The Instruments  /  The Question Sets  /  The risk models

The questions a board should ask about your risk models

Eight questions for a world that stopped being stationary. Models rarely fail loudly — they keep answering, with great confidence, questions from a world that has ended.

Which regime was this model trained on — and does that regime still exist?
Every model — risk, pricing, credit, planning — was trained on data from some regime. This is the question to ask of each, by name.
Do we have a list of every model the institution materially runs on — and when was each last re-validated against post-2022 data?
The stock–bond correlation inverted in 2022 and stayed inverted. Anything validated only against the prior regime is validated against a world that ended.
What does each model assume stays constant — correlations, volatility bounds, tail frequency — and what is the current evidence each assumption still holds?
Assumptions are not wrong for being assumptions. They are wrong for being unexamined.
Where are we diversified on paper but concentrated in behaviour?
An index of 505 stocks can behave like roughly 49. Concentration at the index, sector, and name level does not average out — it multiplies.
If the hedge we rely on inverted — as the stock–bond hedge did — what else in our framework rests on the same minus sign?
A generation of portfolio construction and risk mathematics was built on one correlation’s sign.
How quickly would we know the distribution moved — and through what signal?
The professional pricers of risk reprice visibly when it does. US motor insurance has moved at several times its historical rate. What is our equivalent instrument?
Is the financial reset a tailwind or a threat for our business — and at what threshold does the tailwind become the threat?
Most balance sheets carry both answers at once. The threshold is the useful number.
Who is empowered to say “the model is wrong” — and when did anyone last say it?
A model nobody may challenge is no longer a model. It is a policy wearing mathematics.
The point of these questions is not to distrust models — it is to know, model by model, which world each one still lives in. The readings several of them draw on are computed, with method shown, in the Ledger at bharatravuri.com/ledger.

Built to be carried into your institution. Take it to the room that decides.

From The Instruments at bharatravuri.com — the frameworks are open; the value is what’s brought into the room. Free to print, forward, and table. — Bharat Ravuri, The Practitioner Thinker.
All question sets
An invitation

If any of this meets where you are, I would welcome the conversation.

No intake process, no gatekeeper. A note reaches me directly, and I reply myself — usually within a day or two.

This comes straight to my inbox. I read every note personally and reply myself — not a team, not an auto-responder.

Message received

Thank you — your note has reached me.

It comes straight to my inbox, and I reply personally — usually within a day or two. I look forward to the conversation.

Or reach me directly
Mobile
Call or message — India (IST)