Can I upload CrossTabs to ChatGPT?

Can I upload crosstabs to ChatGPT?Ray Poynter, 21 March 2025

In two recent training sessions, I ran, I was asked the same question twice, “Can I upload crosstabs to ChatGPT for it to analyse and help me find messages in the data?”

I do not tend to do this as I typically upload raw data rather than market research crosstabs. However, I was able to share some general thoughts, such as starting by asking ChatGPT to tell you things you already know so you can cross-check that it has interpreted the data correctly.

Today, I spent some time playing with uploading tables to ChatGPT 4.5 and here are my key recommendations.

  1. PDF versus Excel. I found it much easier to upload and get ChatGPT to interpret the tables in PDF format. In Excel, the file used a lot of empty cells and merged cells, and the underlying Python seemed to struggle to make sense of it. To make that work, I needed to tweak the Excel file AND do more steps in the import.
  2. Check the totals, rows, and columns. The most frequent problem I encounter when uploading data is that the AI does not read all of the rows and columns. I tackle this problem by asking the AI to summarise the rows and columns, something I can easily cross-check and which ensures that the AI has encompassed all of the information.
  3. Give the AI some easy tasks. Ask it how many tables there are, how many people are in the data, and what the columns are — all things that are easy to check.
  4. Give it some harder tasks. For example, ask it to combine some data and filter by things like ‘Only using groups with a base of 30 or more.’ Then, create a net satisfied score and a table showing the net satisfied score, sorting from the highest to the lowest.
  5. Ask for the summary you want, making your question clear. In general, don’t ask ChatGPT to simply analyse or summarise the data. Tell it what the underlying question is, and be specific about what you want to understand. For example, tell it that your client wants to know which customers to target, so please describe which groups of people the client should focus on.

A simple Example:

This example is based on the closed questions from a NewMR study conducted in 2024, looking at levels of optimism in market research around the world.

Here is the table I uploaded:

Data Tables

I uploaded the Table and asked ChatGPT to read the file and be ready to answer questions. Below I set out a short example of the process I discussed earlier.

Prompt: “How many interviews were collected in Europe?”

Response: “The total number of interviews collected in Europe (including the UK) is 128, as indicated in the provided crosstabulation data”
This was very easy to check.

Prompt: “What is the base for Europe in the optimism section?”
I asked this because I knew this was an exception.

Response: “The base for Europe in the optimism section is 125 respondents. This slightly differs from the overall total of 128 respondents initially indicated, suggesting there may be a small number of respondents who did not answer the optimism question, leading to a slightly reduced base size.”
This confirmed it was reading the right numbers from the right place.

Prompt: “Create a table that Shows the base for each Region, Role, and level of Optimism for each of the three sections”
This is the last check that ChatGPT understands my data.

Response:

Table 1

Prompt: “For each of the three variables listed above, show me the percentage who are optimistic (Optimistic + Very Optimistic)”
One advantage of using an AI is that we do not have to use a scripting language or scripting precision to ask for outputs.

Response:

Table 2

Prompt: “Ignore those headings (e.g. Academic) that has a base of less than 30, create list of the groups from geography and role that show the level of optimism and sort that list from highest to lowest”

Response:

Table 3

I don’t know why it chose to include one of the ‘Other’ rows 😊

Prompt: “If the question is how does optimism and pessimism vary by role and region, please write a summary of what this data suggests?”

Response:
“Summary of Optimism and Pessimism by Role and Region:

The survey data indicates clear variations in optimism and pessimism levels within the insights and market research industry when segmented by respondent role and geographic region.

By Region:

  • Respondents from Asia Pacific show notably high optimism, with nearly three-quarters (72%) optimistic or very optimistic. This suggests confidence and positivity in this region towards the industry’s future.
  • In contrast, Europe (59%) and USA & Canada (59%) respondents have lower optimism levels compared to Asia Pacific but still indicate generally positive sentiment.
  • Other regions, such as Africa & Middle East and Latin America, have smaller bases but suggest moderate optimism levels (around 50–57%), implying more cautious outlooks.

By Role:

  • Optimism is highest among Research or Consultancy Suppliers (65%), highlighting confidence within professional research service providers about the industry’s trajectory.
  • Research Buyers/Users (58%) and Suppliers to the Research Industry (56%) express comparatively lower optimism, which might indicate greater awareness of market challenges or uncertainty from those directly commissioning research or providing operational tools and services.

Overall Patterns:

  • Across both role and geographic regions, a majority maintain an optimistic outlook. However, there are marked variations reflecting different economic environments or perspectives shaped by industry roles.
  • Pessimism is relatively low overall (around 11%), suggesting most professionals lean towards optimism or neutrality rather than negativity.

Conclusion:

The data indicates generally positive industry sentiment but underscores regional differences, with Asia Pacific professionals most optimistic, while roles associated with direct purchasing or support services exhibit more tempered optimism, possibly reflecting distinct operational pressures or market expectations.”

Remember, this was a simple example of how to do it. In a real project the data would be bigger, the analysis longer, and the checking more complete.

Learn More

Do you want to know more about using AI to process Quant data? I have two courses coming up soon.

The Research Society, Australia, also hosts the following courses, timed for Australia.

Leave a Reply

Your email address will not be published. Required fields are marked *