Add three field principles on ownership, standards and correction

CHANGELOG.md

@@ -52,6 +52,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   signal per input, cross-checked against the chokepoint list) and a check that the
   origin record still exists and is being read before playbook training.
 
+- **`docs/principles-from-the-field.md`** — added three more field principles with worked
+  go-to-market examples. Under **Ownership:** *when a team underperforms, check who owns
+  what before you check who's good* (most of what reads as a talent problem is ownership
+  geometry — a capable person whose boundary is two millimetres wrong destabilizes the
+  people around them, and the assembly, not any part, is the unit of strength). Under
+  **Daily discipline:** *a rule you keep having to remind people about will get dropped —
+  put it in the workflow instead* (distance from the workflow, not size, is what makes a
+  standard heavy; ask who pays the cost each time it's applied, and use the post-change
+  reminder count to check the artifact isn't decorative). Under **Character & judgment:**
+  *score who changed after feedback, not who agreed with it* (names the selective hearer
+  and the compliant non-changer, and why interviews reward the second). Refreshed the
+  "Where these land in the kit" map with all three as candidate updates.
 - **`docs/principles-from-the-field.md`** — added four more field principles with worked
   go-to-market examples, in two new sections. **Dependencies:** *name the few chokepoints
   your revenue flows through, and who controls each one* (a short list of narrow

docs/principles-from-the-field.md

@@ -13,6 +13,9 @@ Where a principle maps onto a specific part of this kit, that's noted at the end
 **The ownership gap — proximity is not ownership; close the distance.** The people closest to a motion often feel *less* ownership of it than an outsider feels for something trivial. A sports fan who never touched the field says "*we* won" and means it; meanwhile the person whose work actually moves the number talks about "the company's goal" as if watching from the stands. That inversion is expensive — a team of spectators executes a fraction of what the same team executes when each person treats the result as theirs. Two moves close it. First, make each person's part **legibly load-bearing**: a system has no decorative parts, and you only notice the small component when it fails and everything stops — so name, per person, the specific thing that breaks in the motion if their piece slips. Second, when someone knows *what* to do but isn't doing it, check whether it's a competence gap or an ownership gap — corrections don't touch an ownership gap; reconnecting the person to why their part moves the number does.
 **Example:** A rep who is "executing the playbook" but flat is rarely fixed by another playbook correction. The move that works is making their slice of the number legibly theirs — *this segment is yours; if it stalls, here's exactly what the company doesn't hit* — then getting out of the way.
 
+**When a team underperforms, check who owns what before you check who's good.** The reflex when a function stops producing is to assess the people — are they strong enough, motivated enough, do they need coaching. That reflex is aimed at the wrong variable most of the time. The higher-yield question is *geometry*: is each responsibility defined precisely enough that no work falls between two people, and is anyone carrying something that belongs to a stronger part of the system? A capable person whose boundary is slightly wrong doesn't just underperform — they destabilize the people around them, which reads as a team problem and gets treated as a talent problem. The human knee is the clean version of both halves: it survives more than ten times the load its strongest single ligament can hold, because about fifty parts share that load in a specific arrangement — so the assembly, not any part, is the unit of strength, and an excellent operator working detached from the system produces individual output rather than system output. And when surgeons rebuild one, most failures trace not to weak tissue but to a healthy graft positioned about two millimetres off: too tight and the joint stiffens, too loose and it gives way. Two millimetres of overlapping or missing ownership does the same thing to a team, and it's invisible in a performance review because everyone involved is competent.
+**Example:** A pipeline stalls and the manager starts coaching the reps. The actual defect is that nobody owns the handoff between qualification and demo — it's *nearly* owned by two people, so each assumes the other has it. Mapping every responsibility to exactly one name, and treating anything with two names or zero names as the defect, fixes in an afternoon what a quarter of coaching didn't touch.
+
 ---
 
 ## Alignment
@@ -30,6 +33,9 @@ Where a principle maps onto a specific part of this kit, that's noted at the end
 **A deal "waiting on them" still has work on your side — close all of it.** When a deal or decision sits with the other party — their board, their procurement, their budget cycle — the waiting is real, but it quietly expands: "waiting to hear back" grows from the one genuinely blocked item to ten things that never were, and the whole motion around that account pauses. The discipline is a *your-side checklist* per wait-state item, driven to zero: references pre-briefed, the security questionnaire pre-answered, the contract template pre-redlined, the implementation plan drafted, the champion armed for the internal sell. Two payoffs: when their door opens, you close in days instead of re-mobilizing for weeks — and the record shows the delay was never yours, which matters more than it sounds when the relationship hits its next hard moment.
 **Example:** Two deals get a verbal yes on the same Friday. One team spent the wait-state weeks "monitoring"; kickoff takes a month of scrambling. The other drove its your-side list to zero during the wait; the paperwork goes out Monday. Same yes, four weeks of difference — decided entirely during the period that looked idle.
 
+**A rule you keep having to remind people about will get dropped — put it in the workflow instead.** Standards behave like a weight: held at arm's length a light one is punishing within a minute; worn against the body the same weight is barely noticed. Distance, not size, is what makes a standard heavy. A quality bar that lives in someone's head and gets applied by reminder is held at arm's length — every application costs effort, the effort is paid by whoever does the reminding, and the bar quietly goes the first week that person is busy. The same bar encoded in the template everyone starts from, the definition-of-done in the ticket, or a field the system won't let you skip costs almost nothing per use, because it's no longer a separate act of will. This is why teams abandon standards they still agree with: they don't drop the ones they dispute, they drop the ones that are expensive to apply — and the ambient pressure runs one way, because lowering a bar is immediately cheaper for everyone. Your job isn't to have more willpower than that pressure; it's to move each standard from arm's length into the system. The diagnostic is one question per standard: **who pays the cost each time it's applied?** A person remembering means it's decaying. A system means it's held.
+**Example:** A team keeps being told that no deal advances without a named economic buyer. It holds for three weeks and slips. Making the field required to advance the stage ends the reminders permanently — and if the reminders *don't* stop after the change, the artifact is decorative and the standard still isn't held.
+
 ---
 
 ## Decisions
@@ -102,6 +108,9 @@ Where a principle maps onto a specific part of this kit, that's noted at the end
 **Hire for who holds up under pressure, not who looks polished — then give them time before you judge them.** When you pick a hire, a bet, or an account to pour effort into, don't buy the polished surface — the slick résumé, the strong first call — buy the trait that survives a hard quarter: composure under pressure and willingness to be corrected. A rough-but-durable pick in a good process beats a polished-but-brittle one, because the value gets added by the work you put in afterward, not by how finished the pick looked on day one. And that work is slow and mostly invisible, so you'll be tempted to write someone (or something) off on a bad snapshot and quit too early — don't grade a half-finished thing as if it were done, and match your effort to the stage: heavy correction early, fine adjustments later, and the wrong one at the wrong time does damage. (The image behind it: the most famous marble statue was carved from a block two earlier sculptors had rejected as flawed and thrown out — the result came from the carver, not the stone.)
 **Example:** The candidate with the rougher résumé but real coachability and composure under pressure outperforms the polished one who cracks the first hard quarter. And the account that looks like a write-off at week six is often just early in the work, not failed — judged on the snapshot it gets cut; judged on where it's heading it becomes the reference.
 
+**Score who *changed* after feedback, not who agreed with it.** Two people take identical feedback. One hears a description of what they must change and does something about it inside a week. The other hears a description of a problem elsewhere, agrees warmly, and changes nothing. Both look equally engaged in the room, and the gap between them compounds quietly until a bad quarter makes it visible all at once. Track the behavior, not the reception — experience is a stock, but the willingness to act on correction is the *flow rate*, the only mechanism by which an operator's judgment updates as fast as the market moves. Two failure modes are worth naming because they look different and cost the same. The **selective hearer** comes out of a review focused on the part they already do well and files the corrective part under what someone else needed. The **compliant non-changer** accepts every point, argues with none, and produces no behavioral difference — the more dangerous of the two, because it defeats the usual detection method, which is watching for resistance. This also explains why interviews are bad at reading coachability: they reward the second type. The only reliable read is longitudinal — give one real piece of correction on a real call, then look for a changed behavior inside a defined window.
+**Example:** A manager's review notes say the same thing about a rep three cycles running, and each time the rep took it well. Keeping one line per person per cycle — *the specific behavior that changed because of last cycle's feedback* — turns "he's coachable" into something checkable, and two consecutive blanks becomes a finding rather than a feeling. Run the same row against your own name first; an empty one usually means nobody in the org is correcting the operator.
+
 ---
 
 ## Attention & inputs
@@ -149,6 +158,9 @@ Several are now wired into the kit's rituals (below); treat the rest as candidat
 - **Guard inputs / hold the vision** → the [`pre-compact`](../.claude/hooks/pre-compact.sh) hook already re-injects priorities and the latest log so a long session doesn't lose the thread — the same instinct, applied to your own attention: curate the inputs, and re-state the longer-term thesis on a cadence.
 - **A problem going quiet is not the same as a problem being fixed** → *candidate* for your close-out ritual ([`/end-of-day`](../.claude/commands/end-of-day.md)) and [`/reconcile`](../.claude/commands/reconcile.md): before anything is marked resolved or green, require a one-line *root* — what actually changed — so a quieted symptom isn't logged as a fixed one. Pairs with the deal-and-account status fields in `ops/`.
 - **Hear both sides before you assign fault** → *candidate* for whatever review reads win/loss and escalations ([`/weekly-review`](../.claude/commands/weekly-review.md)): put both accounts on the record — the direction/standard side and the execution side, or the account team's and the customer's — before attributing a miss.
+- **Check ownership geometry before you check the people** → *candidate* for [`/weekly-review`](../.claude/commands/weekly-review.md), extending the ownership read: before assessing anyone in an underperforming area, map every responsibility in it to exactly one name, and treat two names or zero names as the defect. In `ops/`, the two-millimetre errors are the ones where a responsibility is *nearly* right — which is why they survive a review.
+- **Put the standard in the workflow, not in a reminder** → *candidate* for your stage definitions in [`ops/pipeline.md`](../ops/pipeline.md) and any rule in [`.claude/rules/`](../.claude/rules/README.md): a standard you restate more than twice in a quarter belongs in an artifact — a required field, a template section, a gate that won't advance — and the reminder count afterward is how you check the artifact isn't decorative.
+- **Score the change, not the agreement** → *candidate* for whatever cadence you use for coaching and for [`/weekly-review`](../.claude/commands/weekly-review.md): one line per person per cycle naming the behavior that changed since the last one, with two consecutive blanks treated as a finding. Run it against your own name first.
 - **Removing the blocker doesn't remove the habit** → *candidate* for any process, tool, or rule change you roll out: name the old habit it replaces and a re-education plan (cadence + owner), and set an observed *behaviour* — not the switch — as the done signal. Relevant to adoption tracking in `ops/customers.md`.
 - **Two teams stop fighting when both own the same number** → *candidate* for [`/weekly-review`](../.claude/commands/weekly-review.md) and the cross-function loop in `ops/feedback-log.md`: before mediating a sales↔marketing or product↔success handoff, put both sides on the *same* number rather than refereeing the interface. Pairs with the one-number idea in [`docs/operating-model.md`](operating-model.md).
 - **A last-contact list of every account, oldest first** → *candidate* for [`/weekly-review`](../.claude/commands/weekly-review.md) and the account book in `ops/customers.md`: keep a last-contact date on every tier-one relationship and work the oldest first — the same stall discipline `ops/pipeline.md` applies to deals, since silence precedes the churn metric.