Announcing o11's backing from Y Combinator.
Academic Researchers: AI Data Analysis in Google Sheets
Academic research runs on data, and that data almost always passes through a spreadsheet at some point. Whether you are a social scientist coding survey responses, a public health researcher tracking longitudinal outcomes, or an economist organizing panel data, Google Sheets is where the messy middle of research happens. It is where raw exports land, where variables get recoded, and where preliminary analysis shapes the direction of your study.
The problem is that this messy middle eats an enormous amount of time. A 2024 survey by the UK Data Service found that researchers spend roughly 60 percent of their analysis time on data preparation rather than actual analysis. You know the drill: columns mislabeled by survey software, inconsistent date formats across collection waves, free-text responses that need categorical coding, duplicate entries from participants who submitted twice. None of this is intellectually demanding, but it is tedious and error-prone, and a single miskeyed recode can quietly corrupt downstream results.
Most researchers are not programmers. They know enough Sheets or Excel to get by, maybe some basic R or SPSS. But translating a research question into the right sequence of formulas, filters, and pivot configurations is slow work, especially when you are juggling teaching, grant writing, and the actual thinking that makes research worthwhile.
Survey Data Cleaning and Normalization
The first bottleneck in any survey-based study is getting your raw data into a usable state. Survey platforms like Qualtrics, SurveyMonkey, and Google Forms each export data differently. Column headers are often question IDs rather than variable names. Likert scale responses might be stored as text (“Strongly Agree”) in one wave and as numbers (5) in another. Missing data is coded inconsistently: sometimes blank, sometimes “N/A”, sometimes -99.
o11 For Google Sheets understands your data layout and can handle these transformations in plain language. Instead of writing nested IF statements or VLOOKUP chains to recode variables, you describe what you need.
“Recode column F from text Likert responses to numeric 1-5 scale where Strongly Disagree = 1 and Strongly Agree = 5. Flag any responses that don’t match these categories in a new column G.”
“Find duplicate participant IDs in column A. For duplicates, keep the row with the most recent timestamp in column B and move the other to a new sheet called ‘Removed Duplicates’.”
o11 reads the structure of your workbook, identifies the relevant columns, and executes the transformation. It does not just fill in a formula for one cell and leave you to drag it down. It processes the entire dataset, handles edge cases, and gives you a clean audit trail of what changed. When your IRB asks how you handled data cleaning, you have a clear record.
Statistical Analysis with Pivot Tables and Descriptive Statistics
Once your data is clean, the next step is exploratory analysis: descriptive statistics, frequency distributions, cross-tabulations. In a traditional workflow, you would manually configure pivot tables, write AVERAGE and STDEV formulas, and spend time formatting output tables to match your reporting needs.
With o11, you describe the analysis you want and it builds the entire structure.
“Create a summary statistics table for variables in columns D through L. Include mean, median, standard deviation, min, max, and count for each. Exclude rows where column C is marked as ‘incomplete’.”
“Build a pivot table on a new sheet showing the mean score for each condition group in column E, broken down by demographic category in column H. Add a column showing the difference between groups.”
o11 creates properly labeled pivot tables, adds the formulas, and formats the output so it is readable. It understands that academic users need precision: decimal places matter, labels should use variable names not column letters, and summary tables should be structured for easy transfer into a paper or poster.
This is not a chatbot generating a block of text about your data. o11 operates directly in your spreadsheet, building real formulas and real pivot tables that you can inspect, modify, and verify. Your advisor can open the sheet and see exactly how every number was calculated.
Research Dataset Cross-Tabulation and Correlation Analysis
The most time-consuming analytical work for many researchers is exploring relationships between variables. You need cross-tabulations to understand how responses distribute across categories. You need correlation matrices to identify which variables move together before running your formal models.
Building a correlation matrix manually in Google Sheets means writing CORREL formulas for every variable pair, which for 10 variables means 45 individual formulas. Then you format it into a readable triangle matrix. Then you do it again for a different subset of your sample.
“Generate a correlation matrix for the numeric variables in columns D through M. Format it as a lower triangle with values rounded to 3 decimal places. Highlight correlations above 0.7 or below -0.7 in bold.”
“Cross-tabulate the responses in column F by the categories in column J. Show both raw counts and percentages. Add a row for column totals and a column for row totals.”
o11 builds these structures in your sheet with the actual formulas visible and editable. If a reviewer asks you to add a variable or run the analysis on a different subsample, you can ask o11 to update the existing table rather than rebuilding from scratch.
Before and After: A Research Data Workflow
Before o11: You export 2,400 survey responses from Qualtrics. You spend three hours renaming columns, recoding Likert scales, and removing duplicates. You spend another two hours building pivot tables for descriptive statistics. You manually construct a correlation matrix, which takes 90 minutes and a round of error-checking when two values look wrong. Total: a full working day before you have even started the real analysis.
After o11: You export the same 2,400 responses. You describe the cleaning steps in three prompts. o11 recodes, deduplicates, and flags issues in minutes. You ask for descriptive stats and a correlation matrix, and they appear on new sheets, properly formatted and labeled. You spend your time interpreting the patterns instead of building the infrastructure. Total: under an hour of active work, with more confidence in the output.
Why o11 Instead of ChatGPT or Generic AI
You could paste your data into ChatGPT and ask it to analyze your survey results. But ChatGPT does not operate inside your spreadsheet. It generates code snippets or text summaries that you then have to manually transfer back. It cannot see your actual data structure, cannot build formulas in your cells, and cannot update a pivot table when you add new data.
o11 works inside Google Sheets as a native layer. It reads your workbook structure, understands relationships between sheets, and creates real spreadsheet objects: formulas, pivot tables, conditional formatting, charts. When your co-author opens the file, they see a normal Google Sheet with transparent calculations, not a black box.
For researchers who need reproducibility, auditability, and the ability to hand off a working spreadsheet to a collaborator or a journal reviewer, the difference is fundamental. Your analysis lives in the spreadsheet where anyone can verify it, not in a chat transcript that disappears when you close the tab.
