How XENTHK works
Eight steps from a file to an answer you can defend. Each step is visible, and wherever the data does not settle a question, you decide.
-
Bring the data
Add a file or connect a source. Files are read on your computer: CSV, Excel, JSON and delimited text. Sources are product analytics such as Countly, Matomo and PostHog, and any web address that returns JSON. Database connections with a user name and password follow.
Several sources can sit in one workspace. None of them is “the main one”: the question decides which data matters.
-
Read every field
XENTHK reads each field's name and values and proposes what it means: an identifier, a date, a price in a currency, a count, a category. Names it does not know are looked up and shown to you with their source.
You confirm the reading on one screen, including what one row stands for. Nothing is analysed before you confirm.
-
Check the data's health
Rows shifted by a stray comma, the same value written in different capitals or spaces, empty cells and placeholder codes are found while the file is read.
XENTHK never fixes them on its own. It shows each case once and asks: leave the shifted rows out, count the spellings as one, keep them as written.
-
Understand the question
You ask in your own words. XENTHK maps the question onto what your data can compute: totals, counts, distributions, relationships, ratios and the formulas of 1,712 KPIs.
When the data has no field for what you ask, it says so and offers the closest measure the data does hold.
-
Choose the analysis
The route follows the question. “Which group has the most” is a total by group. “How do two measures move together” is a correlation. “How spread out” is a distribution with its quartiles and outliers.
Every answer states the route it took, so you can see why a number is what it is.
-
Compute on your computer
The data engine runs in your browser and computes every table, total and coefficient there. Your rows and your results are not sent anywhere.
Large files stay fast because only the columns a question needs are read into each computation.
-
Check every number, limit every claim
Each number in an answer must come from a computed table, or the answer is not shown. A correlation is reported as association, never as cause. A difference is not called significant without a test.
A proxy stays a proxy: completed sales are reported as sales activity, not as demand.
-
Show it, keep it, report it
Each result comes with the charts that suit it, ready to switch, and exports to CSV and Excel. Mark the answers that matter with “Add to report”.
Today the report gathers the marked answers with their tables and claim limits, ready to print or export. The written report, which merges repeated findings and checks them against each other, is in development.
What runs where
On your computer
- Your files and the records read from your sources
- The data engine, which computes every table
- The analyst model, which reads fields and writes answers
- Every result, chart and export
At XENTHK
- Your account and sign-in
- Encrypted delivery of the model and the engine after you sign in
- An encrypted relay to the web addresses you connect; it carries bytes it cannot read
- The record that you accepted the usage notice
What XENTHK will not do
- Invent a number that no computed table holds
- Call a correlation a cause
- Call a difference significant without a test
- Change your data on its own, or keep your rows
See it on your own data.
Request access, add a file and ask your first question.