Curve construction is an engineering discipline

Most curve engines fail for reasons that have nothing to do with maths.
Curve construction isn’t a mathematical problem. It’s an engineering discipline.

The theory has been understood for decades. Bootstrapping, interpolation and optimisation are familiar tools for any quantitative team.

Yet many curve engines still behave unpredictably once they hit production.

Why?
Because live environments are hostile.
- Incomplete data
- Illiquid instruments
- Regime shifts
- Sudden market stress

In that reality, curves must do far more than “fit the data”. They must recalibrate reliably, remain explainable, and behave consistently across trading, risk and governance.

That requires discipline beyond selecting the right formulae.
In practice, robust curve construction is about engineering choices:
– calibration routines chosen for stability, not elegance
– explicit assumptions rather than implicit defaults
– full traceability of inputs and outputs
– the ability to reproduce historical curve states exactly, when questions arise

When curves are engineered this way, everything downstream improves. Pricing models become more stable. Risk metrics become more credible. Trust increases.

Pricing and risk should never be asked to compensate for fragile market-data preparation.

Clean inputs are not a luxury.
They are the foundation.

If this resonates with your production environment, let’s compare experiences.

‍