Chapter 33: Design Hybrid Delivery Intentionally

The meeting where every number was true

It is the monthly steering committee at Meridian, a month after wave two opened under the conditions chapter 24 set, and the agenda’s one item is the December window for clinics four, five, and six. Every number on the wall is true. None of them can be added.

The construction superintendent reports first, because the contractor has a fixed drawing set, a liquidated-damages date, and the habit of speaking first. Clinic four is at seventy-four percent overall by the contract’s progress rules. The structure is complete, the services are in, the fit-out is underway, and the certificate of occupancy is forecast for mid-November, about eight weeks out, subject to the inspectors’ calendar.

Marcus Chen, the platform lead, reports next. The release candidate for the network-wide capability is green on the integration test, and the clinic-four configuration is nine items in the backlog. The team runs two-week cycles at three items a cycle, so the configuration can ride the release after the one scheduled for the third Friday.

The migration lead reads her numbers without looking up. Sixty-two percent of the legacy records are migrated, about forty-one thousand of sixty-six thousand, at a reconciliation rate of two thousand records a week. The remaining twenty-five thousand clear in about twelve and a half weeks. The archive backlog at clinics four and five is the slow part, and the reference tables are still gated by the privacy review.

Sam Otieno has the training numbers. Three of the six cohorts for the final wave are complete, the super-user roster for clinic four is named, and two of the seven are double-booked with the existing-clinic rollout, the capacity question he raised in the corridor last month.

Esther Njeri closes the stream reports. The privacy assessment for the new sites is scheduled, subject to the sub-processor amendment that chapter 24 made a condition, and the certification body’s calendar is the constraint, not the project’s.

The finance director asks the question that turns all the true numbers into an answer no one can give: “Can clinic four open in December?”

The contractor’s seventy-four percent cannot be added to Marcus’s nine items or to the migration lead’s forty-one thousand records. Three of six cohorts is not a number that translates into a gate date without an assumption. “Assessment scheduled” is a row with no date. The room does what rooms do when the numbers cannot be added: it turns to Dana Okafor, because Dana is the person who has, for eighteen months, carried the translations in her head. She knows, or believes she knows, that the migration is the long pole; that the certification body’s calendar will slip before the contractor’s will; that Sam’s double-booked super-users will surface at the training week; and that the December window closes against the grant’s month-thirty-six line, which the finance director has now said aloud. She can answer the question. That is precisely the problem.

This chapter is about the difference between being the person who can answer and designing a system in which the answer does not depend on one person’s head. Chapter 13 chose the hybrid deliberately at Meridian: construction predictive, workflows iterative, the rollout incremental, the platform adaptive inside gated releases, and it named the seams and the seam discipline that would carry them. Chapters 31 and 32 ran the predictive clock at BlueLine and the adaptive clock at KijaniPay, and both closed pointing here, to the world where the clocks multiply and must be coordinated as one intentional architecture. What the book has not yet named is the thing that happens between the choice and the architecture: the hybrid that is designed on paper and coordinated by heroics in fact. The certification gap of chapter 24 was found one week before the wave-two gate, by a pre-gate review the calendar happened to schedule. The training gap of chapter 11 was found after go-live, by a training lead walking the floor. The wave-two gate was resolved by a steering decision under pressure, well made, on the record, and it should not have needed the pressure. Each seam has been resolved when it broke. The hybrid exists. Nobody designed the coordination.

The thesis of the chapter, in one sentence: hybrid delivery is the normal state of complex work, and the mastery is designing it intentionally, which means naming the unit being hybridized, owning every seam with a named accountable person, synchronizing the streams on shared milestones and integration evidence, translating between the clocks with stated rules and spoken assumptions, routing change through the path its reach requires, funding and reporting the mix honestly, refusing the worst-of-both-worlds shapes, and re-tailoring, with a record, when the seams fail, because the architecture is a model and the evidence updates it. The decision this chapter improves is the one the finance director’s question forces: not whether the December gate holds, but whether the project coordinates its own mixed methods deliberately, or leaves the coordination to the heroics of the person who happens to be in the room.

Hybrid is the default, and the default is accidental

The first fact the project leader needs is that hybrid is not a fifth brand and not a compromise between two pure ways of working. It is the normal state of any project large enough to have more than one kind of work in it. The pure predictive project, where every component, phase, team, and contract runs on one baseline with one gate cadence, is an abstract corner, rare outside a single-contractor build with a frozen design. The pure adaptive project, where every component is discovered in use with no predictive stream anywhere, is an equally abstract corner, rare outside a small software team with no regulatory interface and no physical asset. Everything between, and most real projects are between, is a hybrid, because the drivers that push work toward one clock or another do not align by project. They align by component.

Chapter 13’s map made the mechanism visible: requirements uncertainty and solution uncertainty vary within a project, not only between projects. A clinic building has low requirements uncertainty and low solution uncertainty, so it runs predictively, because the drawing set is knowable and the change is expensive. A clinical workflow has high requirements uncertainty, so it runs iteratively, because the definition of good emerges from use in a live consult. A platform integration has high solution uncertainty, so it runs adaptively, because nobody has configured this record system for this referral flow before. The same project sits in several corners of the map at once, and that is a fact about the work, not a failure of design. The failure is the inference drawn from the fact: because the project is one project, the organization assumes one way of working, one calendar, one report, one baseline, and the work that does not fit the assumption runs anyway, unofficially, in the gaps between the org chart’s boxes and the contract’s clauses.

The forces that produce hybrids are worth naming, because each one predicts where the seams will be. Contracts hybridize: chapter 20 showed the sourcing decision running per stream, the fixed price attaching where solution uncertainty is low and the time-and-materials where it is high, and every such contract creates a seam between the commercial models, because the incentives and the acceptance criteria differ on either side of it. Professional specialization hybridizes: the construction crew, the platform team, the migration analysts, the trainers, and the privacy officer run on different rhythms, different evidence habits, and different calendars, and the seams between them are where their calendars do not touch. Regulation hybridizes: the certification gate runs on the certification body’s calendar, not the project’s, and no amount of adaptive fluency moves it. And uncertainty itself hybridizes: the work whose outcome is knowable runs predictively, the work whose outcome emerges runs adaptively, and the line between the two is drawn by evidence, not by preference.

The field signal that the hybrid is accidental is specific, and it is the one the opening scene lived: the seam that is resolved only when it breaks. The accidental hybrid looks, from the outside, like a project that works, because the seams get resolved. The resolution is exactly the problem: it happens under pressure, late, at the moment the evidence is least complete and the stakeholders are most committed, which is the worst possible moment to make the decision. The cost of the accidental hybrid is not the rework alone, though the rework is real. It is the concentration of the coordination in the people who can translate, and the loss of the coordination the day they are not in the room.

The intentional hybrid is the same work with the seams owned. The difference is not the choice of methods; chapter 13 already made that choice, and it was the right one. The difference is the architecture: the streams named, the unit of hybridization stated for each, the seams listed with owners, the synchronization points scheduled with their evidence, the translations written down with their rules, and the re-tailoring decision recorded. Nothing about the intentional hybrid requires more ceremony. It requires that the coordination the heroics were providing be designed instead, because design is the only form of coordination that survives the absence of the coordinator.

The unit decides the design

The move that makes a hybrid designable is the naming of the unit being hybridized, and it is the question chapter 13 promised but did not yet make operational: hybrid by what? The word hybrid, unqualified, is a mood. Hybrid by component, phase, risk, team, contract, or governance layer is a design. Six units cover the field, and each has its own question, its own minimum viable form, and its own failure.

Hybrid by component names the part of the product or project that runs on which clock. The building is predictive; the platform is adaptive; the workflow is iterative; the rollout is incremental. The question is which component, and the minimum viable form is a one-line statement per component, the life-cycle decision matrix of chapter 13 in operating shape. The failure is the component hybridized by default rather than by decision: the platform called adaptive and run predictively because the vendor’s quote required a fixed scope, or the building called predictive and changed weekly because the community liaison keeps bringing neighborhood feedback into the construction meeting. That is not wrong feedback; it is the wrong channel.

Hybrid by phase names the stage that runs on a different rhythm within one stream. Discovery runs on probes and learning; the build runs on gates; commissioning runs on hold points. The question is which stage, and the minimum viable form is the phase boundary with its own entry and exit evidence. The failure is the phase boundary that is invisible: the discovery that drifts into the build and brings its uncertainty with it, or the gate that treats commissioning as a ceremony because the physical work is done.

Hybrid by risk class names the work that is clocked by its risk rather than by its method. The regulatory evidence runs on gates because the certification body moves on its own calendar and the consequence of lateness is a closed clinic. The clinical safety demonstration runs on gates because a floor is not a trade. The question is which risk, and the minimum viable form is the row in the obligations register of chapter 24 that carries its own clock. The failure is the risk class clocked by the stream’s calendar instead of its own: the privacy assessment scheduled against the project’s convenience, which is how the gap in chapter 24 was found one week before the gate.

Hybrid by team names the teams that run on which rhythm. The construction crew works to the gate calendar; the platform team works to the iteration; the migration analysts work to the reconciliation cadence; the trainers work to the cohort schedule. The question is which team, and the minimum viable form is the team’s contract with the seam: what the team owns, what it needs, and when it delivers. The failure is the team told it is adaptive and then governed by the predictive calendar anyway: the standup that reports to the steering committee’s monthly gate, the clock mismatch chapter 30 named as the hybrid rhythm’s failure, the gate that waits for the cadence meeting’s month, the slice that ships past the gate’s date.

Hybrid by contract names the commercial model per stream, the sourcing decision of chapter 20 in operating shape: the fixed price on the integration and the migration, the time-and-materials on the build, the framework on the equipment. The question is which contract, and the minimum viable form is the acceptance definition written into each contract, with the seam evidence, which chapter 13 made the nonnegotiable: the release evidence, the integration test, the rollback drill, the readiness checklist row, written as the definition of acceptance so that a release cannot be accepted by the vendor’s standard and fail the adoption corridor. The failure is the contract that hybridizes the commercial model without hybridizing the acceptance: the fixed price that transfers the integration risk to the vendor, who prices it into the quote, which chapter 13 priced in the decision room.

Hybrid by governance layer names the control layer that runs on which cadence. The daily coordination runs on the standup; the review runs on the iteration; the steering runs on the month; the gate runs on the evidence. The question is which layer, and the minimum viable form is the cadence map of chapter 30 with the output of each layer protected. The failure is the layer hybridized by accident: the gate that happens to fall on the month because that is when the committee meets, which is the difference between a gate and a meeting that chapter 8 drew, the gate as a decision on evidence, the meeting as a calendar event that merely meets.

The table the room needs is small, and it is worth building once, because the naming test is the chapter’s entrance requirement: if you cannot say which unit is being hybridized, you do not have a hybrid, you have a mess, and no coordination design can rescue a mess whose units were never named.

At Meridian, the naming is not hard, and it is the chapter’s first act. Construction is a component running predictively, plus a contract, the fixed-price drawing set with the liquidated-damages date, and a phase, the commissioning window with its hold points. The platform is a component running adaptively inside gated releases, a contract, the time-and-materials with the acceptance evidence pack, and a risk class, because the integration carries the clinical safety case. The migration is a component running on its own throughput rhythm, with the privacy review as a risk-class clock on its reference tables. The workflow design is a component running iteratively, and the certification is a risk class running on the certification body’s gate calendar. Six units, all named in an afternoon, and the naming does the design’s first work: it tells the room which clocks exist, which is the precondition for deciding where they must synchronize.

Draw the architecture, then the seams

The second act of the intentional hybrid is the architecture drawing, and its grammar is the one the book has used since chapter 3: the system map with the relationships, not the decoration. The drawing is simple, and the simplicity is the point. Each stream is a horizontal band, running on its own clock from left to right. The bands do not touch. The synchronization points are vertical lines that cross all the bands, and each vertical line is a shared milestone where the streams must meet with evidence. The bands are the life cycles; the vertical lines are the seams made visible; and the space between the lines is where each stream is free to run on its own rhythm without interference, which is the entire point of the architecture: the hybrid is not the blurring of the streams, it is the deliberate separation of them, with the meetings scheduled and owned.

Figure 33.1: The component-based hybrid model at Meridian, the
streams as bands, the synchronization points as vertical lines.
Each band runs on its own clock; each vertical line is a shared
milestone where the streams must meet with evidence. A stream
that cannot produce its evidence at the line becomes the line's
constraint, which is how the architecture makes the long pole
visible before it becomes a crisis.

  CLOCK                      PREDICTIVE              ITERATIVE
  CONSTRUCTION     |---------|=========fit-out========|
  (clinic 4)       |  frame  | commissioning, hold    |
                   |         | points, C-of-O         |
  CLOCK                      ADAPTIVE (2-week loop)
  PLATFORM         |--release candidate--|--config 9 items--|
  (gated releases) |  integration test   |  third Friday     |
  CLOCK                      THROUGHPUT
  MIGRATION        |-- 41,000 of 66,000 records --|
  (records/week)   |  active set  |  archive 9,000   |
  CLOCK                      WAVES
  TRAINING         | cohort 3 | cohort 4 | cohort 5 | cohort 6 |
  (cohorts)        |  super-user roster, double-booking        |
  CLOCK                      GATES (certification body)
  REGULATORY       | amendment | assessment | certificate |
  (certification)  |  sub-processor 4     |             |
                                                          |
  SYNC POINTS      capability    certification    WAVE GATE
                   release       gate             December
                   (3rd Friday)  (evidence)       (readiness
                                                    checklist)

The drawing does three jobs. It shows the streams as separate, which is the honest state: the construction crew is not going to become adaptive because the platform team is, and the certification body is not going to run on sprints. It shows the synchronization points as the only places the streams must meet, which is the permission the streams need to run on their own clocks without apology. And it shows the constraint: the vertical line where the streams’ forecasts disagree most, the line where one band’s evidence cannot be ready when the others are. That is how the drawing makes the long pole visible before it becomes a crisis, which is the difference between the architecture and the heroics.

The architecture’s operating form is the interface calendar, and it is the field output this chapter’s reader should be able to produce for their own project. The interface calendar is the seam discipline of chapter 13, the interface register of chapter 31 applied across methods, in time: a table whose rows are the interfaces between the streams, and whose columns are the interface, the evidence it requires, the owner, the cadence, and the gate or trigger. The calendar is small, because the interfaces are fewer than the org chart suggests. At Meridian, there are the construction-to-platform seam, where the certified building must receive a platform release with its integration evidence; the platform-to-migration seam, where the reference tables are gated by the privacy review; the workflow-to-training seam, where the training follows the workflow standard Hana owns; the training-to-adoption seam, where the super-user program meets the adoption corridor of chapter 11; and the certification-to-platform seam, where the adaptive releases must never outrun the certification evidence. That last row, chapter 24 proved at the cost of a week of pressure, is the one that cannot be waived. Five rows, five owners, five clocks, one calendar.

The seam owner is the load-bearing assignment, and it comes from the governance map of chapter 8: one person per interface, with the decision right the interface requires. The seam owner does not do the work on either side; the seam owner owns the crossing, which means owning the evidence that crosses, the date it crosses, and the escalation when it will not. The discipline of the interface calendar is the chapter 30 rule in its hybrid form: a row without an owner is a seam without a meeting. The field signal is the question the room cannot answer because the answer lives in no one’s job description, which is exactly the question the opening scene asked and nobody could answer, because the December gate had no owner, only a finance director who wanted the date and a project leader who carried the translation.

The failure of the architecture is the failure chapter 30 named in its hybrid register: the clock mismatch. The gate that waits for the cadence meeting’s month. The workflow slice that ships past the gate’s date. The two rhythms that never touch, which is how the clinics opened with the training gap that chapter 11 exposed, the trained-but-not-adopted thirty-five percent, a seam that existed and was never calendared. The architecture prevents the mismatch not by slowing the fast stream or speeding the slow one, but by naming the points where they must touch and the evidence they must bring. Then the mismatch is discovered at a synchronization point, with time to decide, instead of at a gate, with time to react.

Shared milestones and the evidence that opens them

The synchronization points are shared milestones, and the discipline of the milestone is the one chapter 16 taught: a milestone is evidence, not calendar decoration, and every date names what becomes true and checkable. The shared milestone goes further, because it is owned by more than one stream, and the evidence it requires is produced by more than one stream, which means the milestone must name the evidence in the language of the stream that must trust it. At Meridian the shared milestones are the capability release, the certification gate, the wave gate, and the adoption corridor reviews, and each one is a sentence, not a date. The capability release becomes true when the release candidate passes the integration test and the configuration for the next clinics is loaded. The certification gate becomes true when the certification body’s assessment is complete and on file, the row chapter 24 named as the one that cannot be waived. The wave gate becomes true when the readiness checklist’s five rows, five owners, one gate, the instrument chapter 29 built, are all ready, which is the choreography of the rows arriving on the same date, the five windows of physical readiness, workflow readiness, data readiness, training readiness, and regulatory readiness, each with its owner and its evidence. The adoption corridor review becomes true when the behavior at the point of work is inside the target corridor, the week-eight and week-sixteen gates of chapter 11 that decide the next waves.

The integration evidence is the content of the shared milestone, and its rule is the chapter’s most important single sentence: no stream signs for another stream’s row. The construction superintendent certifies the building; the platform team attests to the release’s integration evidence; the migration lead attests to the reconciliation; Sam attests to the competency checks; Esther’s row is the certification body’s to complete, not the project’s, the structural fact chapter 24 proved with the finance director’s parallel-connection proposal. The gate reads the rows, and the rows are the honest voice, because each row’s evidence is produced by the control that the row names, and the row that cannot be produced on time is the row that decides the gate, which is how the architecture turns the finance director’s question from a hope into a forecast.

The floors inherit the discipline chapter 24 built: a floor never degrades into a provisional. The privacy certification, the accessibility report, the clinical continuity demonstration, the escalation rehearsal: these rows move only through the obligations register, never through the schedule’s convenience, and the gate’s conditional release, which chapter 29 ran as the decision-room drill, is legitimate only when the condition names the control, the owner, the date, and the consequence. The shared milestone is where the floors and the provisionals are sorted, and the sorting test is the one chapter 29 taught: which row, if it slipped, would make the opening unsafe, and which row, if it slipped, would merely be late. The first is a floor. The second is a provisional with a date. The gate is the decision that sorts them, on evidence, on the record, and the architecture’s purpose is to make that decision early enough to be a decision rather than a rescue.

The definition of done that crosses the seams

Every stream has its own definition of done, and the hybrid’s first discipline is to let it. The word done, as the book’s usage rule insists, must specify its level, and in a hybrid the level is the first thing to settle. At the work level, the contractor’s done is substantial completion against the drawing set; the platform’s done is the increment’s definition of done, the statement the team owns, the standard the Scrum Guide, November 2020 edition, by Ken Schwaber and Jeff Sutherland, describes in its own vocabulary, which chapter 32 rendered in the book’s words; the migration’s done is a reconciled record, not a moved one; the training’s done is a passed competency check, not an attended session, the chapter 11 lesson in measurement form; and the certification’s done is the assessment complete and on file. Each of these is local, and the local definitions must not be blurred, because blurring them is how the compromise hybrid begins: the increment counted as done because the demo was impressive, the training counted as done because the room was full, the migration counted as done because the records moved, with the reconciliation gap discovered later.

At the wave level, done is the cross-method definition of done, and it is the field output the hybrid’s reader most needs, because it is the only instrument that makes the gate decisionable. The cross-method definition of done is one statement, written in the language of the gate, that names what must be true for the wave to proceed, with each requirement traceable to the stream that produces it and the evidence that proves it. The definition chapter 10 built at Meridian spans the seams, and the wave gate is its operating form: the platform behavior passes the clinical workflow review; the audit trail passes the privacy certification checks; the migration reconciliation reports zero unexplained gaps; the clinicians scheduled for the clinic complete the competency check; and the continuity drill shows a clinic operating on paper and recovering within the target time. The statement is short. The point is that it is one statement, owned by the gate, not five statements owned by five streams, because the wave either is ready or is not, and the room that holds five definitions of done can always find one that is green.

The seam-level definition of done is the discipline that makes the cross-method statement credible, and it is the piece the streams most often skip. Each seam needs its own done, written in the receiving stream’s language. The construction-to-platform seam is done when the platform release is accepted on the integration evidence, not on a demo, and when the rollback drill has run, the seam evidence chapter 13 made the contract’s nonnegotiable. The platform-to-migration seam is done when the reference tables pass the privacy review and the reconciliation for the migrated set reports zero unexplained gaps. The workflow-to-training seam is done when the training is written against the behavior the workflow requires, not against the screens, the chapter 10 error in reverse. The training-to-adoption seam is done when the super-users are rostered at the desks during the transition, not when the cohorts are scheduled. The seam-level done is what makes the interface calendar’s rows decisionable: the row’s evidence is the seam’s done, and the owner signs the row, which is the difference between a calendar and a hope.

Translate between the clocks

The third act of the intentional hybrid is the translation, and it is the act the opening scene failed. The streams do not speak the same language: the construction stream speaks percent complete against a baseline, the platform stream speaks backlog items and release candidates, the migration stream speaks records and reconciliation, the training stream speaks cohorts and competency, and the certification stream speaks evidence rows and gate dates. None of the numbers is wrong. None of them can be added. The seventy-four percent, the nine items, the forty-one thousand records, the three of six cohorts, and the assessment scheduled are five true statements about five different things, and the room that treats them as one statement about readiness has committed the hybrid’s central error: the translation without a rule.

The translation rules are four, and they are simple enough to hold in one sentence each. First, name the unit: every number in the room names what it measures, the physical progress rule for the construction, the flow forecast for the platform, the reconciliation rate for the migration, the cohort plan for the training, the evidence row for the certification, and the unit is stated before the number is used, because an unnamed unit is where the optimism enters. Second, convert only at the seam: the conversion happens in the interface calendar’s row, with a stated rule and its assumptions, not in the report, because the report is where the conversion is read by people who will not check the rule. Third, the conversion never increases certainty: a range converted to a single number is a promise wearing a conversion’s clothes, the chapter 15 discipline in hybrid form, and the conversion preserves the range or it is not a conversion, it is a hope. Fourth, the receiving stream states the evidence it needs, and the producing stream states what it can give, and the seam owner reconciles the two, because the translation is a negotiation about evidence, not a computation about numbers.

The translation is a decision, not an arithmetic, and the decision is where the hybrid’s honesty lives. The contractor’s seventy-four percent, converted to a gate date, requires the assumption that the remaining twenty-six percent runs at the same rate and that the inspectors’ calendar holds. The platform’s nine items, converted to a date, requires the assumption that the cycle time holds and that the release does not surface a rework. The migration’s two thousand records a week, converted to a completion date, requires the assumption that the archive backlog does not degrade the reconciliation rate. Each conversion is a bet, and the translation table is where the bets are written down: the stream, the number, the measurement rule, the conversion, the assumption, the owner. The table is the translation theater’s antidote, because the theater, which this chapter will name in full, is the conversion that happens in no one’s table, in the report, in the head of the person who happens to be in the room.

The integrated forecast is the translation’s output, and it is the instrument that answers the finance director’s question with a forecast instead of a hope. The integrated forecast builds the wave-level date from the streams’ rates, with the assumptions spoken, and the arithmetic at Meridian is the chapter’s worked example. The gate is December, twelve weeks out. The construction forecast: eight weeks to the certificate of occupancy, four weeks of float. The platform forecast: the release in three weeks, the configuration in about six or seven, roughly six weeks of float. The training forecast: four weeks once the workflows finalize. The certification forecast: the amendment closes in two weeks, the assessment runs six, so about eight weeks. And the migration forecast: twenty-five thousand records at two thousand a week, twelve and a half weeks, the long pole, outside the gate. The integrated forecast makes the constraint visible without any heroics: the December gate holds only if the migration changes, and the change is the subject of the re-tailoring section. The forecast’s job is to name the constraint while there is still time to decide what to do about it, which is the difference between the architecture and the rescue.

Change at different speeds

The fourth act is change control, and the hybrid’s version is the routing question: which streams does this change touch? A change in a hybrid is not one event; it is the same event landing on several clocks, and the mastery is routing it through the path each stream’s clock requires, at the speed each stream’s clock runs. The construction baseline moves through formal change, the baseline-change record of chapter 31, because the drawing set is the promise and the change is expensive, and the contractor’s liquidated-damages date does not move because the change was oral. The platform backlog changes on the iteration, the backlog refresh of chapter 32, because the empirical loop’s planning is its change control, and the change that cannot wait for the iteration is the change that should not be a platform change at all: it is an escalation with an owner and a date. The migration’s sequencing changes with its reconciliation cadence, and the training plan follows the workflow standard, not the calendar, the order chapter 11 taught, the environment first, the behavior second.

The obligations move on their own path, and the path is the obligations register of chapter 24: a risk may be accepted, an obligation may not, and the change that touches a floor, the privacy certification, the accessibility, the clinical continuity, moves only through the gate, with the record, and the owner who can say no. The routing rule is the chapter’s sentence: a change that crosses streams crosses through all their paths, and the seam owner integrates the impact, so that the change record carries the full footprint, the streams touched, the owners, the dates, the consequences, in one place.

The worked routing at Meridian is the policy change, and it is the example this chapter sets up for chapter 39 to complete. The district health authority issues a new referral pathway policy: clinics will hand patients to the hospital through a new referral form with a new data field and a new handoff procedure, and the policy is mandatory from the first of the new year. It lands in Marcus’s backlog as a software change, “add the referral field to the discharge form,” because that is where it arrived. The routing question exposes what the software change hides: the field is a data-field change, so it touches the platform, the release plan, and the certification evidence, because the audit trail must record what the field records; the handoff is a workflow change, so it touches the workflow standard Hana owns, and the clinical governance review; the procedure is a training change, so it touches the cohort plan Sam owns, and the competency check; and the acceptance is a readiness change, so it touches Nora’s acceptance criteria and the wave gate’s definition of done. One policy, five streams, five paths, one record. The software change was not wrong; it was incomplete, and the incompleteness is exactly what the routing question is for: the change that touches the workflow and the training and the acceptance cannot be routed as a software change, because the streams that will fail the gate are the ones that never saw the change.

The speed discipline is the counterpart: the streams do not change at the same speed, and they should not. The construction change moves on the gate cadence, and forcing it onto the iteration would be theater. The platform change moves on the iteration, and forcing it onto the gate cadence would be ceremony. The policy change touches both, and the seam owner integrates the impact, which means reconciling the speeds: the data field can ride the next release, the workflow standard needs the clinical review, the training needs the cohort, and the gate needs the evidence. The change record carries the dates each stream can make, with the seam owner responsible for the latest date, because the change is done when the slowest stream is done, and the record says so.

Fund and report the mix honestly

The fifth act is the money, and the hybrid’s funding discipline is the budget line that spans the streams without pretending they are one. The cost model of chapter 18 runs per stream: the construction contract with its progress payments and its liquidated damages; the platform’s time-and-materials with the acceptance evidence pack; the migration effort, priced by the records and the reconciliation, not by the month; the training with its cohorts and its super-user capacity; the certification with the body’s fees. The budget is the sum, but the control is per stream, because the variance that matters is the stream’s variance: the construction overrun against the fixed price, the platform’s actuals against the estimate, the migration’s rate against the plan. The report that sums them into one number has performed the translation without a rule, which is the hybrid’s central error applied to money.

The funding gates inherit the rolling authorization of chapter 8: the grant pays from the clinic-month the doors open, at two hundred fifty thousand units per clinic-month, the economics chapter 5 established, and the December window is not a preference, it is the grant’s arithmetic. The three clinics of the final wave at two hundred fifty thousand a clinic-month: a month’s slip costs seven hundred fifty thousand units that the grant will never pay, because the grant ends at month thirty-six and the eligibility runs on the clinic-month, and the cost of the slip is the cost of the delay, the chapter 5 lesson in the hybrid’s register, which is why the finance director asks the question and why the question deserves a forecast, not a hope.

The reporting discipline is the one the reader carries from chapter 1’s promise: the grant officer counts wait times and screening uptake, not percent complete, and the steering committee reads the forecast, the decisions, and the decision requests, not the streams’ victories. The one-page report of the hybrid carries the streams in their own units, the measurement rule per stream, the integrated forecast at the wave level with the assumptions spoken, the open decisions with their owners and dates, and the decision requests the gate needs. Where color is used, the status color carries text, because the color that means ready in the construction stream means something else in the certification stream, and the report that reconciles the streams must reconcile their colors too. The failure is the report that performs the translation the streams refused: the single percent-complete figure for the wave, built from numbers no stream can reproduce, which is the translation theater in reporting form. The field signal is the number in the report that the stream owners do not recognize as theirs.

Worst of both worlds

Every hybrid has its failure shapes, and the honest book names them, because each one is a plausible path from the opening scene, and each one is walked by competent people for reasons that make sense at the time. The compromise hybrid dilutes every stream to fit one calendar: the platform team gets a two-week cycle that ends at the monthly gate, the construction gets an adaptive drawing set, the migration gets a percent-complete number, and the certification gets scheduled around the project’s convenience, the exact mistake chapter 24 refused. The compromise is the most common shape because it is the most polite: everyone gives a little, and what everyone gives is the integrity of their own clock. The field signal is the room where nobody can say what a green status means, because the green has been averaged from colors that meant different things.

The identity hybrid is the compromise wearing a label: “we are hybrid,” meaning no baseline anywhere, no gate that holds, no iteration that ships, the license to drift that chapter 32 named in its adaptive form, applied to the whole project. The field signal is the stream that cannot state its own baseline or forecast: the construction whose schedule is a memory, the platform whose product goal is a poster, the migration whose rate is a hope. The identity hybrid is the umbrella failure, and its repair is the chapter’s whole method: the unit named, the seam owned, the evidence defined.

The seamless fiction is the opposite shape, the org chart’s belief that one project means no seams need owning: the interface calendar has no rows because the project is one team, the gate is a meeting, and the seam is discovered at the moment of failure, the certification gap found one week before the wave-two gate, the training gap found after go-live, the interface specification found at the dry run. The field signal is the question that no one’s job description owns, and the cost is the rework and the heroics, the coordination that was being provided by pressure.

The double baseline is the two calendars: the construction schedule and the platform roadmap, both current, both moving, never reconciled, with the wave gate reading a date that no stream owns and both streams disowning at the gate. The field signal is the meeting where the construction says the gate is the contractor’s date, the platform says the gate is the grant’s date, and the migration says the gate is nobody’s date. The cost is the gate’s credibility: the date that cannot fail because no one owns it, which fails anyway, at the worst time. The repair is the shared milestone dictionary, the single statement of what becomes true and checkable, owned by the gate.

The translation theater is the number that no stream can reproduce: the backlog items converted to percent complete for the report, the records converted to weeks without the assumption, the wave shown ready on a figure that the streams could not sum. The field signal is the report number that the stream owners do not recognize, and the repair is the translation table with its units, its rules, and its assumptions, the conversion that happens at the seam, with the owner.

And the heroics coordinator is the person, and it is the one the opening scene was about: the project leader who carries every translation in her head, who can answer the finance director’s question because she alone holds the five streams’ forecasts, who resolves every seam when it breaks because the seams are her job description, the chapter 25 Dana in her most expensive form. The field signal is the question only the project leader can answer: the corridor commitments, the calendar that lives in the head. The heroics coordinator is not a failure of the person; it is a failure of the architecture, and the repair is not to work harder. It is to move the coordination from the head into the interface calendar, with owners, evidence, and clocks. The coordination that lives in one head is the coordination that dies when the head leaves, and the project that depends on the heroics of the coordinator is the project that has not designed its own coordination.

Re-tailor when the seams fail

The architecture is a model, and models meet evidence, and the evidence will disagree with the model, which is not the architecture’s failure, it is its normal operation, and the discipline is the re-tailoring decision. Chapter 13 promised the re-decision triggers, the approach as a hypothesis with explicit conditions under which it is re-decided; chapter 22 gave the trigger discipline, the condition, the evidence, the owner, the date; and the hybrid’s re-tailoring is the two joined: when a seam fails, the project re-tailors deliberately, with a record, and the re-tailoring is a design decision, not an emergency.

The re-tailoring has five moves, and the room chooses among them with the same discipline it uses for any decision: the context, the options, the evidence, the choice, the record, the chapter 4 form. Decouple: the seam is split, the stream that can meet the gate proceeds, the stream that cannot carries its incompleteness as a dated provisional with an owner and a trigger, the chapter 29 discipline, and the December arithmetic makes the move concrete, the active-set migration as the floor, the archive backlog as the dated provisional, the clinic opening on the records that must be complete and carrying the records that may complete after, with the reconciliation and the privacy review as the conditions. Resequence: the wave’s order changes, clinic five opens after clinic four rather than with it, the chapter 5 sequencing discipline applied to the seam’s failure. Buffer: the float is added at the seam, the two weeks between the migration’s completion and the gate that the forecast shows is missing, the chapter 23 resilience in scheduling form. Change the clock: the platform releases at capability milestones rather than on the calendar, or the migration runs a second reconciliation stream through the archive, the clock changed where the evidence requires it. And split the wave: the three clinics of the final wave become two and one, the December gate holding for the pair, the third following on its own gate, the sequencing that turns a wave failure into a forecast.

The re-tailoring’s floors are the obligations register’s rows, and the rule is absolute in the way chapter 24 made it: the re-tailoring never moves an obligation row. The privacy certification, the accessibility, the clinical continuity, the escalation rehearsal: these are not re-tailored; they are met, and the re-tailoring decides what the project does around them, the decoupling, the resequencing, the buffering, the clock change, the wave split, never the row itself. The field signal of the healthy re-tailoring is the record: the context, the options considered, the evidence, the choice, the floors preserved, the new trigger dates, and the review date. The re-tailoring that happens without a record is the drift that chapter 31 named in its predictive register, the architecture quietly renegotiated at every failure until there is no architecture, only the calendar that survived.

The wave-three architecture

The worked application is the meeting the opening scene interrupted, and it is worth following in sequence, because the sequence is the chapter. The room is the steering committee’s review, four weeks after the December question: Dana, Elena Marchetti, Marcus, Nora Kariuki, Sam, Hana Lindqvist, Esther, the migration lead, the construction superintendent, and the finance director, with the five streams’ numbers on the wall and the December gate, twelve weeks out, in the agenda’s title. Dana opens with the chapter’s first act: the naming. The room names the units in an hour, and the naming is the design’s first decision: construction, a predictive component under a fixed-price contract with a commissioning phase; platform, an adaptive component inside gated releases under the time-and-materials with the acceptance evidence pack; migration, a throughput component with the privacy review as its risk-class clock; workflow, an iterative component; training, a wave component; certification, a risk-class component on the certification body’s gate calendar. The room draws the bands and the vertical lines, and the drawing makes the seams visible: the construction-to-platform seam, the platform-to-migration seam, the workflow-to-training seam, the training-to-adoption seam, the certification-to-platform seam, and the December gate. Five seams, five owners, one calendar.

The interface calendar is built in the same session, and it is the session’s real work, because the calendar is where the heroics move into the architecture. The rows take the shape the chapter taught, and two carry the session. The certification-to-platform row names the evidence, the assessment complete and on file, the owner, Esther, and the trigger: the amendment closes in two weeks, or the submission date moves with the consequence named. The platform-to-migration row names the reference tables passing the privacy review and the reconciliation at zero unexplained gaps, owned by the migration lead with Esther’s review. The rest take their owners and cadences quickly: the construction-to-platform row, Marcus, on the third-Friday release; the workflow-to-training row, Hana’s standard finalized before the cohort plan, Sam’s training written against the behavior; the training-to-adoption row, Sam and Nora, with the double-booking of two super-users named as the row’s first risk.

The cross-method definition of done is written on one page — one sentence in the gate’s language, each row traceable to the stream that produces it — and the room argues about it for an hour, which is the hour well spent. The argument is about the migration row, because the migration is the long pole, and the long pole is where the definition of done does its work. The room must decide whether the wave’s done includes the archive backlog, and the decision is the re-tailoring’s first move: the decoupling. The active records, sixteen thousand of the twenty-five thousand remaining, are the floor: the clinic cannot open without the records its clinicians will read. The archive, nine thousand records, is the provisional: it can complete after opening, on a dated plan, with the reconciliation and the privacy review as its conditions, at two thousand a week, four and a half weeks, inside the wave’s hypercare, with the trigger that the reconciliation reporting unexplained gaps would stop the clinic’s opening and the archive’s schedule together. The integrated forecast is recomputed with the decoupling, and the constraint moves: the migration’s active set clears in eight weeks, inside the gate; the certification clears in eight; the construction in eight; the platform in six or seven; the training in four; and the December gate holds, with the archive as the dated provisional and the finance director’s question answered with a forecast: clinic four opens in December on the floor, with the provisional named, owned, and dated.

The change routing runs through the same session, because the referral pathway policy arrived the day before, and the room routes it the way the chapter teaches: the policy touches the platform, the release plan, and the audit trail, Marcus; the workflow standard and the clinical review, Hana; the training and the competency check, Sam; the acceptance and the wave gate’s definition of done, Nora; and the record carries the five streams, the five owners, and the latest date, the change done when the slowest stream is done. The funding is confirmed in the session’s last act: the budget lines per stream, the funding gates tied to the evidence, the grant’s clinic-month arithmetic spoken aloud, two hundred fifty thousand per clinic-month, the December gate’s value, and the report’s shape, the streams in their own units with the integrated forecast and the decision requests, one page, the grant officer’s wait times and screening uptake in the footer, where they belong.

The machine’s boundary is drawn in the session too, because the hybrid is a place where automation earns a bounded role. The machine can draft the interface calendar from the streams’ records, the rows, the evidence, the cadences, the draft the room corrects; it can flag the vocabulary mismatches, the seventy-four percent sitting beside the nine items with no reconciliation, the report number no stream can reproduce, the draft signal the room verifies; and it can compute the integrated forecast from the streams’ rates, the weeks, the ranges, the long pole, the arithmetic with its assumptions stated. The human check is the chapter’s content: the unit naming, the seam owners, the floors, the gate decision, the re-tailoring, the routing, and the data boundary, the patient records, the migration files, the reference tables, the clinical data, never entering an unapproved system, the book’s standing rule in its strongest form, because the hybrid multiplies the systems the data might flow into, and the boundary multiplies with it.

The discipline that carried the session is the chapter’s whole argument, and it is worth stating once plainly: the hybrid is not the blurring of the streams, it is the deliberate separation of them, with the meetings scheduled, owned, evidenced, and translated. The coordination that the heroics were providing, the translations in Dana’s head, the seams resolved under pressure, moves into the architecture: the interface calendar, the shared milestones, the cross-method definition of done, the translation table, the change routing, the integrated forecast, and the re-tailoring record. Then the answer to the finance director’s question does not depend on who is in the room, which is the difference between the hybrid that works by design and the hybrid that works by heroics, until the day the heroics are not in the room.

The most common next failure is the one the December gate will face in November: the streams meet at the certification gate, the rows disagree, the calendar’s row with the oldest owner is the one nobody updated, and the temptation is the compromise, the gate that softens the definition of done, the provisional that acquires no date, the report that averages the colors, the reversion to the accidental hybrid the architecture was built to replace. The architecture, like every model in this book, is only as honest as the evidence it is fed, and the evidence is only as honest as the owners who feed it. The next chapter turns to the scale where the seams multiply beyond five: the multiple teams, the dependencies between them, the system-level flow, the integration events, the architecture of this chapter becoming the dependency board of chapter 34, the coordination that must scale from the clinic’s seams to the program’s.

Practice

One. A quick check: name the unit. For each scene, say which unit is being hybridized, component, phase, risk class, team, contract, or governance layer, and the one seam the scene is most at risk of leaving unowned. (a) A hospital wing is built to a fixed drawing set while the interior fit-out is redesigned iteratively with the ward nurses, because the workflows keep changing the layout. (b) A platform’s data infrastructure is bought on a fixed price while the features are built on time-and-materials, with the acceptance defined once in the contract. (c) A product team ships features on two-week cycles while the regulatory submission runs on a quarterly evidence gate, and the team keeps scheduling the submission against the sprint calendar. (d) An organization’s PMO imposes a monthly gate on all work, and the product teams report their iterations at the gate. (e) A crisis response runs daily huddles for the field while the donor reporting runs on a formal weekly cadence, and nobody owns the conversion between the huddle’s tonnage and the report’s dashboard.

(a) is a phase hybrid with a component inside it: the building is the predictive component, the fit-out is the iterative component, and the unowned seam at risk is the moment the fixed shell must accept the changing interior, the drawings, the services, the certificate, because nobody owns the moment the two meet. (b) is a contract hybrid: the unit is the contract, and the seam is the acceptance, the evidence the fixed-price stream and the time-and-materials stream must both produce, the integration test, the rollback drill, the readiness row, written into the contract as the definition of acceptance, which chapter 13 made the nonnegotiable; the scene’s risk is the acceptance defined once for the whole platform, which the fixed-price stream will meet and the adaptive stream cannot. (c) is a risk-class hybrid: the regulatory submission runs on its own clock because the consequence of lateness is a license, not a sprint, and the seam at risk is the calibration, the team scheduling the submission against the sprint calendar, the clock mismatch, the risk clocked by the stream’s calendar instead of its own. (d) is a governance-layer hybrid that has collapsed into the compromise: the monthly gate is a governance layer, the product teams’ iterations are a delivery layer, and the two can coexist only if the gate reads the iterations’ evidence rather than forcing the iterations to end at the gate; the scene as written is the gate that falls on the month by accident, the meeting that merely meets. (e) is a team hybrid, the field team on the daily clock and the donor reporting on the formal cadence, and the unowned seam is the translation, the conversion between the huddle’s tonnage and the dashboard’s number, with the rule, the unit, and the assumption, which the scene shows nobody owning. The common error is naming the method instead of the unit: “that is agile,” “that is waterfall,” which is the identity debate chapter 13 refused, when the question is which unit, which seam, and who owns it.

Two. A field drill: build the interface calendar. Take the project you lead or the one you work on, the one that is hybrid whether it admits it or not. Name the streams that actually run on different clocks, the construction, the platform, the migration, the workflows, the training, the regulatory evidence, the operations handover, whatever your work contains. For each interface between the streams, write the row: the interface, the evidence it requires, the owner, the cadence, and the gate or trigger. Then answer: which row has no owner, and what is the question that only the person who currently carries the translation can answer?

The drill passes when the calendar is small, five to eight rows, and every row has an owner from the governance map, not from the org chart, and when the unowned row is the one the room can name without a search. The most common failure is the calendar that lists the meetings instead of the interfaces, the steering committee, the review, the standup, which is the cadence without the seam, the meeting that merely meets; the repair is the interface, the thing that crosses, the evidence, the owner. The second failure is the row whose evidence is in the producing stream’s language only: the release signed off by the platform team, with no row for the receiving stream’s acceptance; the repair is the receiving stream’s language in the evidence field, because the seam is where the evidence must be trusted by someone who did not produce it. The third failure is the owner who is the project leader, every row, which is the heroics coordinator in table form; the repair is the governance map’s assignment, the seam owner as a decision right, not a favor.

Three. A field drill: write the cross-method definition of done. For one gate in your own hybrid, the go-live, the release, the handover, write the one-page definition of done: what must be true for the gate to open, written in the language of the gate, with each requirement traceable to the stream that produces it and the evidence that proves it. Then sort the rows into floors, the rows that cannot move, and provisionals, the rows that may carry a date, and write the condition for each provisional: the control, the owner, the date, the consequence. Name the row that your organization is most likely to treat as a provisional when it is a floor.

The drill passes when the definition fits on one page and the floor sort is defensible in the room. The most common failure is the definition written in the language of the most powerful stream: the gate defined by the construction certificate, with the training, the migration, the certification hidden in a footnote; the repair is the gate’s language, the sentence the receiving stream, the operations lead, the regulator, the user, would ask, and the row answered in that language. The second failure is the floor sort done by the schedule: the row that is convenient to move becomes the provisional, which is the chapter 24 mistake, the finance director’s parallel connection; the repair is the counterfactual test from chapter 29, which row, if it slipped, would make the outcome unsafe, and the unsafe row is the floor whatever the calendar says. The third failure is the provisional with no date and no owner, the condition that is a hope; the repair is the four-part condition, the control, the owner, the date, the consequence, and the review at which the condition is judged.

Four. A decision room: the December gate. It is four weeks after the opening scene at Meridian, and the numbers are on the wall: the December gate twelve weeks out; the construction at seventy-four percent, the certificate of occupancy forecast in eight; the platform’s nine configuration items at three a cycle, the release in three weeks; the migration at forty-one thousand of sixty-six thousand records, two thousand a week, twelve and a half weeks for the remaining twenty-five thousand, with the active set at sixteen thousand and the archive at nine thousand, the archive the slow part, and the reference tables gated by the privacy review; the training at three of six cohorts with two super-users double-booked; the certification subject to the sub-processor amendment, the assessment six weeks after submission, the submission two weeks after the amendment. The options: (a) hold December on the current plan and accelerate the migration by adding a second reconciliation stream for the archive, funded, pulling the reconciliation analyst from the existing-clinic rollout, which the capacity plan shows is the analyst’s only free window; (b) hold December by decoupling, the active set as the floor, the archive as the dated provisional with the reconciliation and the privacy review as its conditions, opening clinic four on the floor and completing the archive inside hypercare; (c) move the gate to January, keeping the wave whole and the migration complete before the first patient; (d) hold December on the current plan and complete the migration in parallel after opening, with no date and no owner for the archive, because the clinic needs the active records and the archive is administrative. Decide the move and say what the record must carry.

The defensible answer is (b), and the reasoning is the chapter’s whole method in one decision. The forecast says the migration is the long pole, twelve and a half weeks against the gate’s twelve, and the decoupling is the re-tailoring that respects the floor and the provisional: the active set, sixteen thousand records at two thousand a week, clears in eight weeks, inside the gate, and the archive, nine thousand records, completes in four and a half weeks inside hypercare, with the reconciliation and the privacy review as its conditions and a trigger, the reconciliation reporting unexplained gaps, that stops the opening and the archive together. (a) is defensible in intent and risky in execution: the second reconciliation stream pulls the analyst from the existing-clinic rollout, the capacity conflict chapter 25 surfaced in the corridor and chapter 19 priced, and the acceleration buys a few weeks of float at the cost of a bottleneck the plan already named. (c) is the perfectionist start wearing a schedule costume: the floor rows are met by the decoupling, and the January gate costs the three clinics a month of service, seven hundred fifty thousand units at two hundred fifty thousand a clinic-month, for an archive the floor does not require. (d) is the chapter 24 mistake wearing new clothes: the parallel migration with no date and no owner is the finance director’s proposal, the provisional that acquires no date, and the archive that “is administrative” is exactly the row whose reconciliation protects the records the clinicians will read, the data-quality floor that chapter 36 will price when the clinics open. The record must carry the unit naming, the seam owners, the cross-method definition of done, the floor and the provisional with the four-part condition, the trigger, the integrated forecast with its assumptions, and the review date, because the gate decision is only as credible as the record that carries it.

Five. A decision room: route the policy. The district’s referral pathway policy arrives at Meridian with a mandatory date: clinics will hand patients to the hospital through a new referral form, a new data field, and a new handoff procedure, from the first of the new year. The platform lead has logged it as a software change, “add the referral field to the discharge form,” and it is sitting in the backlog with a size and an owner. The wave gate is in December, and the new year’s policy lands inside the final wave’s hypercare. The options: (a) let the software change ride the release, because the field is a field and the clinic can start the handoff on paper; (b) route it as the change record requires, the platform, the release, and the audit trail under Marcus, the workflow standard and the clinical review under Hana, the training and the competency check under Sam, the acceptance and the wave gate’s definition of done under Nora, with the seam owner integrating the dates and the record carrying the full footprint; (c) defer the whole policy until after the wave, because the gate is in December and the policy is in January; (d) treat the policy as a certification change only, because the audit trail must record what the field records, and let the workflow and the training follow the data. Decide the routing and say what the record must carry.

The defensible answer is (b), and the reasoning is the routing question: which streams does this change touch. The field is a data-field change, so it touches the platform, the release plan, and the certification evidence, because the audit trail must record what the field records; the handoff is a workflow change, so it touches the workflow standard and the clinical review; the procedure is a training change, so it touches the cohort plan and the competency check; and the acceptance is a readiness change, so it touches the wave gate’s definition of done, which the wave’s clinicians will be assessed against while the new policy is already mandatory. (a) is the software-change trap: the field rides the release and the workflow, the training, and the acceptance never see the change, which is the incompleteness the routing question exists to catch. (c) defers a mandatory policy, which is not a deferrable event, it is a change in the environment the gate must absorb, and the deferral is the seamless fiction, the seam discovered at the moment of failure, the first patient handed over under a policy nobody trained. (d) gets the certification half right: the audit trail is real, but the workflow and the training are not optional follow-ons, they are the change, the behavior, the chapter 11 order, environment first, behavior second, and the routing that treats the behavior as following the data has inverted it. The record must carry the full footprint, the five streams, the five owners, the dates each stream can make, the latest date as the change’s date, and the seam owner’s name, because the change is done when the slowest stream is done, and the record says so, the routing that chapter 39 will show in full.

Six. The mastery drill: find and repair the conflicting control assumptions. Here is a hybrid plan, one paragraph, from a regional health network opening its final clinics. Read it and find the conflicts, then repair them, then name the field signal that revealed each conflict. “The network opens the final three clinics in December. The construction schedule is the baseline: it holds the December date, moves only through formal change, and reports percent complete against the drawing set. The platform team runs two-week cycles, refreshes the backlog by outcome, risk, and evidence, and reports nine items remaining for the clinic configuration. The migration is measured in records migrated, with reconciliation at zero unexplained gaps, and the remaining archive records are expected to clear after opening. The training is measured in cohorts completed, and the competency check is administered in the classroom. The gate report shows the wave at eighty-five percent ready, computed from the construction percent complete, the platform items, the migration records, and the cohorts. The referral policy change is in the platform backlog as a software change. The readiness checklist has five rows and four owners; the certification row is owned by the privacy officer, who reports the assessment scheduled.” Then answer: what would happen in this plan, in what order, and what would the architecture of chapter 33 change?

The plan’s conflicts are the chapter’s failure shapes in one paragraph, and the drill’s repair is the chapter’s method. The first conflict is the double baseline: the construction schedule holds the December date and the platform’s backlog refresh moves the configuration, two calendars, two owners, and no shared milestone dictionary, and the field signal is the gate report’s eighty-five percent, which no stream can reproduce, because it is computed from four numbers with four units, the translation theater, the conversion that happened in no one’s table. The repair is the shared milestone, the wave gate, with its sentence, what becomes true and checkable, and the integrated forecast built from the streams’ rates with the assumptions spoken, which would show the migration as the long pole, the archive at nine thousand records at two thousand a week, four and a half weeks after the active set, and the December gate holding only with the decoupling, the active set as the floor, the archive as the dated provisional, the reconciliation and the privacy review as its conditions. The second conflict is the seamless fiction: the readiness checklist has five rows and four owners, and the unowned row is the one the gate will discover at the gate, the row that chapter 24 proved, the certification, which here has an owner, Esther, and a status, “scheduled,” which is not evidence, it is a hope, and the field signal is the row that no one can update without the certification body, the structural fact that the row’s clock is not the project’s; the repair is the row’s gate and trigger, the assessment complete and on file, the amendment’s date, the consequence. The third conflict is the clock mismatch in the training row: the competency check is administered in the classroom, which is the box ticker’s evidence, the attendance counted as training, the chapter 11 lesson, and the field signal is the cohort number that never appears at the desk; the repair is the seam-level definition of done, the training done when the super-users are rostered at the desks and the check is administered at the point of work. The fourth conflict is the change routing: the referral policy in the platform backlog as a software change, the incomplete footprint, the field signal the policy that touches the workflow, the training, and the acceptance and appears once, in the platform; the repair is the routing, the five streams, the five owners, the latest date. The fifth conflict is the heroics coordinator implied by the whole plan: the eighty-five percent computed by someone, the translations in one head, the field signal the number no stream owns; the repair is the interface calendar, the translation table, the seam owners, the coordination moved from the head into the architecture. The order of failure in this plan: the gate report is trusted through November because the numbers are green in their own units; the certification row surfaces in early December as “scheduled,” and the gate softens it, the compromise hybrid; the clinic opens with the archive unreconciled and the referral policy trained on the old handoff, and the first unexplained gap in the reconciliation and the first mis-routed referral are discovered after opening, the accidental hybrid’s signature, the seam resolved when it broke. The architecture of chapter 33 changes the order: the conflicts are found in September, at the synchronization points, with time to decide, because the unit is named, the seam is owned, the evidence is defined, and the forecast is integrated, and the gate in December is a decision rather than a rescue.

Seven. The transfer question. On the project you lead, or the one you work on, where is the hybrid that is accidental? Which streams actually run on different clocks, and which of them was chosen, and which drifted, and when did the last seam break, and how was it resolved, by a design or by a heroics, and who was in the room? Can you name the unit, the component, the phase, the risk class, the team, the contract, the governance layer, for each stream, and what is the unit you cannot name, and what does that say about the seam that is unowned? Who owns each interface between your streams, and which row of the interface calendar has no owner, and what is the question that only the person who currently carries the translation can answer, and what happens when that person is not in the room? What are your synchronization points, and are they shared milestones with evidence, the wave gate, the capability release, the certification gate, the adoption corridor, or are they dates on two calendars, both current, both moving, never reconciled? What is the cross-method definition of done at your gate, and is it one statement in the gate’s language, or five statements in five streams’ languages, and which row of your gate is a floor that your organization treats as a provisional, and who would be harmed by the slip, and are they in the room? What does your report carry, the streams in their own units with the integrated forecast and the decision requests, or the single percent complete that no stream can reproduce, and what is the number in the report that your stream owners do not recognize as theirs? Where do your changes route, the software change that is really a workflow change, the policy that touches the training and the acceptance and appears once, in the backlog, and who integrates the impact, and what does the change record carry? When the seams fail, do you re-tailor with a record, the context, the options, the evidence, the choice, the floors preserved, the triggers dated, or do you drift, the architecture quietly renegotiated at every failure until there is only the calendar that survived? The durable principle: hybrid delivery is the normal state of complex work, and the mastery is designing it intentionally, the unit named, the seam owned, the milestone shared with its evidence, the translation made with its rules, the change routed through its reach, the mix funded and reported honestly, the worst-of-both-worlds shapes refused, and the re-tailoring recorded, because the coordination that lives in one head is the coordination that dies when the head leaves. The most common next failure is the one the December gate will face in November: the compromise, the gate that softens the done, the provisional that acquires no date, the report that averages the colors, the reversion to the accidental hybrid, because the architecture is only as honest as the evidence it is fed, and the evidence is only as honest as the owners who feed it. The next chapter turns to the scale where the seams multiply beyond five: the multiple teams, the dependencies between them, the system-level flow, the integration events, the coordination that must scale from the clinic’s seams to the program’s, which is the subject of the next chapter.*

Notes

• The composite cases remain author-created illustrative material. The Meridian wave-three planning session, the monthly steering review four weeks after the wave-two gate, the five streams’ reports, the December gate twelve weeks out, the construction at seventy-four percent with the certificate of occupancy forecast in eight weeks, the platform’s nine configuration items at three a cycle with the release in three weeks, the migration at forty-one thousand of sixty-six thousand records at two thousand a week, the training at three of six cohorts with two super-users double-booked, the certification subject to the sub-processor amendment, the referral pathway policy, the interface calendar, the cross-method definition of done, the decoupling decision, and all named characters, are teaching constructions consistent with the facts established in earlier chapters: the six-clinic program with the shared platform, the promise of six clinics delivered by December, the grant window at month thirty-six, and the grant’s access outcomes, wait times, and screening uptake from chapters 1 and 5; the grant economics of two hundred fifty thousand units per clinic-month and the sequenced one-clinic-per-quarter plan from chapter 5; the four accountability seats, the governance map, the external reviewer, the conditional kickoff with Nora’s operational ownership, Hana and Nora’s workflow standard decision, and the data migration plan approval from chapter 8; the definition of done spanning the seams, the migration reconciliation at zero unexplained gaps, the product and project scope separation, and the privacy certification from chapter 10; the wave structure with clinic one live and clinics two and three as wave two, the ninety-five-percent-trained, thirty-five-percent-adopted gap, the super-user program, the adoption corridor above eighty-five percent by week sixteen, the week-eight and week-sixteen gates, the terminal moved where the paper was, the transition target, the change fatigue from the facility consolidation and the new staffing model, Sam Otieno, Thelma, and the classroom competency check lesson from chapter 11; the hybrid choice at month eight, the predictive construction, the iterative workflows, the incremental rollout, the adaptive platform inside gated releases, the seam discipline, the release-and-gate map with the wave gate, the capability releases, the certification gate, and the adoption corridor twelve weeks after, the seam evidence as the contract’s nonnegotiable, and the vendor’s implementation lead from chapter 13; the vertical slices and the deliverable-oriented decomposition with the certification evidence and the training attached to the module from chapter 14; the roadmap, the milestone dictionary, the milestones as evidence, the integrated plan layers, and the two-pocket forecast discipline from chapter 16; the schedule and flow models of time from chapter 17; the cost categories, the time-phased budget, and the funding gates from chapter 18; the capacity plan, the double-booked super-users, the training room, the migration files, the test datasets, the reference tables gated by the privacy review, and the test environments from chapter 19; the make-or-buy, the fixed price on the integration and the migration, the time-and-materials on the build, and the acceptance evidence pack from chapter 20; the acceptance matrix and the clinical escalation competence as a precondition from chapter 21; the trigger discipline from chapter 22; the recovery time objective and the recovery point objective and the exercised plan from chapter 23; the wave-two pre-gate review, the privacy certification held by the fourth sub-processor, the records of processing, the accessibility report, the continuity plan, the escalation rehearsal under hypercare, the invoice workflow with the one-signature control, the obligations register, the assurance map, the floor rows, the finance director’s parallel-connection proposal, and the steering decision held on the record from chapter 24; the heroics coordinator, the migration window for the existing clinics, the vendor’s change window closing Friday, the double-booked wave-three trainers, and the corridor commitments from chapter 25; the readiness dashboard with its five windows and five owners and one gate, the provisional and the nonnegotiable, the conditional release with the control, the owner, the date, and the consequence, and the wave-three decision room from chapter 29; the hybrid clock of the seam calendar and the clock mismatch, the gate that waits for the cadence meeting’s month, and the slice that ships past the gate’s date from chapter 30; the interface register, the baseline-change record, the formal change path, the progress rules, and the ninety-percent syndrome from chapter 31; the empirical loop, the product goal, the outcome-based backlog, the definition of done, the flow forecast, and the pseudo-agile behaviors from chapter 32. The teaching numbers introduced here are fully reproducible from the text: the migration, sixty-six thousand records total, forty-one thousand migrated, twenty-five thousand remaining, sixteen thousand active and nine thousand archive, at two thousand records a week, the active set clearing in eight weeks, the whole set in twelve and a half weeks, and the archive in four and a half weeks after opening; the construction at seventy-four percent with the certificate of occupancy forecast in eight weeks against the gate’s twelve, four weeks of float; the platform’s nine items at three a cycle, about three cycles, six weeks, with the release in three; the training at three of six cohorts with the final four weeks once the workflows finalize; the certification, the amendment in two weeks, the assessment in six, about eight weeks; the gate twelve weeks out; and the economics, three clinics at two hundred fifty thousand units per clinic-month, a month’s slip at seven hundred fifty thousand units, all stated with their assumptions as teaching judgments rather than measurements. The standards references are described in the book’s own words: the current official Scrum Guide, November 2020 edition, by Ken Schwaber and Jeff Sutherland, for the adaptive stream’s definition of done and its events, and the general international guidance on project management, ISO 21502:2020, whose advice to tailor the delivery approach to the context of the work underlies the hybrid’s premise, the same premise the PMBOK Guide, Seventh Edition, Project Management Institute, 2021, makes when it groups development approaches as predictive, hybrid, and adaptive with delivery cadence as a separate dial; all are cited for their structure, never as guarantees of success.