The Data Storytellers Podcast

The episode is predominantly a biographical career walk-through with long personal tangents; genuinely useful ideas - like two-quarter profitability turnaround via risk-model changes, or the semantic-layer-plus-insights-layer framework for AI analytics - surface occasionally but are never developed with enough depth to be actionable. Filler and anecdote dominate the runtime.

“in the span of two quarters, we were able to actually move the company from, uh, being unprofitable to profitable”

“I don't think it's uh, the right concept of here I'm just going to pass a bunch of raw data to an LLM, um, and expect it to give me all the answers”

The semantic-layer-plus-insights-layer framing for grounding AI analytics is a modestly fresh angle, but the majority of takes - trust in data first, AI won't replace humans wholesale, hallucinations are real, L5 autonomy as analogy - are widely circulated ideas in the data/AI discourse with no meaningful first-principles development.

“how do I connect the dots between different tables and connect that to a presentation someone just did around m why revenue went up and doing the diagnostics around that”

“here's then all of the work that it's going to show you back so you can build that trust”

Mangahas is a genuine multi-discipline practitioner who simultaneously ran analytics, FP&A, and fundraising at a Series C/D marketplace, led data transparency at Meta, and navigated a marketplace through COVID-induced market collapse - an unusually broad operational résumé. He is not a C-suite unicorn founder or public-company CEO, but he has credibly done hard things at meaningful scale.

“I was also, since I was head of fpa, also, uh, playing a large part of the fundraising process for the next round of fundraising for the Series D. So it was fundraising, leading the FP&A team, as well as leading analytics”

“when I joined, it was a team of one. Uh, and actually the scope was, was limited. It was just focused on their biggest account, which was Walmart”

There are scattered concrete anchors - Series D at over $100M, 85% car utilisation figure, the 90/10 and 75/25 host protection-plan splits, two-quarter profitability turnaround, team scaling from 1 to ~20 - but the actual data and analytics work that produced these outcomes is described almost entirely in vague terms, with no model names, experiment results, or causal mechanisms named.

“We raised a little over 100 million”

“85% of the time they're just sitting in your driveway”

The host regularly injects extended personal anecdotes (car rental in Rhode Island, European vacation, selling his London company) that displace guest insight time, and at one point explains the Big Four accounting firms to a six-year PwC veteran. Questions are open biographical prompts with no follow-up pressure, no numbers challenged, and no disagreement surfaced across 84 minutes.

“So the big four. So we have, uh, ey, we have PwC, we have McKinsey. Right. And. And is it BCG?”

“So I was flying to Rhode island for a baptism, and I flew through New York. Right, right. And I, in advance, was, like, very prudent about the. The car.”