ds@mcpanalytics.aiEmail your data and your question and get an answer about your own data, with the method shown.Email your data and question. Get an answer.
Modify an analysis: how it works | MCP Analytics
Modify

Change the question without starting over.

Analyses are rarely right first time: you see the answer and realise you asked slightly the wrong thing. Modifying builds a new version from the same question, data and column mapping, with your change added, and leaves the original as it was.

Modify an analysis Commission a new one instead →

Opens your analyses. Pick one, describe the change, and nothing is built until you submit it.

  1. Start from what existsAn analysis you already have at Brief or Deck depth, with its question, its data and its report. A Snapshot cannot be modified; create a new one instead.
  2. Say what should be differentAdd a variable, split by a new group, change the outcome, tighten a threshold. In plain language, as before.
  3. A new build from the same starting pointYour change is added to the original question with the instruction to keep everything else the same, and the build runs again on the same copy of the data, with the same column mapping, at the same depth.
  4. The link is recordedThe new version stores which analysis it came from and the change you asked for, so you can tell which answer came from which question.

What stays the same, and what can change

Kept: the question with your change added, the data, the column mapping and the depth. Rebuilt: everything else, including the choice of method, because the build runs again. Before comparing numbers between the two reports, read the method section of each.

Worked example

What a modification sends to the build

Illustration. The parent is a real published study; the change is an example, and no modified version of this study is published.

The parent. IBM HR Employee Attrition Drivers, a commissioned study built from the question below on the public IBM HR attrition dataset. Its summary table shows 30.5% of employees who worked overtime left, against 10.4% of those who did not, across all departments together.

The change. You read it and want to know whether that holds at every pay level. You type the change in plain words, and the build receives the original question with your change appended, in this form:

Which employee attributes actually predict attrition?

Apply these changes (keep everything else the same):
- Split the overtime finding by monthly income band.

What it says. The question, data, column mapping and depth come from the parent, so the new version answers the same question with one thing changed, on the same rows.

What it does not say. The build runs again, so it can choose a different method or different checks than the parent did. Compare the method sections of the two reports before comparing their numbers.

By MCP Analytics Team. Updated November 2026.

The other things you can do

CreateCommission an analysis for a question nobody has asked…RunRun an analysis again, on data that has changed. BrowseStart from a prebuilt analysis, already built and tested.