The Shift Code

The episode delivers a handful of genuinely non-obvious empirical findings - particularly the 1-in-200 statistic and uniqueness bias as a red flag - but the conversation is padded with meandering AI chit-chat, mutual PMI promotion, and well-known platitudes about optimism bias borrowed from Kahneman and Tetlock. Insight rate is uneven rather than consistently dense.

“1 in 2 projects delivers on budget or better, but only 1 in 12 projects, about 8% deliver both on budget and on time or better... then that is about half a percent. So 1 in 200, 1 in 250 projects, depending on the sector, that deliver all these things together”

“the single biggest red flag that our data showed was the idea that my project is unique”

The uniqueness bias finding and the 20-30% planning investment threshold are genuinely fresh empirical contributions, but the episode also leans heavily on recycled frameworks (Kahneman's WYSIATI, Tetlock's chimp analogy, planning fallacy) and clichéd examples like the Sydney Opera House that circulate constantly in project management discourse. The AI discussion at the end is entirely generic.

“the single biggest red flag that our data showed was the idea that my project is unique, you know, so if people involved in the project... say, well, this is unique, the single biggest warning flag for having a massive cost blowout”

“the better projects... that invest somewhere between 20 and 30%, those actually perform better”

Alexander Boutier is a genuine empirical researcher from Oxford Global Projects with access to a 22,000-project database and a co-authored book with Bent Flyvbjerg - substantive credentials, not a carousel podcast guest. However he is primarily an academic analyst rather than an operator who has personally delivered large projects at scale, which limits the practitioner depth.

“our big database now of more than 22,000 projects across all sectors, from IT and organizational change to classic construction, transportation infrastructure, energy, defense projects”

“for my latest book, the how to Measure Anything in Project Management, we looked through our database”

The episode is notably strong on named examples with real numbers: the Lidl half-billion-euro IT write-off, Westinghouse/Vogtle bankruptcy, Carillion collapse linked to specific payment milestone failures, Denver airport's 18-month baggage system struggle, the NBER 40%-vs-5% AI productivity gap, and the core 1-in-200 statistic. This lifts the episode well above average on this dimension.

“in Germany, a supermarket chain called Little, they built a big IT system to digitize their accounting. They spent half a billion euros on this... it took one CEO and 15 months after go live, the system was... retired”

“Westinghouse went bankrupt over Vogtler in the UK here, our largest domestic construction company, Carillion, went bankrupt because of delays to their project and sort of they got startup of starved of cash because they missed some payment milestones on three different things”

The host asks a few genuinely pointed questions - directly challenging the guest on Agile and calling out the 'column A and column B' hedge - but too often pivots to promoting PMI frameworks and his own research rather than pressing the guest harder. The AI section devolves into mutual enthusiasm with no substantive pushback or follow-up on the corporate productivity gap claim.

“So is this because you're working mostly on very large projects or would it also apply basically, are you against Agile, let me ask it”

“I knew you would say that. Right, but how do you know?”