From noise to a named human being with a date.

The job

Establish ground truth before opinion enters the room. Metrics exist so that the RAG status is computed, not debated.

The standardFive properties every metric must carry

1. A number. Not a direction, not a vibe.
2. A target or baseline. A number without a comparison is trivia.
3. A trend. Movement beats level. A green metric heading down fast is more urgent than a red metric recovering.
4. A threshold. Pre-agreed bands that set RAG automatically. If you're arguing about whether a domain is amber or red, you set the thresholds too late.
5. A freshness stamp. A metric older than the review cycle isn't green — it's grey. Unknown is a status, and it should be an uncomfortable one.

Technique1 lagging + 2 leading per domain

Most domains die of metric bloat. Cap it hard:

1 lagging metric — the outcome the domain is accountable for. Slow, honest, unarguable.
2 leading metrics — the inputs that cause the lagging one 2–8 weeks out. These are what you can actually act on this week.

If you only ever review the lagging metric, you are steering with the rear-view mirror and the review is theatre.

TechniquePre-commit the thresholds

Write them once, out of session:

Delivery — On-time %. Green ≥ 90. Amber 80–89. Red < 80. Or red regardless if trend is down two consecutive weeks.

That second clause is the one everyone skips. Level and velocity both trigger.

Watch forFailure modes

• Vanity metrics. Numbers that only go up. If a metric has never been red, it isn't measuring anything.
• Metric bloat. Fifteen KPIs per domain means no one knows which one matters. You have optimised for looking thorough.
• Retrospective threshold-setting. Deciding the number is "fine, actually" after seeing it. This is where honesty dies.
• Silent staleness. A green card sitting on data from six weeks ago.

The job

Triage. From everything the metrics say, surface the one deviation that most deserves the next four minutes.

Not a list. One. The discipline of picking one is the value.

TechniqueRank by Materiality × Velocity × Reversibility

When several things are wrong, choose using three axes:

1. Materiality — how much does it cost if it plays out fully?
2. Velocity — how fast is it moving? A slow leak can wait a week. A fast one can't.
3. Reversibility — can we recover if we let it run? Irreversible damage jumps the queue regardless of size.

A small, fast, irreversible problem beats a large, slow, recoverable one. Most leaders get this backwards because size is the most visible axis.

TechniqueClassify the shape — blip, trend, or break

Acting on blips is how leadership teams generate churn and lose credibility. Naming something a blip is a legitimate, valuable alert output — it explicitly buys the domain a week of peace.

TechniqueThe null alert

"No alert. Metrics are within band, trend stable, data fresh."

This must be an allowed and respected outcome. If every domain must produce an alert, you have built an anxiety machine and your team will manufacture problems to look engaged. A healthy Monday MAX has 3–6 domains with no alert.

Watch forFailure modes

• The laundry list. Everything is an alert, therefore nothing is.
• Alert as complaint. "Marketing isn't giving us leads" — that's an accusation, not a deviation. No number, no date, no consequence.
• The perennial alert. The same alert every week for two months means the Execute step is broken, not the alert.
• Green-washing. The metric is red but the alert is soft because the owner is in the room. Handle this with pre-set thresholds (see M).

The job

Get from symptom to mechanism. This is where the drill earns its name and where 80% of reviews fail — because it's the only step that requires thinking rather than reporting.

TechniqueThe four-cause taxonomy

Before you 5-Whys anything, classify the cause. The class determines the fix, and misclassifying it is why the same problem returns.

The most expensive mistake in management is treating a Structure problem as a Skill or Will problem. You retrain, you exhort, you replace the person — and the system produces the identical failure through the next occupant of the chair. Conversely, papering over a Will problem with new process is how organisations accumulate bureaucracy: every rule in a mature company is a scar from someone who wouldn't, treated as though they couldn't.

Ask: "If I swapped this person for the best person I know, would the outcome change?"
No → Structure.
Yes → Skill or Will. Then ask: "Do they know how, and have they got what they need?" Yes → Will. No → Skill.

TechniqueIs this domain the cause, or the victim?

Domains contaminate each other. This is precisely the Integration term in your equation — most red domains are downstream of a red domain nobody looked at.

Cash is red
← because Delivery is late (revenue recognition slips)
← because People is under-resourced (two open roles at 90 days)
← because Leadership hasn't approved the headcount.

Four red cards. One root. If you MAXED each of those four domains independently, you'd generate four action sets, three of which are treating symptoms — and you'd burn capacity you don't have.

So the X-Ray asks two questions, not one:

1. What is the mechanism inside this domain?
2. Is this domain's red actually the shadow of another domain's red?

Mark it explicitly — a contamination link from a domain's X-Ray to its upstream domain. Over time the heat grid stops being ten independent columns and starts revealing the causal spine of your business.

TechniqueFalsify before you commit

Once you have a candidate cause, ask:

"If this cause were true, what else would I expect to see? Do I see it?"

Candidate: "Churn is up because onboarding got worse."
Prediction: "Then churn should be concentrated in cohorts onboarded after March, and NPS at day 30 should have dropped."
Check it. If churn is evenly distributed across cohorts, your cause is wrong — and you were about to spend a quarter rebuilding onboarding.

Ninety seconds of falsification saves a quarter of misdirected work. This is the highest-ROI habit in the entire drill.

TechniqueBounded 5-Whys

Five Whys is fine but it runs away — three iterations, then stop. Anything deeper than three usually lands on "because of our culture," which is unactionable and lets everyone off the hook.

Watch forFailure modes

• Symptom re-statement. "It's late because it's behind schedule."
• Person-blame as a cause. Almost always a Structure problem in disguise.
• Abstraction escape. "Culture," "communication," "alignment." Unfalsifiable, unfixable, comfortable.
• Cause shopping. Picking the cause whose fix you already wanted to do.
• Solving in the X. The room jumps to solutions before the mechanism is agreed. Enforce the wall: no solutions until the mechanism is stated aloud and accepted.

The job

Convert the mechanism into a committed decision. Note the word: decision, not action. Several perfectly good decisions produce no action at all — and that's the part almost every review framework gets wrong.

The standardThe six decision types

Force the room to name which one it is. Naming the type is what stops the drift.

Accept and Observe are first-class outcomes. Most weekly reviews fail because the implicit rule is "every red must generate an action." So you generate ten actions a week, complete four, and carry six. Within a quarter you have sixty open actions, credibility is gone, and the ritual becomes something people survive rather than use.

An explicit Accept is a leadership act. "Yes, support response time is worse than we'd like. We are choosing that trade-off until October to protect the migration. Domain is green because we are performing to intent." That's a decision. The domain isn't red — it's deliberate.

An explicit Observe needs its trip-wire, or it's procrastination with a job title:

"Observe. If debtor days exceed 60, or if it's still above 45 on 1 September, this converts to Act automatically."

The standardThe action standard — four fields, non-negotiable

An action without all four is a wish:

1. Owner — a person. Not a team. Not "we." Teams don't get out of bed.
2. Date — a specific day. Not "next month," not "ASAP."
3. First step — the concrete next physical act. "Draft the change-control gate" is a project. "Book 30 minutes with Priya on Thursday to agree the gate criteria" is a first step. Actions without a first step are the ones that don't start.
4. Definition of done — how we'll know, in one observable sentence.

TechniqueSize to the cadence

The review is weekly. Therefore an action must be completable, or meaningfully advanceable, in a week.

If it can't be — it's not an action, it's a project. It goes to the Priorities list (top 5 company / top 3 CEO), not into the domain drill. Mixing week-scale actions with quarter-scale projects in the same list is the single fastest way to make a review list untrustworthy.

TechniqueCheck the total, not the item

Each action looks reasonable in isolation. Ten domains × two actions = twenty new actions, in a week, on top of the actual job. Before closing the session, look at the aggregate:

"We've created 14 new actions. Last week we completed 6 of 11. We are writing cheques the organisation cannot cash."

Then cut. Ruthlessly. The willingness to leave a red domain with zero actions this week — because capacity is spent elsewhere — is the mark of an operator rather than a list-maker. Say it out loud: "Delivery stays red. We're not acting on it this week. Cash is the priority. We'll re-look Monday."

That's honest. Carrying a fake action is not.

TechniqueTwo-way vs one-way doors

Two-way door (reversible): decide fast, decide in the room, don't gold-plate. Speed is the value.

One-way door (irreversible — a hire, a fire, a repricing, a public commitment, a platform migration): do not decide inside a 5-minute domain slot. The decision type is Escalate or Invest, with a proper decision process outside the review.

Watch for one-way doors being casually walked through because the room had momentum. The drill's tempo is a feature for reversible calls and a hazard for irreversible ones.

Watch forFailure modes

• Discussion mistaken for decision. Everyone leaves feeling aligned; nothing was decided.
• The action list as anxiety relief. Creating actions feels like control. It isn't.
• Zombie actions. The same action rolls forward four weeks. That's not an action problem, it's a Delegate problem or an Accept-in-denial.
• The heroic verb. "Improve," "review," "look into," "align on." All of these mean nobody knows what to do on Tuesday morning.

The job

Transfer ownership with authority — not a task, an outcome. This is the step that determines whether MAX10 scales you or traps you.

The diagnostic metricCEO Action Share

% of this week's actions owned by you.

Track it every session. It is, quietly, the single most revealing number the whole system produces.

Plot it alongside RAG. A business whose domains are going green while CEO Action Share stays at 60% is not healthy — it's load-bearing on one person, and it will regress the moment that person takes leave.

TechniqueState the authority level explicitly

Ambiguous authority is the number one cause of stalled delegation. Name the level, every time:

Most delegation failure is a level mismatch: you think you gave them 4, they heard 2, so they wait for you — and you conclude they lack initiative. Say the number.

The trajectory of a domain owner from level 2 → level 5 over a year is your leadership development plan, and it's more honest than any competency framework.

TechniqueThe four confirmations

Before the domain closes, the owner should be able to state, unprompted:

1. The outcome — what "done" looks like, in their own words. (If they can only recite the task, ownership hasn't transferred.)
2. Their authority level — the number above.
3. Their resources — what they have, and what they're allowed to spend or commit.
4. Their trip-wire — the condition under which they must come back to you early, not at the next review. ("If it slips past Thursday, or if it needs more than $2k, you escalate immediately — don't wait for Monday.")

The trip-wire is what makes level 4 and 5 delegation safe. Without it, delegation is either abdication or micromanagement, and leaders oscillate between the two because they've never defined the middle.

TechniqueThe monkey check

Whose calendar does this now live on?

If, at the end of the domain, the next physical action sits on your calendar — the monkey jumped back onto your shoulder. It happens through phrases like "leave it with me," "I'll just have a quick word with them," "let me look at it first." Each one is a delegation reversal disguised as helpfulness.

Ask it out loud, at the close of every domain: "Whose calendar?"

Watch forFailure modes

• Delegating the task, keeping the metric. They do the work; you stay accountable; nobody grows.
• Team ownership. "Ops will handle it." Ops will not handle it.
• Authority mismatch. Responsibility without decision rights. This is the fastest way to burn out a good operator.
• The silent re-take. You quietly do it yourself because it's faster. It is faster — once. Then it's yours forever.
• No trip-wire. Discovering on the following Monday that it went wrong on Tuesday.

The MAXED-out definition of done

A domain is MAXED out when it passes all five gates:

• RAG is derivable from the numbers by someone who isn't you. Data is fresh.
• One alert (or an explicit no-alert), stated as a deviation with a consequence and a date.
• Root cause stated as a mechanism, classified (Signal / Structure / Skill / Will), and checked for upstream contamination.
• A named decision type. Any actions carry owner + date + first step + done-condition. Aggregate load is within capacity.
• Named human owner, explicit authority level, resources confirmed, trip-wire set. Monkey check passed.