Classical fights — how moat-on-moat analysis works pre-AI

Why this chapter exists

Most moat-on-moat analysis still happens the way it did in 1995. A strategist names the defender's primary moat, names the attacker's primary moat, and asks which one is structurally vulnerable to which. Helmer's seven-power vocabulary — Scale, Network, Counter-Positioning, Switching Costs, Brand, Cornered Resource, Process Power — is the cleanest version of that grammar. Used well, it tells you the shape of the fight before you argue the specifics.

This chapter walks through six recurring patterns from that grammar. They are the baseline: classical theory taken on its own terms, before chapter 09 argues the picture is incomplete in three specific ways. Each pattern names a defender moat type, an attacker moat type, the verdict the matchup typically resolves to, and a concrete fight where the pattern played out. Read together, they show what a strategy room can do with seven moats and a willingness to commit to a specific structural diagnosis.

Six classical patterns

1. Different-wedge consolidator beats incumbent on a cannibalization bind

The defender holds a Switching-Costs moat in one customer wedge and cannot ship the new wedge's product without breaking its own pricing model. An attacker who picks a different wedge with a different pricing model wins, not by being better but by being structurally unreachable. Procore versus Autodesk on the general-contractor side is the canonical AEC case. Autodesk owned the architect and engineer seats, but its per-seat licensing model could not enter the GC office without cannibalizing the design-side business that funded everything. Procore picked a customer (the GC), a buyer (the operations leader, not the designer), and a pricing model (annual platform subscription tied to construction volume) Autodesk could not match, and won the field-software category outright. The classical name for this is Counter-Positioning beats Switching Costs when the cannibalization bind is structural — not when the incumbent is merely slow.

2. Pricing-arbitrage attacker stays independent because the incumbent cannot follow

The defender holds dominant Switching Costs, but the price the customer pays for those switching costs is high enough to fund a permanent gadfly. The attacker prices on a different model — perpetual license against subscription, consumption against seat — and survives indefinitely as a pricing-pressure relief valve, because the incumbent cannot drop its price without breaking its own P&L. Bricsys / BricsCAD versus Autodesk has run this play for twenty years. BricsCAD is a perpetual-license CAD package that reads and writes the same DWG file format as AutoCAD, priced at roughly a quarter of Autodesk's subscription cost over a typical hold period. Autodesk cannot drop its seat pricing to match without rebuilding its own revenue base. Bricsys does not need to win the category; it needs only to remain reachable. The same lineage runs through the AI era, where per-outcome and per-home pricing models pick up the same lever with sharper teeth (chapter 13 returns to this).

3. A network defender beats a brand attacker more lopsidedly than people expect

The defender holds Network Economies — liquidity between users on a platform. The attacker has a strong brand but no network of its own. The conventional intuition is that a great brand can pull users away one at a time. The actual mechanic is harsher. A brand without liquidity has no path against a dense network, because the user does not buy the platform's reputation; they buy the other users on it. eBay's buyers showed up because the sellers were there; the sellers were there because the buyers were. A trusted, well-known catalog retailer with no two-sided liquidity could not chip at that with reputation alone. The mismatch is structural: the network defender wins almost every cell of this matchup, because the brand attacker has no mechanism to recruit the other side of the marketplace.

4. Cornered Resource beats almost everything when the resource is a regulated right-of-way

The defender holds an exclusive input gated by a regulator: spectrum, landing slots, banking charters, port access, plan-review authority, payment-rail integrations. Once issued, these are non-replicable on any timeline a competitor can finance. Plaid is the modern canonical case: thousands of bank integrations built under regulatory constraints that would take a new entrant years and a different relationship structure to replicate. They beat Scale because the gate is regulatory, not capital. They beat Networks because the network forms on top of the right, not around it. They beat Counter-Positioning because there is no alternative model the regulator has yet authorized. The qualifier "almost" matters — regulatory regime change, substitute technology, or antitrust unwinding can dilute the right — but while the regime holds, the moat is as deep as moats get in classical theory.

5. Counter-Positioning beats Brand only when the cannibalization bind is structural

This is Helmer's most-cited matchup and the one most often misapplied. The argument is not that brand never matters; it is that brand cannot save an incumbent whose own P&L structure forbids it from shipping the challenger's product. Vanguard ate active asset managers' brands not by competing on reputation but by counter-positioning on a structure — mutual ownership plus index funds — that the active managers' existing AUM economics could not match without dismantling their own fee base. Netflix did the same to Blockbuster: late fees were the bind, brand could not pay the bill of cannibalizing them. The discipline is in checking the bind. If the incumbent could ship the new model and chose not to, the moat is not Counter-Positioning — it is management failure dressed up. Brand will win those, given enough time. Brand will not win when the structure forbids the response.

6. Process Power survives revelation but not delegation

The defender holds replicable organizational know-how that competitors cannot copy even when the playbook is published. Toyota Production System is the canonical case: every detail has been written down, taught, and toured for decades, and almost nobody else does it. The know-how lives in tens of thousands of small habits that have to be acquired in sequence by a culture that grew them. Process Power is robust against revelation. It is not robust against delegation. If a capable model can produce the output the process firm produces, at lower cost, the moat compresses regardless of whether anyone else can replicate the culture. That second test — revelation versus delegation — is the one classical theory had no reason to draw, because in 1995 there was no plausible delegate. Chapter 12 returns to it.

The pattern under the patterns

Each classical moat type has a distinctive failure mode. Scale fails when the segment fragments. Networks fail to multi-homing or counter-positioning. Switching Costs fail to migration tooling and structural cannibalization binds. Brand fails when the consequence of being wrong outruns reputation. Cornered Resources fail when the regime changes. Process Power fails when know-how decomposes into something a model can absorb. The seven-power vocabulary is most useful not for naming what a moat is but for naming the matchup in which a moat is structurally vulnerable. Good attackers find that matchup; lazy attackers fight every cell at once and lose every one of them.

The classical strategist's job, in this picture, is to name the shape of the fight before fighting it. That naming is enough to predict most outcomes most of the time. It is not enough for three specific kinds of fight that the AI era has made common.