Best way to achieve this Data Analysis use case?

API
,
1

Goal: Have OpenAI Analyse the sheet where I store all my business data (Email Subscribers List, Event Registrants, Product Purchases etc) which are divided by sheet with following columns (Name, Email, Date of Purchase/Registration), Source etc).

I wanted to use OpenAI to check this sheet on a periodic basis and give me analysis, it;s finding, trends, comparison compared to last run (eg: if the period I set it 1 week, then on the 1st week it finds one trend, next week it shows that compared to last week, we saw xyz increase)

Question:
My question is what;s the best way to go by doing this?

I am confused on using CustomGPT, or OpenAI API’s (Chat completion or Assistant) etc.

If anyone has any recommendations, feel free to let me know.

Once I figure this out, I’ll share the automation template & guide below so anyone can do this after me as I think it could be a great use-case

1 Like
2

Automation? Personal information? That means API, not ChatGPT.

“Sheet”? Like a spreadsheet program? That means automated extraction from that into a form of natural language the AI can understand and where it can keep track of the data type (csv is not great, because the header key with columns becomes far separated from the rows of data with growing length.

Then you must consider the actual quantity of information. The AI model cannot take every invoice ever, and every customer record, and give a high-quality attention to a large amount of data, even if it can fit into the long context length of new models that allows a growing chat history. You’ll need to make sub-tasks.

Or simply talk to the AI: let the AI help you come up with a data extraction solution in code to a database format like pandas or SQLite, where your code can do the numeric processing on queries. And then guide you to a better customer record solution than some Google doc.

3

Thanks for the suggestion.

It’s stored in Google Sheet. The quantity is very big, like Email Subscribers sheet has like 180k+ rows.

What would make sense in that case? Don’t think I can configure it to read all that.

Only thing I can think of is using CustomGPT as that can analyze a bunch of data at once but don’t think we can call onto a CustomGPT via API’s right

4

You overestimate what ChatGPT can do.

If you have a file download into CSV or a simple spreadsheet file, that can be placed into code interpreter, and the Assistants AI can write Python scripts to process data. It can only receive back 16k characters of Python output, though.

5

Sorry for the delayed response.

Thanks for the clarification.

Just wanted to know what all possibilities are there.

So my only option for such a use case is to feed data up to 16k characters of data via CODE Interpreter.

Hmm, that’s a bummer, thought I could do so much more.

Doesn’t CustomGPT’s have the ability to feed itself data without any limits? Anything I can do in that regard that you can think of