Upload Your Spreadsheets and Let Claude.ai Do the Rest
Move over ChatGPT! Claude.ai is allowing users to upload spreadsheets and pdf’s for quick analysis. And all of this without running into the constant token limits found with ChatGPT.
Claude.ai, which is free to users, allows more robust data set uploads and is able to keep it all straight in its AI brain. No uploading your data piecemeal hoping connections will be made despite the token limits. What you need is a large language model able to all of your data and analyze it as one coherent piece.
Sign up for Claude.ai, from Anthropic, and experiment for yourself. Below are some of the best prompts for extracting the most effective conclusions from your data.
- Summarize this dataset in a few sentences: [insert dataset]. Provide the model with the actual dataset or a description of it. This encourages the model to digest the main points and output a concise summary.
- What are the main insights from this data? This elicits high-level observations about trends, relationships and conclusions without requiring much analysis from the model.
- What trends do you notice in this data? This prompt works well if the data has a time element or progression. The model can pick out upward, downward or stable trends.
- What are the most interesting relationships in this data? This focuses the model on correlations and associations between variables in a broader dataset. It may pick out unexpected relationships.
- Compare variable X and variable Y – are they correlated? What’s the relationship? Substitute your variables of interest here. This prompts a deeper look at the interaction between two specific variables.
- What conclusions can be drawn from this data? This asks the model to make analytic leaps from the data to higher-level conclusions. It may make interesting connections but the logic may need verification.
- What are the limitations of this dataset? What additional data would help draw more robust conclusions? This asks the model to make analytic leaps from the data to higher-level conclusions. It may make interesting connections but the logic may need verification.
- Generate a graph showing the relationship between variable A and variable B. Providing the model with specific variables prompts it to visualize the relationship in graph form. Useful for quick insights.
- Cluster the data into 3-5 groups based on similarities. Describe each group. This reduces complex data into distinct categories that can reveal subgroups and patterns. Useful for segmentation.
- Outline the key takeaways from this analysis in bullet points. Asking the model to synthesize bullet points forces succinct, logical conclusions tailored for communication.
- Write a short blog post/article explaining the key insights from this data.
- Suggest additional analyses that could provide more insight into this data. This tests whether the model can think critically about what further analyses might yield more insights.
The key is to ask clear, directed questions and prompt the model to summarize, interpret, and visualize the patterns in the data. You can iterate on the prompts to go deeper into the analysis. Adding some basic context about the goals and data can also help the model provide more relevant insights.
By Ariel Penn, October 17, 2023