Ask MapleA research agent that works inside the analysis — not in another tab.

01 · What it is

An assistant with access to the data, the engine and the screen.

02 · What you are looking at

The Ask Maple panel, part by part.

Table Studio on the left, Ask Maple on the right. The conversation is about the table.

1

The view on the left

Whatever you were working on — here, Table Studio showing purchase intent by shopper type, shaded for significance. Maple can see this view: its question, banner, filters and the numbers in it.

2

Panel header

‘Ask Maple’ and a subtitle naming the context — ‘Purchase intent × Shopper type’. It tells you what Maple is looking at, so you know what ‘this’ means when you ask about it.

3

Your question

Typed in plain English, in the bubble at the top: ‘Why are health-led shoppers so different?’ No column names or syntax needed; the context comes from the view.

4

The answer

Maple’s reply cites the cell on screen — 44.1% vs 26.9%, significantly higher than every other type, adjusted p = 0.008 — and names two things that could explain it. The numbers come from the table, not from the model’s memory.

5

Suggested actions

Two cards, each a concrete analysis Maple proposes to run next, with a one-line reason and a Run button. Pressing Run performs it in the product; nothing happens until you press.

6

Tool steps (when working)

When Maple is doing something — coding a column, fitting a model — each step appears in the panel as it runs and is marked Done. Later screens show this. There is no hidden work.

7

The prompt box

At the bottom: ‘Ask about what’s on screen…’. The Standard / Pro toggle chooses which model answers; Pro uses a more capable model for harder questions.

8

Copy

The small copy icon under an answer puts the text on the clipboard — for a note, a slide or an email — with the numbers exactly as the engine produced them.

03 · One investigation

Why are health-led shoppers different?

1. 01

   Explain — from the page

   “Why are health-led shoppers different?”

   What Maple does: reads the table on screen and answers from it — 44.1% vs 26.9%, significantly higher than every other type (adjusted p = 0.008 for this cell). Then it suggests two next steps, each a button: a chart of intent by flavour filtered to health-led shoppers, and a key drivers run on purchase intent.

   

2. 02

   Chart — opened beside you

   Press Run on the first suggestion.

   What you are seeing: the left pane has changed. The table has been replaced by the suggested chart — purchase intent by flavour preference, with a filter chip reading Shopper type is Health-led and the base, n = 372. Yuzu & ginger leads. This is a real AddMaple chart: you can change the filter, restyle it, save it to the story.

   

3. 03

   Code — what they wrote

   “What do they say about taste?”

   What Maple does: runs Analyze open-ended text on the 1,204 free-text taste comments, tagging each with the themes it expresses, and opens the result in the table view.

   What you are seeing: each comment beside the theme chips applied to it, with the words that earned each theme highlighted; a side panel of how often “Too sweet” appears in each shopper type; and Maple’s reading — 21% overall, but only 9% among health-led shoppers, significantly lower.

   

4. 04

   Build — a new variable

   “Create a ‘Likely to buy’ column: Definitely would buy + Probably would buy, everyone else ‘Not likely’.”

   What you are seeing: two tool steps — Preview segmented column, then Create segmented column — and the answer: 672 respondents (55.8%) are likely, 532 are not. On the left, a new card, Likely to buy, has appeared at the top of the summary dashboard. It is now a column in the study, ready to be a banner or an outcome.

   

5. 05

   Segment — with diagnostics

   “Is ‘health-led’ really one group?”

   What Maple does: runs a segmentation — fits candidates for k = 2 through 6 — and opens the Model comparison and diagnostics table. It reports three groups in the best-fitting run (Flavour explorers, Price-first sceptics, Habit buyers), notes that a second run on ratings only prefers four with less certainty, and saves the best fit for the researcher to review against shopper type. The diagnostics are the real ones; how to read them is on the Audience groups page.

   

6. 06

   Draft — then you publish

   “Save this as a finding for the client.”

   What you are seeing: a story dashboard — AddMaple’s shareable report page — with a finding block Maple has drafted: a headline, a paragraph citing the numbers, three stat tiles (55.8%, 44.1%, 3 audience groups) and the filtered flavour chart. Maple’s message ends: “It’s a draft until you publish it — review the blocks, then Publish to create the share link.” The Publish button is the researcher’s.

   

04 · The boundaries

What Maple does, and what stays with you.

• Maple proposes

  Suggested charts, cleaning batches, segment solutions, story drafts — all arrive as proposals you can see before they take effect.
• Every step is visible

  Each tool call is shown as it runs: column detail, create segmented column, cluster run candidate, save finding. There is no hidden work.
• Every change is reversible

  Variables, labels, groupings and story blocks are recorded changes on top of untouched source data. Undo any of them.
• You publish

  Maple can draft a finding into a story. Only the researcher publishes it — and only after reading it.

05 · Words used on this page

A short glossary.

Agent

An AI assistant that can take actions in the product — run analyses, create columns, draft blocks — not only answer in text.

Page context

What Maple can see: the project’s columns, labels and filters, and the view currently on screen.

Tool step

One action Maple takes — ‘Analyze open-ended text’, ‘Create segmented column’ — shown in the panel as it runs.

Suggested action

An analysis Maple proposes, with a Run button. Nothing runs until you press it.

Segmented column

A new column built by grouping answers — Likely to buy = Definitely + Probably. The same thing as a NET.

Story dashboard

AddMaple’s shareable report: findings, stat tiles and charts the reader can filter. Maple can draft into it; the researcher publishes.

Standard / Pro

Which model answers. Pro switches to a more capable model for harder questions.

MCP

The open protocol through which external agents (Claude, Cursor and others) can use the same capabilities on a project you give them access to.

06 · Questions

What researchers ask us about Maple.

What can Maple see?

The project you are in: its columns, labels, filters and the view on screen. Ask about the table in front of you and it answers from that table. It does not see other projects or anything outside the study.

Where do the numbers in its answers come from?

From the same engine that draws the tables and charts, run with the same filters and base. Maple does not estimate numbers; it reads them from the computation it just ran, and the result opens beside the conversation so you can check it.

Can it change my data?

It can create variables, recode labels, group columns and draft story blocks — as recorded, reversible changes over untouched source data. It cannot publish a story, delete source data, or act outside the project.

Do I have to use it?

No. Everything Maple does is also available by hand — Table Studio, Key drivers, Audience groups, the story editor. Maple is the fast path to the same tools and the same results.

Can I use it from Claude, Cursor or my own tools?

The same capabilities are exposed through AddMaple’s MCP server, so an external agent can run the same analyses on a project you have given it access to — with the same boundaries.