Module 5 · Build #3 — The Know-Your-Numbers Dashboard
The kickoff: from files to dashboard
Now Claude reads your files, tells you what they can honestly answer, and then builds the page. We do it in two steps on purpose — understand your data before you chart it.
Open the folder in Claude Code
- Open your
numbers-dashboardfolder in Claude Code, the same way you opened your project folder in the earlier builds. - Make sure permissions are set so Claude can work without stopping to ask on every step (“Auto accept edits” in the Desktop app). If you set up sample data, run that prompt from the last lesson now.
Step 1 — read first, don’t build yet
Paste this first. It tells Claude to look, not touch — so you can catch a wrong or misread file before it becomes a wrong number on your page.
Step 1 — read first (don't build anything yet)
Look at the files in my data/ folder. For each one, tell me in plain English: what it contains, what date range it covers, and what business questions it could answer. Don't build anything yet.
Read what Claude comes back with. If it says a file covers the wrong dates, or it grabbed the wrong export, this is the moment to fix it — drop the right file into data/and ask it to look again. When the summary matches what you expect, move on.
Step 2 — build the dashboard
Now the build. Notice the prompt asks Claude to explain its plan in plain English before writing anything — and to refuse to invent numbers it can’t back up.
Step 2 — build the one-page dashboard
Now build my dashboard — a single HTML page I open in my browser, generated from the files in data/. Requirements: 1. Show: money in this month vs last month, my busiest services or products, new vs returning clients if the data supports it, and a simple chart of the last 6 months. 2. Create a refresh script: when I drop new exports into data/ and run it, the page rebuilds with the latest numbers. 3. Keep every number honest. If the data can't answer something, show "not enough data" — never estimate or invent. Add a small footnote saying which file each number came from. 4. Clean and readable — big numbers, clear labels, no clutter. Use only what's really in my files. Before you write any code, tell me your plan in plain English. Keep it simple — this runs on my computer, just for me.
Claude will lay out its plan, then build. When it’s done it’ll point you to an HTML file — open it and there’s your page. Big numbers, a trend chart, a footnote under each number telling you which file it came from.
Before you believe a single number on that page, check two or three of them by hand against the system they came from. Open last month’s revenue in your accounting tool and confirm it matches. Count this week’s bookings and confirm the page agrees.
AI can misread a column or double-count a row and state the wrong total with total confidence. Ten minutes of spot-checking is the difference between a page you trust and a page that quietly lies to you. If one number is off, tell Claude which one — it’ll find out why and fix how it’s reading the file.
Found a number that’s wrong? Don’t try to fix it yourself — just tell Claude what it should be and where you got it:
When a number is wrong
This number is wrong — it should be [X — for example: 18,400] according to [SYSTEM — for example: "my QuickBooks revenue report for last month"]. Find out why the dashboard is reading it differently and fix it.
Your dashboard is open in your browser, and you’ve spot-checked two or three numbers against the real systems and they hold up. You’re looking at an honest page of your own business — probably for the first time without waiting on anyone.
More learning resources
Still want another angle? These are hand-picked and vetted to line up with this lesson — from creators and sites worth your time.
