Product
Import, clean, test, chart, predict and report, without the code.
CLAIR is data analysis software for spreadsheets and survey data. Ask a question in plain English and get an answer you can check: answers to questions about your data show the rows and the calculation. By default it all runs on your own computer, and analysis works offline once it is set up, with a licence check online at least every 14 days.
Ask
Ask in plain English. See where the answer came from.
Type a question the way you would ask a colleague. CLAIR runs a calculation on your data and shows you what it did.
- Questions about your data: the rows and the calculation are shown.
- Statistical tests: the test, its inputs and its assumption checks are shown.
- Predictions: the columns the model used and its accuracy are shown.
- Written summaries: the AI model wrote them, and they say so.
What Verified means. The label means the number came from a calculation that actually ran on your data and passed CLAIR's automatic checks. That does not prove the question was understood the way you meant it, so the method and the query still deserve a look.
You ask
What was the average hospital stay last quarter?
4.2 days
VerifiedThe average of the length of stay column across 1,240 rows.
Example answer, not real data. In CLAIR, an answer to a data question shows the rows and the calculation behind it.
Files
Open almost any data file.
Drop in a spreadsheet, a statistics package file or a database, and CLAIR works out the columns for you. Nothing to convert first.
Spreadsheets
Statistics packages
Data files
Cleaning
Understand the file before you trust it.
CLAIR reads the structure for you: column types, totals rows, and codes like -99 that usually mean no answer. It proposes fixes in plain sentences and you confirm them. Many changes can be undone, but not every operation can, so keep your original file.
- North120
- South-99
- Subtotal Sales315
- West95
-99Looks like "no answer", not a sale. CLAIR offers to make it blank, and you confirm.
SubtotalLeft out of the totals, and named so you can see it.
Statistics
The right test for the question.
Choose what you want to compare and CLAIR runs a suitable test, checks its assumptions and tells you if they do not hold. A test weighs the evidence under its assumptions, and an effect size describes how big a difference is. Neither shows what caused it.
Did scores differ between the two classes?
- Group sizes
- Class A n = 48, Class B n = 51
- Evidence under the test's assumptions
- p = 0.004
- Effect size: how big the difference is
- Cohen's d = 0.60, medium
- 95% CI for d: 0.20 to 1.00
Assumptions checked
- Similar spread
- Enough rows per group
Example result card, not real data.
Charts and dashboards
Charts when you need them, an overview when you do not.
Build a chart in a few clicks, or open the overview dashboard CLAIR prepares for each file. In CLAIR, click a mark to see the rows behind it.
Example charts, not real data. In CLAIR, click a bar to see the rows behind it.
Predict
Find useful predictors, then predict the outcome.
Pick a column you care about and CLAIR ranks the columns most associated with it, builds a model, and shows which columns the model used and how accurate it was. For harder problems, Deep Analysis runs a longer search that can be resumed if it is interrupted.
A pattern is not proof of cause. A column that predicts well may only be standing in for something else, and a prediction can be badly wrong for situations unlike the data it learned from. Treat what you find as a starting point to check.
CLAIR is for exploring and researching data. Do not use its predictions or other outputs to decide, rank or screen individual people for clinical care, a student's admission or grading, hiring or pay, credit, insurance or housing, or benefits and law enforcement decisions. The Acceptable Use Policy has the full list.
Most associated with the outcome
- Hours of tutoring86
- Prior grade64
- Attendance41
- Commute time12
Example, not real data. A pattern in the data is not proof of cause.
Deep Analysis
Can be resumed- Tried several kinds of model
- Tuned each one
- Checked on rows it had not seen
- Searching for a better one
A longer search that picks up where it left off.
Experiments
Plan runs, or pick the next one.
If you control the settings, such as a recipe, a machine or a process, CLAIR can plan all the runs up front or suggest the next one from your results so far. It first checks that your file looks like an experiment, and says so if it does not.
- 11602071.2
- 21802076.8
- 31803079.4
- 419030try next
Example. CLAIR suggests the next settings worth testing from the runs so far.
Exports and the record
Reports you can hand on, with a record of what was done.
Export a report as a PDF, PowerPoint or HTML. Results stay tied to the version of the data that produced them, and CLAIR writes a plain-language Methods paragraph from the steps you took. Exports with AI-written text name the model used. That helps you document your work, but it is not a full record of your methods, so keep your own notes too.
Methods
Values of -99 were treated as missing. The two groups were compared with a t-test, and its assumptions were checked.
Example. The paragraph only describes what was done in your session.
Try it on a spreadsheet of your own.
It is free during early access, and by default your data stays on your computer.
The Mac version is expected in late October 2026. Windows is available now.Free during early access, email sign-in. Mac expected in late October 2026.Free during early access for Windows 10 and 11. Mac expected in late October 2026.