Financial Modeler's Corner

There are genuine ideas here - the xirr-naming-convention critique, the three-part decomposition of Excel, and the AI-needs-standardised-language argument - but they are spread thin across a 51-minute episode padded with a hug anecdote, keyboard-warrior banter, and rapid-fire yes/no filler. The insight rate is modest, not dense.

“when people write or when people do an IRR calculation in Excel, they often use the Xirr formula and they will write in their spreadsheet xirr as the row label... But xirr is just an Excel function. It's not a real thing.”

“Will the Excel as a creative tool die? Yeah, quite possibly... Will the Excel file format die? No. So the middle thing will endure”

A handful of fresh framings appear - applying Hotelling's law to modellers sharing semantic space, the contrarian claim that sensitivities have 'almost zero value', and the nuanced Excel-as-three-things split - but most of the episode recycles well-known practitioner concerns about silos and jargon without pushing them to a genuinely novel conclusion.

“Sensitivities have almost zero value because no one can quantify the variation that they expect. Whereas scenarios are real cases that correspond to something that we know could happen in real life.”

“if the financial modelers can go into a space, a kind of semantic space and share a common place, then people will find it”

John Yeldham is a genuine senior practitioner who created methodologies at BDO UK and Forvis Mazars, has deep project-finance and infrastructure experience, and is building a training product - not a career podcaster or abstract thought leader. He lacks the marquee name or scale of a true industry-shaping operator, which caps the score.

“John created the BDO UK modeling methodology, renewed the methodology at UH Forvis Mazars”

“I had the FD sat there almost having a nervous breakdown, and at the end, we got that number, and he gave me a hug”

The episode offers illustrative micro-examples (the car park algorithm, the district heating non-linearity, the xirr label critique) but almost no hard data - no named clients, no revenue figures, no measured outcomes, no published research beyond a passing reference to 'pedagogical research'. Claims like 'more than half of all models' are stated without sourcing.

“I did one for a large car park company and it needed to effectively work out the priority of car parking spaces for bookings in advance”

“I've done district heating, which has plants and so on. And when the temperature changes in a city with district heating, it's a non linear relationship”

The host asks decent second-level questions (why hasn't language standardised, what will it take) and occasionally pushes back lightly on sensitivity vs scenario, but he more often validates and shares his own anecdotes - the extended hug story, the keyboard-warrior digression - rather than pressing the guest on weak or unsubstantiated claims. Rapid-fire yes/no adds little substance.

“Why do you think the language hasn't become standardized?”

“I think they're a little helpful when the relationship may not be what you expect to figure out how sensitive something is”