PROMPTS
Build 3 Codex Dashboards From Your Customer Data ⚡
Spreadsheets are the worst way to find patterns in your business. These three Codex prompts turn your order, buying-hour, and review data into premium.
I run an e-commerce brand and I make decisions from my customer data every single day. But spreadsheets are my worst enemy because I’m a visual learner. So I had Codex build me three live dashboards from my data. And now I can see patterns in 5 seconds that would take me an hour to find in a CSV.
This guide gives you the exact three prompts I used. I run Shopify and Klaviyo, but none of this is platform-specific. As long as you can export a CSV from your store, your email tool, and your reviews, these prompts work for you.
Each dashboard is one self-contained HTML file that opens directly in your browser. No servers, no hosting, no extra tools to install. You hand Codex a CSV, give it the prompt, and 5–10 minutes later you have something that looks like Apple built it.
What You’ll Need
Codex. OpenAI’s coding agent. You’ll need a paid plan with code execution. openai.com/codex.
A CSV of your orders. Shopify, WooCommerce, BigCommerce, Squarespace, Stripe, Amazon Seller Central. Any platform that lets you export orders with shipping city/state will work. You’ll use this for Dashboards 1 and 2.
A CSV of your reviews. Yotpo, Stamped, Loox, Judge.me, Google Reviews, Amazon reviews, your Shopify reviews app. Whatever you’re collecting customer feedback in. You’ll use this for Dashboard 3.
About 30 minutes. Roughly 10 minutes per dashboard once Codex starts building.
Privacy note
Codex runs your CSV locally during the build. Each dashboard is designed so the final HTML file doesn’t expose raw customer data. Just the visualizations. You can open it on your own computer, share it with a partner or team, and the underlying numbers stay private.
Where Your Customers Live
A premium dark-canvas U.S. Map with a dot for every city that’s ever ordered from you. The top 10 cities show up in red. You can pan, zoom, hover for city + state, and click any dot to highlight it. A “Top 10 Cities” list sits under the map. No raw order counts are visible anywhere. Just the pattern.
When Your Customers Buy
A dark premium dashboard that ranks all 24 hours of the day by buying strength. The top three hours get a gold, silver, and bronze podium. The full 24-hour heatmap sits below with rank numbers in each cell. Not raw order counts. Hover for signal strength + daypart. Click any cell for adjacent hours. The best buying hour appears in the headline so you always know when to send your next email or post.
How Your Customers Talk
A floating word map over a dark animated mesh background. The biggest word is the strongest descriptor customers use about your brand. Click any word and a glass panel slides in showing real review snippets where that word appears. Under the map: one clean section that says “Here’s your marketing angle” with a positioning line built from your top customer language. This is the dashboard that changes how you write ads.
A Few Things To Know
Refreshing later. You don’t need to rewrite anything. Drop a new CSV into the same Codex chat, ask it to rebuild the dashboard using the same spec, and it regenerates the file. The design stays the same. Only the data updates.
Customizing the look. The prompts give Codex a strong opinion, but you can override it. Want the city map to be blue instead of red? Ask. Want the word map on a cream background instead of black? Ask. Codex will rebuild without losing the structure.
Sharing the dashboards. Each one is a single HTML file. You can email it to a partner, drop it in Notion or Slack, or open it on a TV in your office. No login, no install, just open the file.
If something looks off
Codex sometimes nails the data but misses on aesthetics. The fastest fix is to take a screenshot of what you got, paste it back into the chat, and tell Codex what you want different. It’ll iterate in the same chat without starting over.
The real unlock of AI agents is not full autonomy, but human-in-the-loop systems design. Structure your agent inputs with explicit constraints, negative prompts, and automated test checkpoints. When building tools, keep token consumption lean by caching system prompts and isolating tasks into specialized sub-agents.