Two instruments, two different questions. One asks whether your growth mechanism compounds. The other asks whether what it built can be copied.
The Business Loop Diagnostic asks whether the mechanism compounds. Does each turn leave something behind — an asset on your side, retained value on the customer's — that makes the next turn cheaper?
The 8 Moats Diagnostic asks whether the result holds — and, earlier on, which results are even available to you. If a well-funded competitor comes after this, what stops them reproducing it?
Early on, the moats score reads low almost by definition. That's not a reason to skip it. Eight moats is a finite list, and your model only puts a few of them within reach — the diagnostic names which ones, so you can point at two deliberately instead of hoping something accumulates. The question shifts as you mature, from which moats should we be building toward? to is what we've accumulated hard to copy? If you don't know which one you're asking, start with the loop.
Score a company raising its Series A against eight moats and it will report EXPOSED almost by construction — no data asset yet, no regulatory position, no ecosystem, no distribution. Read as a grade, that tells you nothing you didn't know. It's a description of being early.
Read forward, it's the most useful thing the instrument produces. Eight moats is a finite list, and your model only makes a few of them reachable — how you sell, what you touch, and what accumulates each time the loop turns rule most of them out before you start. Naming the two you can actually build, and the four you'd spend years failing to build, is the Series A version of the question. Moats stack by Series C because someone chose them at Series A.
So the number matters later and the composition matters now. A 4/24 and a 6/24 at Series A are the same company; which four points they are is not. That's also why the paid Stress Test adds trajectory — whether each moat is widening or eroding — since direction is the part an early-stage total can't show.
The loop tells you whether anything is accumulating at all. The moats tell you which accumulation is worth pointing at. Running both is how you avoid the two failure modes — compounding into an asset anyone can copy, and defending a position nothing is feeding.
Most companies describe their growth as a flywheel. Most of them have a funnel they're refilling every quarter with money. The difference is whether the output of each cycle becomes an input to the next.
A self-serve check you run on yourself. Name your go-to-market motion and it names which of the five loop types should compound underneath — then, from your current conditions, which of them are actually available to you at your stage right now, and where your effort is going that can't compound. That result appears on the page immediately, with no email required. The written report — three prioritized moves, your closing-door flags, and the scoring rubric — takes an email, and I write it myself within one business day.
Something compounds on your side — the Compounding Asset — and something is retained on the customer's side. Each turn makes the next one cheaper.
The mechanism looks circular in the deck, but the output doesn't actually feed the input. Growth still costs the same per unit as it did last year.
Linear acquisition, honestly described. Not a failure — plenty of good businesses are funnels. But it changes what your growth plan can assume.
Each loop type is defined by what it retains and compounds over time — not by the channel it runs through. Co-created with Krzysztof Czubak.
Cadence bounds how many turns a loop can possibly have run — which bounds what can be observed rather than inferred. A trust loop compounding over years has barely turned inside an eighteen-month-old company, so it can only be assessed on whether it's designed to close. A user loop compounding per interaction may have turned thousands of times in the same company and is fully assessable. This is why the diagnostic takes turns elapsed rather than founding date, and why two companies at identical funding stages get different verdicts.
Compounding is measured in revolutions, not months. A PLG company at seed may have run thousands of loop turns; an enterprise company at Series B with a fourteen-month sales cycle has run three. Judging the second by elapsed time is a measurement error — the loop hasn't had enough revolutions to show.
So intake takes your go-to-market motion and your turns elapsed. Under roughly ten turns, the verdict is whether the loop is designed to close. Past fifty, we can assess whether it is observed closing. This is where the Stress Test goes past the free Diagnostic: the Diagnostic uses your motion to orient — naming the candidate loops and checking which your conditions make available. The Stress Test adds turns elapsed and evidence-graded assessment of whether those loops are observed closing — motion as one input to a graded verdict, not just the starting point.
A founder-led company runs on Trust and Skill; grading it against a User loop produces a false negative — calling a working business a funnel. Each motion names its own candidate loops: community-led runs on User, Content and Trust together; content-led on Content and the authority (Trust) it earns; event-led on who's in the room (User) and the reputation for quality (Trust); product-led on User and Skill. Sales-led runs on Trust; partner-led on Trust or Capital.
The failure mode we design against is judging one motion's loop by another's standards — an enterprise trust loop graded as if it were product-led.
Each claim is marked OBSERVED, INFERRED, or REQUIRES CLIENT INPUT. An early-motion company will come back mostly INFERRED, and that's honest rather than a weakness — it tells you which parts of the verdict rest on public evidence and which rest on reasoning you're entitled to argue with.
Eight moats, each scored 0–3, producing a defensibility position out of 24. Built on Hamilton Helmer's 7 Powers and Gokul Rajaram's eight-moat frame, extended into a repeatable instrument.
Every moat has a way it stops working. That's the argument for scoring all eight rather than defending the one you're proudest of — the failure modes are independent, so depth in a single moat is one bad quarter away from being nothing.
The scale runs on stack depth rather than peak strength, because single moats keep failing — the data advantage a competitor buys, the regulatory position that gets deregulated, the distribution deal that ends. Depth in one moat is a wall; several reinforcing each other is a position. Early on, Exposed is the expected position rather than a warning — what to read at that stage is which two moats the model puts within reach. No single moat is durable — stack them →
Scores across all eight moats with a position and a written rationale for each. Methodology-light — you get the assessment, not the machinery behind it. No cost and no call: submit your company and I run the same instrument used on every scorecard in the published library below.
The rigorous version, with the methodology visible. Every score pressure-tested through five lenses, trajectory analysis on whether each moat is widening or eroding, two diagnostic flags, and head-to-head comparison against two to four named competitors.
A score is an opinion until it survives something. Each moat is tested against all five.
Judge the analysis before you buy it. Every teardown below uses the same instrument, the same rubric, and the same four positions — so you can check whether the reasoning holds on a company you already have an opinion about. Each one was published and argued in public.
Click any company below and I'll email you its full scorecard — next business day.
Public analyses. Not client work. The spread is the point — an instrument returning Fortified on everything would tell you nothing. See all 15 scorecard summaries →
A moat around a market nobody wants is an expensive wall. Diagnostics tell you whether what you've built holds; research tells you whether it's worth holding.
Most engagements start with research and use a diagnostic to pressure-test what it finds. If you're not sure which end to start from, that's a reasonable thing to spend fifteen minutes on.