Drop in a file: a survey, a tracker, reviews, support tickets, a spreadsheet. Ask in plain words. Every figure is computed by the engine, never written by the model. Every claim carries its receipt. When the data cannot carry the claim, it says so.
Request a trialEvery column is classified from its own rows. Personal data is set aside. You check the reading before the room opens.
The model writes a plan. The server checks every step against what the data holds, runs it, and checks the words against the figures before you see them.
A figure cites its variable, its value and its base. A quote is matched character for character. A claim that cannot be matched is never shown.
Crosstabs, trends, gradients, tests, regressions, forecasts, text themes and quotes. One question, one population, one dimension, optionally against another. Every table has a base and a test on each cell.
Installed on your machine. The file never leaves it. Only the question and computed tables reach the model.
Log in at app.talktask.si. Nothing to install. One space per company.
It will not invent a figure, a quote or a respondent. It will not answer a question the data cannot carry. It will not apply a weight, scale a column or mark a value as missing unless the rows allow it. These refusals are counted as correct answers, and they are tested that way.