calseda

Every figure calculated on Calseda opens to the inputs it was made from.

When a supervisor or an auditor asks how a figure was produced, most teams rebuild the answer by hand. On Calseda you open the figure and see its full lineage: every input it used, every input behind those, all the way down to the raw data, and export it as a report. The lineage starts where your data reaches Calseda.

Evidence for BCBS 239, taken from the figure itself.

Four of the Basel Committee's principles for risk data aggregation and risk reporting, and what your team can open on a live figure for each.

ThemeWhat Calseda shows
Governance (Principle 1) Roles from your identity provider decide who may do what. Each value entered by hand carries who entered it and when, and the reason where the record has one. An AI agent's entries are marked "AI agent for" its person.
Accuracy and integrity (Principle 3) The lineage shows each input as the figure used it; if one has changed since, you see both values. The exported report defines each kind of data in it, from your own documents.
Timeliness (Principle 5) Figures recalculated when their data changes, and flagged stale when an input is too old. Calseda measures how long each change took from its source to the figure.
Adaptability (Principle 6) A new question is answered on the live data without changing anyone's figures. An analyst can try a new calculation the same day, without touching production.

Lineage traces values, not database columns.

A figure looks wrong, and risk asks why it is what it is.

Today a team traces it through spreadsheets and batch logs for days, finds a manual adjustment nobody remembers making, and answers with caveats.

On Calseda, risk opens the figure the day the question arrives. Each input shows its value and age, flagged where it was older than its source allows. The adjustment is one of those inputs, carrying who entered it and when, and the answer goes back with the lineage report.

Every row on the demo's limit board opens down to the data as it arrived.

Each input shows the system it came from and its age. Another page maps the data flow through the whole server from what the calculations read, so it can't drift from the code.

A demo's limit board with one row opened to its inputs, sources and ages A demo's limit board with one row opened to its inputs, sources and ages
One figure opened to its lineage, each input with its source and age.
A demo's data flow map from sources through calculations to results A demo's data flow map from sources through calculations to results
The server's data flow, from each source to each result, as it runs now.

Start with the figure that took longest to explain.

  1. Write the calculation, or call the engine you run

    Your own analysts write it, inside your own environment, and Calseda runs the rest. Recordings of real inputs test each new version against the last.

  2. Put last audit's questions to the live figure

    Today's process keeps running beside it.

Make one reported figure answer for itself

Become a design partner: one reported figure and its lineage live in about six weeks, for a fixed fee. At the end you keep the framework, licensed to run what you built on one server.

For your auditor or data office: Regulatory data, one page (PDF).