Tablif Town/Provetale/Problem page
Problem page

How to find which segments explain a revenue change in a spreadsheet

Your monthly revenue dropped and someone asks the obvious question: where did it go? You open the spreadsheet, see the total fall, and start building pivot tables by customer type, region and product. An hour later you have ten tabs and still no clear answer you would put in front of your manager. You need one sentence that says which part of the business moved, with numbers you can defend.

2 min readWritten from provetale.comWelcome page from Tablif · 23 Oct 2026
The short answer

Provetale turns a spreadsheet into a written report. Ordinary code calculates the figures, compares each segment between two periods and works out how much of the change each one accounts for.

On this page
  1. Why the usual way fails
  2. How Provetale helps
  3. Step by step
  4. Tips
  5. Questions

01Why the usual way fails

A total can fall because one group collapsed or because every group slipped a little, and the headline number looks the same in both cases. Splitting the data by hand means choosing which columns to compare, and it is easy to miss the one that matters. Pasting the file into a chat assistant is fast, but language models are not built to do arithmetic on raw rows, so a confident sentence can carry a wrong figure. When the number is wrong, nobody can trace where it came from.

02How Provetale helps

Provetale turns a spreadsheet into a written report. Ordinary code calculates the figures, compares each segment between two periods and works out how much of the change each one accounts for. The AI model only writes the text from those calculated findings, and every number, date and group name in that text is checked against the stored evidence before the report is published. Each statement opens the calculation, the columns it used and the data it came from, so you can show your work.

“The analysis runs on our own code. The AI model is never asked to calculate anything.”
From provetale.com
“Follow a finding to its source columns, comparison and calculation.”
From provetale.com

03Step by step

  1. Upload your spreadsheet and let Quick Analysis calculate metrics, trends and anomalies.
  2. Open the finding about the change and read which segments account for most of it.
  3. Click the evidence to see the source columns, the comparison and the calculation.
  4. Use Explore to ask a focused question about that finding.
  5. Pick the report structure your reader needs, from Executive to Technical, and share it.

04Tips

  • Separate what happened from why it happened. A contribution number tells you which segment moved, not the reason, so treat any cause as something to check with the team that owns that segment.
  • Compare the same kind of data across both periods. If one month has missing records for a segment, the drop may be a data gap, so confirm the source coverage before you act.
  • When a new version of the file arrives next month, compare the versions instead of starting again. Provetale's Monitor view shows the meaningful changes and their evidence.
  • Write the answer as one short sentence with two numbers: the size of the change and the share explained by the top segments. Readers remember that far better than a table.

05Questions

Does the AI see my raw rows?

No. The model only gets findings that code computed, such as figures, column names and group names.

Source: provetale.com
What happens if a sentence fails the check?

It is rewritten up to twice, and if it still fails it is left out of the report.

Source: provetale.com
Which languages can the report be written in?

English, Bengali, Spanish, German, French, Portuguese, Japanese and Arabic, written directly rather than translated.

Source: provetale.com

Provetale turns spreadsheets into written reports where every finding is backed by traceable evidence for teams and analysts.