The Claude Charts Playbook

Claude charts turns a pasted table into an interactive, filterable visual in a single prompt. But feed it the wrong input and you get a broken output - or worse, a chart that looks right but isn't.

This page tells you exactly what works, what breaks, and where the hard limits are.

Data input formats

What Claude reads cleanly

What causes garbage output

Chart types available

What Claude can build

What Claude can't build

The two input mistakes that break everything

Mistake 1 - Skipping the header row

Mistake 2 - Not specifying the chart type

Hard limit on data volume

Claude handles up to roughly 300-500 rows reliably in a single prompt. Above that, accuracy drops - miscalculated aggregations, missed patterns, incomplete visuals.

If your dataset is larger:

→ Pre-aggregate in Attio or Sheets before pasting (summarize by week, not by day)

→ Split into separate charts by segment rather than one combined chart

→ Use the "describe my data" prompt on Subpage 2 first - Claude will tell you what's feasible before you commit

Most bad Claude chart outputs aren't a Claude problem. They're a data formatting problem. The prompt can be perfect - if the data is messy, the chart is broken. This page covers how to format exports from Attio, HubSpot, and raw CSVs so Claude reads them cleanly every single time.

Exporting from Attio

What to export

The three columns that matter most for GTM charts

Exporting from HubSpot

Cleaning HubSpot exports

The "describe my data" diagnostic prompt

Run this before building any chart from a dataset you haven't used before. Claude will tell you what chart types are possible with your data, what decision each one helps you make, and what's missing that would strengthen the analysis.

The prompt

How to read the output

When to clean before pasting vs. letting Claude clean it

Clean it yourself when

Let Claude clean it when

This is the core of the playbook. Five specific moments in a GTM workflow where pulling data into Claude and building a chart is faster than any alternative - and more importantly, directly answers the decision sitting in front of you.

Each entry: the decision, the prompt, the dummy data to test with, and what to look for in the output.

Decision 1 - After a campaign sends: where did reply rates actually drop?

The moment

The prompt

What to look for in the output

Decision 2 - Before a QBR: where do deals stall by ICP segment?

The moment

The prompt

What to look for in the output

Decision 3 - When qualifying a new ICP: is signal strength actually predicting conversion?

The moment

The prompt

What to look for in the output

Decision 4 - When content performance drops: which post type is actually driving results?

The moment

The prompt

What to look for in the output

Decision 5 - When a client asks "is this working": lead source attribution by stage

The moment

The prompt

What to look for in the output

A screenshot of a Claude chat window is not a client deliverable. A standalone interactive HTML file hosted on a clean link is. This page covers exactly how to go from a chart Claude just built to something you can drop into a client Slack or email in under 3 minutes.

Step 1 - Run the "code it" follow-up prompt

The moment Claude builds the chart you need, run this prompt immediately in the same conversation:

The prompt

What you get

Step 2 - Host it on Google Drive

Upload and get a clean link

One thing to know

Step 3 - How to frame it in the delivery message

What not to say

What to say instead

Scenarios

Most weekly GTM reviews are 90 minutes of pulling data from four different places, building a Sheets tab nobody saves, and ending with "let's keep an eye on this." This page is the 15-minute version - a fixed prompt sequence that produces three actionable charts and a written synthesis, every week, from a single paste.

The structure

The review runs in a single Claude conversation. Four prompts. In order. Don't skip ahead.

→ Prompt 1: Diagnostic (what chart types are possible with this week's data)

→ Prompt 2: Performance chart (what changed this week)

→ Prompt 3: Trend chart (is this a blip or a pattern)

→ Prompt 4: Synthesis (what does it mean and what's the next move)

Prompt 1 - The diagnostic

Run this first, every time

Prompt 2 - The performance chart

What changed this week

Prompt 3 - The trend chart

Is this a blip or a pattern

Prompt 4 - The synthesis

The three questions every review answers

The three questions, explained

The full 15-minute timeline

→ Minutes 1-2: Export this week's data from Attio or your tracking sheet

→ Minutes 3-4: Run Prompt 1, read the diagnostic, fix any data gaps

→ Minutes 5-8: Run Prompt 2, review the performance chart

→ Minutes 9-11: Run Prompt 3, review the trend chart

→ Minutes 12-14: Run Prompt 4, edit the synthesis

→ Minute 15: Share the two charts + written summary

The whole thing lives in one Claude conversation. Save the conversation link if you want to reference it next week.

Every prompt from this playbook, organized by use case. Copy directly. Swap in your data. Run.

Diagnostic prompts

Describe my data

Clean my data for charting

GTM chart prompts

Reply rate by channel and sequence step

Pipeline stage velocity by ICP segment

ICP signal cluster scatter map

LinkedIn content engagement by post type and day

Lead source attribution by pipeline stage

Client delivery prompt

Convert chart to shareable HTML

**Weekly synthesis **