Megan Skiendel: Miles — I have been looking forward to this one all week, which probably says something unflattering about my week.
Miles Ashworth: Deeply unflattering, yes. What went wrong?
Megan Skiendel: I read a pitch deck on Monday that used the word 'moat' eleven times in fourteen slides, and not once — not once — did they explain what would actually make copying them hard. It was just... the word, used as a spell.
Miles Ashworth: Good lord. Eleven times.
Megan Skiendel: Eleven. Which is, it turns out, a perfect entry point for what we're actually digging into today — what a real economic moat is, where Buffett's original 1995 framing actually holds, and where the language became, as you'd say, intellectual theatre.
Miles Ashworth: The definition is simple enough on paper — a durable structural advantage that keeps rivals from replicating your position even when they try very hard. What's happened is the temporary advantages — a hot product, a charismatic founder, first-mover luck — these get labelled as moats because the vocabulary flatters the person doing the labelling. And the question I keep returning to is: can you actually identify a real structural moat before it fails? Or is it always forensics?
Megan Skiendel: That's the one. Because a company can look dominant for decades and then lose the thing that made copying it hard — not because a rival outsmarted them, but because that structural mechanism just... stopped being hard.
Miles Ashworth: Quite. And that's the puzzle — eighteen months, sometimes less, and something that looked unassailable turns out to have been a very convincing costume.
Megan Skiendel: That costume line is exactly the problem — and it's what Pat Dorsey was trying to cut through. He spent years at Morningstar studying every company that held above fifteen percent return on invested capital for fifteen years or more, and what he found is that the ones who survived had something mechanically hard to copy. Not just 'doing well.' Mechanically hard.
Miles Ashworth: Mechanically hard meaning — what, precisely?
Megan Skiendel: Okay, think about your phone number. You've had it for — I don't know, fifteen years. Switching isn't just inconvenient. It breaks every contact you have, every app linked to that number, every two-factor code. That's a switching cost. The new carrier doesn't have to be worse. The cost of leaving is what locks you in. That's a structural barrier. That's a moat.
Miles Ashworth: Good lord, yes — SAP is basically this. Multinationals running SAP don't stay because SAP is delightful. They stay because their entire data schema, their trained finance teams, their compliance workflows — all of it is built around this thing. The product could be actively unpleasant and nobody leaves.
Megan Skiendel: Exactly. And Dorsey systematised that intuition into four sources: switching costs, network effects, intangible assets — patents, brand, licences — and cost advantages. Four. That's the whole framework from the Morningstar work. But — and this is the bit that gets dropped — he also separated width from durability. How strong today versus how long it survives.
Miles Ashworth: Wait — those two can point in opposite directions?
Megan Skiendel: They absolutely can. A wide moat built on one fragile mechanism — actually, this is the part that genuinely surprised me when I first worked through Dorsey's framework — it can collapse faster than a narrower moat built across multiple sources. Because you only need the one thing to break.
Miles Ashworth: Which is essentially why he left Morningstar in 2014 and set up Dorsey Asset Management as a concentrated twelve-stock fund — he was betting actual capital that he could identify which mechanisms were genuinely durable, not just wide. That's a rather pointed statement of confidence in the framework.
Megan Skiendel: It is. Twelve stocks. Global. No hiding in diversification. And the bet is that if you've correctly identified the structural barrier — not the performance gap, the barrier — you can hold through noise. The question, and it's one I haven't fully resolved, is whether width and durability are ever legible before the mechanism gets tested.
Miles Ashworth: Legible before tested — frankly, that's the whole problem with switching costs in particular. Because it's the most claimed moat and also the most abused. You can have genuine architectural lock-in, the way SAP does when a finance team has spent three years building custom month-end close reports in its data schema and trained forty people to live inside it — that's not friction, that's institutional memory embedded in software. Migration isn't buying new software; it's burning the library. But then you have the other thing. Where nobody's left yet... and that gets filed as switching costs as well.
Megan Skiendel: The 'nobody bothered to leave' moat.
Miles Ashworth: Exactly. And they are not the same thing at all, but they look identical until someone leaves.
Megan Skiendel: This is what the Morningstar AI re-evaluation is actually surfacing. They went through 132 companies — 132 — and the finding is that AI erodes moats built on workflow friction and application stickiness, while the ones built on proprietary data or infrastructure survive. Which is basically AI telling you which lock-in was real.
Miles Ashworth: Wait — Salesforce sits squarely in that second category then, yes? Because the data your sales team has accumulated in that CRM over eight years, that's not stickiness. That's an actual asset you'd be abandoning.
Megan Skiendel: That's the distinction. Apple's the same logic — the hardware, the software, the services, each layer makes leaving every other layer more expensive. That's architectural. What AI removes is the human effort of migration — actually, no, it reduces it — which exposes whether there was anything structural underneath.
Miles Ashworth: And the thing that survives AI — this is the part I didn't expect — isn't the technical barrier. It's psychological. A procurement director on a Thursday afternoon isn't asking whether the new vendor's product is better. She's asking whether that vendor will still exist in three years. The risk calculus is about vendor survival, not feature parity.
Megan Skiendel: Oh, honestly — that's the part internal teams consistently get wrong. They benchmark features, they never benchmark trust. And that's before we've even touched what happens when you stress-test network effects the same way, which is where this gets considerably more uncomfortable.
Miles Ashworth: Quite. That's a different problem entirely — and a worse one.
Megan Skiendel: Network effects looked like the safe moat — density as destiny, the whole story — but I watched that collapse in real time. We had the two-sided marketplace. More restaurants, more diners, flywheel spinning. Textbook. And then DoorDash entered with better matching and we lost the density advantage in about eight months. The network still existed. It just... stopped mattering.
Miles Ashworth: Wait — the network was intact and it still stopped mattering?
Megan Skiendel: Intact. Because a marketplace with millions of sellers is worthless if the algorithm can't surface the right one at the right moment. Scale without matching quality is actually — no, it's worse than neutral. It's noise. The Morningstar framing nails this: classical two-sided density is obsolete. What accumulates as a real moat now is algorithmic matching capability.
Miles Ashworth: Which is the PayPal problem, isn't it — great product, apparent network effects, and Morningstar still flags it as a cautionary example because the mechanism wasn't as durable as the framing suggested.
Megan Skiendel: Exactly PayPal. The effects were real — they existed — they just didn't compound the way density-story investors assumed. And here's what surprised me: the matching-over-density insight predates AI entirely. This wasn't about language models. Platforms were already proving that quality beats scale before anyone said the words 'large language model.'
Miles Ashworth: Good lord — so the AI disruption conversation is actually arriving late to a party that the platforms already threw.
Megan Skiendel: Years late. Which reframes the ROIC problem too — Pat Dorsey flagged that the ROIC signal degrades for asset-light businesses because the capital denominator approaches zero, inflating returns regardless of actual competitive strength. You're screening for moats and pulling up companies that just have thin balance sheets.
Miles Ashworth: So the signal's broken and the dominant theoretical model is obsolete. We've been mis-rating a rather large number of companies on two separate faulty assumptions simultaneously — that is, frankly, quite a significant mess.
Megan Skiendel: The practical consequence is: any analyst who rated a platform on user count rather than matching quality in 2018 was writing a story about a moat that was already shifting underneath them. The diagnosis had already stopped fitting the disease.
Miles Ashworth: Which leaves you with one question, really. Not 'is this company winning' — that's forensics dressed as analysis. The actual question is: what, specifically, makes copying this business structurally hard? Porter's Five Forces is sitting underneath all of it — a moat is just the permanent suppression of one of those forces. If you can't name which force and why suppression holds, you don't have a moat thesis. You have a mood.
Megan Skiendel: And brand is the one that trips everyone. Because it looks like a mechanism — intangible assets, Dorsey's framework, it's on the list — but it's actually a lagging indicator of past product quality. The moment the product stops justifying it, the brand evaporates. Fast. You can't measure it until you're measuring the collapse.
Miles Ashworth: The castle's already falling. The operational excellence is just furniture at that point. That's — yeah, that's where I land on all of it.
Megan Skiendel: Me too. Honestly, that pitch deck on Monday — I think they'd have been fine if someone had just asked them once: what breaks if a well-funded rival spends three years trying to copy you?
Miles Ashworth: Eleven slides of moat and nobody asked the one question. Quite.
