Tanveer Bhatti argues that modern stress testing must be designed to fail safely, not flawlessly. As systemic risks grow more nonlinear and AI accelerates model complexity, institutions must combine advanced analytics with governance discipline, uncertain
As stress testing expectations
intensify across global financial institutions, the uncomfortable truth is that
models tend to fail precisely when they are most needed.
Tanveer Bhatti, Early Stage Fintech
Investor and former Group Head of Model Risk at Revolut, believes the industry
has yet to fully internalize that lesson.
“I saw this managing model risk in a
G-SIB during 2008,” he says. “The models kept outputting numbers right through
the crisis. Validity collapses at the moment you need it most, and nobody in
the room has standing to say so because the model said it.
“The experience was repeated during
the pandemic. “2020 showed how quickly your calibration set becomes irrelevant.
Outliers appeared in almost every model simultaneously. Backtesting told you
nothing useful at that point,” he explains.
For Bhatti, the issue is
philosophical as much as technical.
“You're asking models built on
repeatable mass phenomena to price non-repeatable events. That's a category
error disguised as a quantitative one.”
What holds under pressure, he argues,
is not apparent precision but explicit uncertainty bounds, sensitivity
analysis, and hard circuit breakers when outputs drift outside credible ranges:
“The framework has to be designed to fail safely.”
Cascading effects present another
structural weakness. In times of stress, pricing, risk, profit and loss,
liquidity, and operations do not move independently.
“Cascades travel through actions,”
Bhatti says. “A margin call triggers collateral posting, that hits funding,
that changes client behaviour, that pressures capital.”
If stress testing models individual
nodes rather than those chains, institutions are mislabeling component analysis
as network resilience, he argues.
He resists the idea that a single
modeling methodology can solve this. Network models, agent-based systems, and
scenario overlays each capture different dynamics.
“The practical answer is
triangulation,” he says. Analytical rigor is meaningless if it cannot be
explained, governed, and executed in real decision environments.
Scenario design sits at the heart of
credible stress testing. Bhatti challenges the common belief that plausibility
depends on historical frequency.
“Plausibility doesn't come from
historical frequency. It comes from a coherent causal story with mechanics that
are defensible.”
The most common error is stacking
negative outcomes without narrative cohesion. “That produces a list of fears,”
he warns.
