Turn One Messy Dataset Into a Strategy Deck People Will Actually Read with Claude Design

Do not throw every file you have into Claude Design. That makes the project slower, burns through usage faster, and usually creates a bloated deck.

Start with one clean spreadsheet or CSV. This workflow can work with YouTube channel data, Facebook ad data, Shopify data, sales exports, customer research, or any messy report you need to explain.

If you have supporting assets, include only the ones that help explain the data. For a YouTube analysis, that might mean thumbnails, titles, and descriptions tied to each video.

Open claude.ai/design, switch to Slide deck, and keep the setup simple.

For this workflow, skip the design system unless your team already has one uploaded. Turn on speaker notes so Claude can keep the slides lighter instead of cramming every detail onto the page.

After the project opens, upload your spreadsheet and any supporting assets.

If you have images or creative files, make sure the filenames or spreadsheet fields make the relationship clear. Claude needs some way to match each asset back to the right row.

The prompt is the important part. Do not ask Claude to simply analyze the data. Push it toward strategy analysis, clear decisions, and useful takeaways.

Turn these files into a strategy deck on performance. Analyze the results by item and extract best practices from the data and supporting assets. Use charts, rankings, and concrete recommendations. Match any images or creative files to the CSV using the filename or matching field. Keep it crisp and presentation-ready.

If you know the categories you want Claude to study, add them. For the YouTube test, the prompt asked it to look at thumbnails, titles, descriptions, and subject matter. Swap those categories for whatever matters in your dataset.

Claude Design can take 10-15 minutes to build the deck. The first time you run this workflow, watch how it thinks through the data instead of sending the prompt and leaving.

Check whether the deck is doing real work. A useful deck should not just repeat numbers back to you. It should show what is working, explain why it matters, and turn patterns into next steps.

Here is an updated CSV with the correct date range. Everything else should be the same. Rebuild the deck using this file instead.

Do not restart the whole project if the first pass is close. Fix the specific issue.

If the deck focuses on the wrong metrics, ask Claude to rebuild around the metrics that best explain performance and de-emphasize vanity metrics.

Rebuild this deck around the metrics that best explain performance. De-emphasize vanity metrics and explain why the chosen KPIs matter.

If the slides are too busy, ask for cleaner hierarchy, fewer words per slide, and stronger contrast. If the charts are weak, ask for comparisons between top and bottom performers, pattern clusters, and visualizations that connect choices to performance.

Tighten the visual style. Make this feel like a polished strategy presentation, not a generic AI deck. Use cleaner hierarchy, fewer words per slide, and stronger contrast.

Add more useful charts. Compare top and bottom performers, show pattern clusters, and visualize which choices correlate with better performance.

In the test run, Claude Design produced a strong deck because the starting prompt asked for best practices and recommendations. It pulled in thumbnails, compared winners and underperformers, and created slides closer to do-this-not-that than a basic analytics recap.