates, deduplicate records, or calculate a forecast. It cannot tell you why a number changed.

Clean the data before interpreting it

Keep one observation per row and one unit per numeric column. Put a currency or unit in the header instead of mixing “€120,” “120 USD,” and plain numbers in the same column. Decimal commas, thousands separators, accounting parentheses, and spreadsheet formulas can be interpreted differently by parsers. Export clean decimal numbers and verify the result rather than relying on formatting.

A blank is not necessarily a zero. If February is unknown, replacing it with zero changes both the total and the average. Decide what a missing observation means in your business process before charting it. Likewise, two rows both labeled March may represent a duplicate, two branches, or separate transactions; the label alone does not establish which. Reconcile duplicates at the source.

Separate calculation from an explanation

The example establishes that April has the largest value. It does not establish that a campaign caused the increase, that April was profitable, or that May will be higher. Those conclusions require more evidence, such as costs, timing, and a comparison group. If you ask an AI assistant to interpret the chart, provide that context and ask it to distinguish observations from hypotheses.

For a short management note, report the period, unit, total, number of records, largest and smallest observation, and any excluded or missing rows. Keep the CSV and your correction notes together. That gives a colleague a way to reproduce the answer instead of trusting the shape of a graphic.

Keep working with the source

Browse the other guides and practice files. Read our editorial approach or report an error. Examples demonstrate a method; they are not customer testimonials or measured business outcomes.