Lesson 0050 · Initiative Lead · Module 3

Steer, Replan, and Close the Initiative

Leadership is keeping the truth cheap to see. Review evidence while options remain, change the plan without disguising the change, and finish the operational and organizational work after launch.

Reading ≈ 9 minPractice ≈ 90 min

Mission tie-in: a system is successful when it changes the intended outcome and remains responsibly owned. Shipping is a transition in that lifecycle, not the finish line.

Knowledge: run one evidence loop

Use a weekly initiative review built from live sources, not a status performance. Product, design, and engineering answer seven questions:

Outcome: what does the latest measure say, by relevant cohort?
Learning: which assumption became stronger, weaker, or false?
Delivered evidence: what integrated behavior can now be demonstrated?
Flow: what is aging or blocked, and where should the team swarm?
Risk and quality: which driver, guardrail, dependency, or operational signal changed?
Decisions: what was decided, what is waiting, and who owns the next evidence?
Forecast: what range is credible now, and what scope trade preserves the outcome?

One-page initiative review · returns pilot, week 4

Outcome: Median 3.2 hours for eligible cohort; support contacts down 18% versus 25% target.
Guardrails: Fraud unchanged; manual queue 34/day under limit; mobile completion is 11 points lower.
Learned: Status visibility drives fewer contacts; label speed is not the limiting factor.
Changed: Move mobile comprehension work ahead of exchange automation; keep carrier design.
Risk: Guest orders are 29% of demand and remain outside pilot; product decision due Friday.
Forecast: Core pilot outcome likely in two weeks; exchange variation removed from this initiative.
Ask: Approve two design sessions with customer-support participants; no staffing change.

The principle behind the principle

A weekly evidence loop is not about discipline; it is about preserving options. The cost of changing a plan grows with the work already committed to the old one, while the value of information is highest when the most options remain. A week is not sacred—it is chosen because it is short enough that at most one week of work has been committed against a belief that turns out to be false, and long enough that at least one slice's evidence has actually arrived. Pick the interval from those two quantities and you will sometimes get three days and sometimes get a fortnight.

Pre-agreed continue/change/stop criteria work on a different mechanism: they remove the identity cost of stopping. A decision to stop made in advance is a policy; the same decision made in the moment is a verdict on the people who proposed the work. That asymmetry, not incompetence, is why teams that never pre-commit almost never stop—and why the criteria have to be written while everyone still expects to succeed.

The loop degrades into theatre in two conditions. When the outcome measure cannot move weekly—a quarterly conversion figure, a retention curve—steering on it produces noise interpreted as signal; name a leading indicator, and label it explicitly as a proxy so nobody mistakes it for the outcome. And when the review has no authority to change scope, it is a status meeting wearing a review's clothes. If nothing on the page can be changed by the people reading it, delete the page and send an email.

Make a change explicit across all affected artifacts. If guest checkout enters scope, update the frame, story map, threats, contracts, telemetry, rollout, and forecast. “We are still on plan” after the plan changed is how teams exchange trust for temporary comfort.

Pre-agree continue, change, and stop criteria. Continue when outcome movement and guardrails support the hypothesis. Change the solution or slice when the opportunity remains but evidence rejects the mechanism. Stop when the opportunity is too small, the intervention cannot move it economically, a guardrail cannot be protected, or strategy has changed. Stopping after cheap evidence is capital allocation, not team failure.

After release, close the initiative deliberately:

Transfer service ownership, on-call, runbooks, access, dashboards, support knowledge, and vendor responsibility to a named steady-state owner.
Remove release flags, temporary adapters, manual pilot paths, obsolete metrics, elevated access, and test data—or record each as debt with a trigger from lesson 0038.
Supersede ADRs whose context changed; update context/container views, contract docs, and architecture fitness functions.
Report the outcome against baseline and guardrails, including segments that did not benefit and costs that moved elsewhere.
Hold a learning review: expected, observed, surprised, decision for next time. Turn at most one or two insights into owned system changes.

A retrospective about feelings can improve teamwork; an initiative review must also improve the organization's memory. Preserve decisions and evidence, not a victory narrative. Future leads need to know which assumptions failed and which shortcuts still charge interest.

How the lead actually read week four

The numbers above are good enough to be dangerous. Median 3.2 hours beats the under-four-hour target. Support contacts are down 18% against a 25% target, which invites the most comfortable reading available: nearly there, keep going, it will close.

The lead does not look at the headline first; she looks at the guardrails, because that is where a working system hides its casualties. Mobile completion is eleven points below desktop. So she asks the question that separates a lead from a reporter: is the 18% a smaller version of the same effect, or a different effect? Splitting the contact reduction by device answers it—desktop is down 27%, mobile down 6%. The mechanism works. The surface does not. Averaged together, those two facts look like partial success; separated, they are one success and one defect, and they need opposite responses.

The obvious move is to add a comprehension workstream alongside the exchange automation already in flight. The lead rejects it on WIP grounds: two parallel bets on a seven-person team means neither produces clean evidence, and the team has just demonstrated how easily two effects hide inside one number. Exchange automation is dropped from the initiative and recorded as a non-goal with a revisit trigger, not quietly deprioritized where it can drift back.

Marta dissents, for a real reason rather than a bad one: the exchange slice was promised to a stakeholder. The lead does not overrule the relationship; she makes the trade explicit in the “Ask” line of the review, so the sponsor chooses between the promise and the outcome with the same information the team has. That is the honest version of scope negotiation—the alternative is deciding on the sponsor's behalf and calling it protecting them.

Two weeks later, support contacts are down 24% with no new mechanism at all, only mobile comprehension work: better refusal copy, a status page that survives a small screen, and one fewer step before the label. The exchange slice never came back, and nobody asked for it—which is itself evidence about the original scope, and belongs in the learning review rather than in a quiet sense of relief.

Field notes · steering under pressure

Pressure response	Damage	Leadership move
hide forecast until certain	Options disappear while confidence grows.	Share range, cause, options early.
add people late	Coordination load rises immediately.	Trade scope; remove blockers.
declare launch a success	Outcome and cleanup become orphaned.	Keep review through stabilization.
skip the closeout	Temporary machinery becomes permanent.	Assign removal and ownership dates.

What people get wrong after reading this

Stopping means the initiative failed.: The failure is spending the remaining budget after the evidence has arrived. Stopping cheaply is the return on discovery—and a team that has stopped one thing well is trusted with larger bets, not smaller ones.
The weekly review is for stakeholders.: It is the team's steering instrument. If the only thing it ever changes is a slide, it has become reporting, and reporting is cheaper to do asynchronously.
Closeout is documentation.: Closeout is the transfer of ownership and the removal of machinery that only made sense during the pilot: the flag, the manual queue tooling, the fake adapter, the elevated support access. Documentation is the residue of that work, not the work.
A green outcome metric closes the initiative.: Not until someone else can operate the service at three in the morning, and every temporary path is either gone or on a dated ledger with an owner. Otherwise the initiative has not ended; it has just stopped being anyone's job.

When this is the wrong move

A weekly review is overhead on a two-week effort—one mid-point check is the whole loop. A formal closeout is waste when nothing temporary was built and the team that built it is the team that will run it; in that case the only real item is the debt ledger. Pre-agreed stop criteria are meaningless when the initiative is a legal or regulatory mandate: the variable there is scope and sequence, not existence, so write change criteria and skip the stop column rather than pretending the option exists. And a learning review held while the outcome is still moving will mostly capture relief; wait for the number to settle, then hold it while people still remember what they believed.

Skill: steer with evidence

The solution fails but the opportunity remains. Choose:

When should forecast risk be shared?

An initiative is operationally closed when:

Capstone: lead a simulated initiative

Choose a real or plausible product brief. In a 90-minute working session, produce the complete leadership packet: initiative frame, assumption map, current workflow, story-map slice, five architecture drivers, three-view architecture packet, one state/contract review, one ADR, walking skeleton, delivery/dependency map, team charter, and rollout contract. Then write the week-four review after inventing one adverse result. The templates live in the Initiative Playbook.

In the lab and the capstone. The lab's phase-4 checkpoint ends the way an initiative does: the telemetry you added has to say something, the rollout gate has to pass or fail on its own, and the temporary scaffolding has to be removed or written down. The worked artifacts are the week-four review and the closeout in the strong capstone packet—note that the review changes the plan on its own page rather than announcing that everything is on track. Compare with the weak version's review and closeout: a status summary with no segment, no ask, no dissent, and a closeout that transfers nothing.

Reveal: how to review the capstone

Trace one line through every artifact: outcome → assumption → user slice → architecture driver → boundary/contract → decision → delivery milestone → test/telemetry → rollout gate → continue/change/stop criterion. Any broken link is either unnecessary work or an unmanaged risk. A strong packet is coherent, not large.

Teach it back

In two sentences, explain to a new lead why stop criteria have to be written before anyone believes they will be needed.
Take the last initiative you were on. Name the week the evidence was already sufficient to change the plan, and name what was actually done that week instead.
Predict the failure: if a team never closes out, what is still running in a quarter, who is quietly maintaining it, and what does the next initiative inherit?

Your win

You can lead a new initiative from ambiguous brief through discovery, system design, team execution, controlled release, evidence-based steering, and responsible operational closure.

Read and watch deeper

Thinking in Systems, Donella Meadows — feedback, delays, and why local delivery measures can misrepresent system outcomes.
Accelerate, Forsgren, Humble & Kim — delivery performance and continuous improvement grounded in evidence.
The DevOps Handbook, Kim et al. — feedback, telemetry, learning, and operational ownership across the value stream.
Escaping the Build Trap, Melissa Perri — ch. 1–4 and 13: outcome over output, and why organizations keep shipping after the evidence has arrived.
Reference: Initiative Playbook · review and closeout, and the debt ledger.

Use your agent-teacher as a hostile reviewer of artifact coherence: it should ask what evidence connects each decision to the outcome, never manufacture the evidence itself.

Mark lesson complete