My research is organised into twelve branches under a single framework, and a framework like that is only as sound as the weakest branch it gathers. I am Dr Sepehr Abrishami, and the season of Reimagining AEC on the ISO standards turns here to the plans every delivery team keeps. This episode is about the delivery plans, the task information delivery plan and the master information delivery plan, as ISO 19650 Part 2 sets them out at clauses five point four point four and five point four point five. A master plan has the same weakness as my framework. It is built from smaller plans, and it holds only if each of them does.
Denholm District Council, which is hypothetical, is procuring the Ferrybank Community Hub, a two-storey community building. Its lead appointed party is Marrow and Vale Architects, which carries its own architecture task team, and two appointed parties complete the delivery team. They are Kirkwall Structures and Penhale Building Services. The project has just begun Stage 3, spatial coordination, in the Plan of Work of the Royal Institute of British Architects, or RIBA, and the Stage 3 information delivery milestone is Friday the eleventh of December 2026. On Friday the sixteenth of October, the person at Marrow and Vale who carries the information management function puts the three task team plans side by side. Each plan is internally consistent. Each one meets its deadline. And once they are combined, one container cannot be finished in time, however the work is arranged. It is five working days short. The lead appointed party now has a choice. It can re-date one line so that the master plan looks consistent, and the council sees a plan that meets its milestone. Or it can tell the council, in October, that a milestone in December is at risk. Only the second is what the standard asks for.
So the question for this episode is how many task team plans become one master plan whose dates hold, and what the lead appointed party must do when they cannot hold. By the end you will be able to build a task plan that records what the standard asks of every container, set its dates backwards from the milestone, aggregate several plans and read the result for two different kinds of conflict, and carry out the three duties that follow, including the one that sends a problem upward.
Six steps get us there. The first three stay with a single task team, covering who owns the plan, the six things it records, and the route from a scope to a schedule. The last three move up to the lead appointed party, covering the aggregation, the arithmetic that tests it, and the duties once the master plan exists. Ferrybank stays with us throughout.
The first step is ownership, and it comes first because every later question about a delivery plan assumes that somebody knows whose plan it is. Part 2 sets out eight activities, and the delivery plans belong to the fourth, appointment, which is carried out for every appointment on the project. Inside it, the order matters more than it looks. The lead appointed party first establishes the detailed responsibility matrix, which gives every information container an owning task team and a point of exchange. It then sets out its information requirements for each appointed party, and so for each task team, meaning what must be delivered, to what level of information need and against which acceptance criteria. Only then does each task team establish its plan. And only after that does the lead appointed party aggregate the task plans into the master plan. A task plan written before the matrix has settled is a task plan written twice. The episode on roles and the responsibility matrix takes the matrix itself much further.
Read clause five point four point four slowly and it settles three questions before it says anything about dates. The first is ownership. Each task team establishes its own plan. Three task teams means three task plans, because the team that produces the information is the only party that knows how long its own work takes. On Ferrybank that means one plan each from Marrow and Vale's architecture team, Kirkwall and Penhale, and none of them written by the lead on another team's behalf. The second is duration. The plan is established and maintained throughout the appointment, and both obligations sit in the same sentence. A plan that was accurate at appointment and untouched since does not meet the requirement. The third is the level of planning. The plan works at the level of the information container. Not the discipline, not the work package, not the drawing set. Every container the team is responsible for gets its own row.
Before a task team schedules anything, the clause lists five matters it has to weigh, and they fall into two groups. Three come from outside the team and fix what the plan has to achieve. They are the project's information delivery milestones, the team's own lines in the detailed responsibility matrix, and the information requirements the lead appointed party has given it. The milestones come first because a task plan works backwards from the dates the project must meet, not forwards from the date the team would like to begin. The other two matters concern the team's ability to achieve it. One is the resources shared across the delivery team, such as a survey model that two teams need in the same fortnight. The other is the time the team needs to produce the information, and here the standard is specific. Production covers four activities, which are generating the information, coordinating it, reviewing it and approving it. Ask most teams how long a model will take and you hear only the generation time. Coordination takes calendar time and waits on other teams. Review needs someone other than the author, and approval needs a named person with authority to release, available on the day. Price generation alone and the shortfall does not disappear. It surfaces at the milestone, in front of the appointing party.
Two task plans can look identical on the day they are signed. One is filed to satisfy the contract and opened again only when a milestone is missed. The other is reviewed on the same cycle as the delivery programme, so the lead appointed party hears about a slip from the task team first. Only the second meets the clause.
Many practitioners learned these plans under PAS 1192-2, Publicly Available Specification 1192 Part 2, the withdrawn British specification for building information modelling, or BIM, in the delivery phase. Both names carried across unchanged, which is the trap, because the 2018 standard gives the task plan a defined content and the master plan a defined set of duties. PD 19650-0, the transition guidance, since withdrawn and cited here as history, mapped old onto new, and the episode on the UK BIM Framework and the move from PAS 1192 covers that history. The same task plan also appears in the operational phase under ISO 19650-3, planned against asset management milestones rather than project ones, with materially the same container fields. A team that can plan for one can plan for the other.
So the first step resolves like this. The task plan belongs to the team that does the work, it lives for the whole appointment, and it is built one container at a time from inputs that already exist.
The second step is the content of the plan, which is the heart of the subject, because a master plan can only aggregate what the task plans contain. For every container, the plan records six things. The name and title. The predecessors or dependencies. The level of information need. The estimated production duration. The information author. And the delivery milestones. Take one Ferrybank container through all six. It is Kirkwall Structures' structural analysis model at Stage 3.
Name and title is two things, not one. The identifier follows the naming convention in the UK National Annex to Part 2 and makes the container unique across the project. The title is what a person reads. Without the title nobody can discuss the container in a meeting, and without the identifier nobody can find it in the common data environment. Dependencies come next. The analysis model waits on the architectural general arrangement at the same stage and on the ground investigation report. The instruction that matters is to record dependencies that sit outside your own task team as well as inside it, because those are the ones that fail, and they are the only ones the lead appointed party can do anything about.
Level of information need is the field most often filled in with the wrong vocabulary. It records how much information the container must carry when it is delivered, in extent and granularity. It is not a maturity band and not a level of detail. The governing standard is ISO 7817-1, which superseded and withdrew BS EN 17412-1, and a live plan still citing the withdrawn document is a finding waiting to be raised. The task plan does not derive this field. It records what the lead appointed party's requirements specify, and where they are silent, the team raises a query rather than guessing. The episode on level of information need in ISO 7817-1 covers the method.
The estimated production duration is the field the master plan cannot be built without. A date says when a container will appear. A duration says how much time the team needs to make it appear, and only the duration survives a change of programme. When a milestone moves, a plan built on dates has to be rewritten line by line. A plan built on durations and dependencies recalculates. Kirkwall's entry for the analysis model is fifteen working days, covering all four production activities, with the architectural general arrangement available at the start of that window. That single number is what lets the lead appointed party test whether a milestone can be met at all.
The information author is the party responsible for producing the container. Name an individual where the team can commit one, and a defined role where it cannot. If nobody can be named against a container, that container is not yet planned, and saying so is more useful than filling the cell. The delivery milestones say which information delivery milestone or milestones the container serves. On UK projects these usually attach to the stages of the RIBA Plan of Work 2020, and the later stages are worth getting right. Stage 5 is manufacturing and construction, Stage 6 is handover and Stage 7 is use. Where one container serves more than one milestone, record each, not only the earliest.
Now put the six against a real template. Marrow and Vale's inherited template carries four of the six. Two are missing. It carries internal start, internal completion and planned issue dates, from which a duration could be worked out, but working it out is not stating it, and nothing downstream can rely on an inference. And it records who self-checked and who approved each container, which names the people giving assurance rather than the party responsible for production. Two added columns are the whole fix.
The same reading turns up something the clause does not ask for at all. A task plan is not required to carry a status code. Every status column in a template is added management content, and it can be tuned freely. It can also carry defects the clause never introduced, and many templates still list the B status codes that the 2021 revision of the UK National Annex deprecated. The episode on the common data environment deals with status codes properly.
That resolves the second step. Six fields, recorded for every container, with the duration and the author stated rather than inferred, and everything else in the plan understood as a choice rather than a requirement.
The third step is getting from a scope to a schedule, and the governing principle is that the scope comes first and the dates come last. The team reads its own lines in the detailed responsibility matrix, not the fee proposal or the brief, because the matrix defines which containers it owns. It turns that scope into a container list, one row for every model, drawing, schedule, report or specification it delivers. It puts a duration against each one, estimated by the team itself in working days under conditions it can state out loud. It chains the dependencies, looking outside its own team as it does. It names an author for every container. Only then does it set dates. A team that starts with the dates produces a plan that describes its hopes rather than its work. On Ferrybank, Kirkwall's scope becomes eight containers, each identifier carrying Kirkwall's own originator code.
The dates are set backwards, and this is where most plans go wrong. The information delivery milestone is not the task team's deadline. On Ferrybank the milestone, Friday the eleventh of December, is the point by which the council has reviewed and accepted the Stage 3 information model. The council needs ten working days to review and accept, so the delivery team has to submit by Friday the twenty-seventh of November. Before that, Marrow and Vale needs five working days to review and authorise the model, so the task teams must deliver to it by Friday the twentieth of November. That is Kirkwall's real date. Three weeks, fifteen working days, separate the milestone printed on the programme from the day the work must be finished, and a task plan dated to the milestone is late on the day it is written.
A duration stops being credible in two familiar ways. The first is coordination priced as effort. A team prices two days of coordination and books two days of programme, when coordination waits on other teams issuing and on a clash review held on a fixed day, ready or not. The second is a shared resource assumed to be free, such as a specialist analyst or a technical authority who has to sign something. The fix for both is the same. State the assumption in the plan, and record the shared resource as a dependency, so that the party able to resolve a conflict can see it.
Dependencies come in two kinds, and only one matters to anyone else. Inside the team, a dependency is sequencing the team can fix itself, such as Kirkwall's superstructure model waiting on its own analysis model. Outside the team, it is sequencing the team cannot fix, such as the analysis model waiting on Marrow and Vale's general arrangement. Record the outside kind precisely, by the predecessor's identifier and owning team, with when production can begin once it arrives and what the team will do if it arrives late.
Because the plan is maintained, the practical question is what obliges an update, and the principle is that anything changing an input to the plan changes the plan. Some triggers come from outside the team, such as a milestone moving, a requirement changing a level of information need, or a predecessor slipping. Others come from inside, such as a container being split or withdrawn, an author leaving the appointment, or a duration proving wrong once production starts, which is information rather than failure. If any of these happens and the plan does not move, the lead appointed party is aggregating a document that misdescribes the work.
So the third step resolves. A task plan is an input to someone else's arithmetic. A missing duration becomes an estimate somebody makes on the team's behalf, and a missing dependency becomes an assumed independence, which is the most expensive assumption in delivery planning.
The fourth step moves up a level, to the master information delivery plan, and the single most useful word in clause five point four point five is aggregate. The lead appointed party aggregates the task plans to establish the delivery team's master plan. It does not write a schedule and then check the task plans against it. An authored plan carries dates nobody committed to. An aggregated plan carries dates each team put its name against, so a conflict shows up as arithmetic rather than as an argument. It also settles accountability. If a team misses a date the lead invented, the plan was wrong. If it misses a date it gave, the work slipped, and only that is a problem anybody can manage.
The clause names four matters the lead appointed party weighs as it aggregates. Two look sideways at the task teams. The detailed responsibility matrix is the check that every container in the delivery team's scope appears in somebody's task plan, and in only one. The dependencies between task teams are where aggregation earns its place, because a cross-team dependency is invisible inside any single plan. The other two look upward. They are the time the lead appointed party needs to review and authorise the information model, and the time the appointing party needs to review and accept it. Neither appears in any task plan, because neither belongs to a task team. The aggregation adds them, and that is why a master plan is always tighter than the task plans that fed it. On Ferrybank these are the two windows we met in the third step. Marrow and Vale's five working days run from Monday the twenty-third to Friday the twenty-seventh of November, and the council's ten run from Monday the thirtieth of November to the milestone itself.
Where the master plan goes next is what gives it force. Under clause five point four point six it is one of the documents taken into account in the lead appointed party's completed appointment documents, alongside the BIM execution plan it accompanies, and it is managed under change control for the rest of the appointment. So a change to the master plan is a change to an appointment document. It follows the project's change procedure, the appointing party can see it, and the delivery team cannot adjust it quietly among itself when the dates stop working. The episode on the BIM execution plan covers that document set.
Two cases catch people out. The first is the small appointment. A lead appointed party with no appointed parties of its own still needs a master plan, and in that case the master plan is in effect its task plan. Guidance Part A, the UK guidance on the information management function and resources, works this through with the example of a client's project manager, an organisation that is a whole delivery team in itself, with nobody to aggregate from. It still produces information against requirements, so it still needs a BIM execution plan and a master plan, although both are likely to be less detailed. What scales with the appointment is the content of the plans, not whether they exist. The second case is a chasing failure rather than a knowledge failure. Two task teams submit their plans and one does not. The master plan is built from what arrived, the gap is logged as outstanding rather than as a risk to a milestone, and the late plan eventually arrives carrying dates that do not fit. The conflict then surfaces at the milestone instead of at the aggregation. The clause expects the aggregation to be completed. Every team is chased, and a missing plan is treated as a risk to a milestone, which is a category that obliges somebody to be told.
That resolves the fourth step. The master plan is collected, not composed, it adds two windows that belong to nobody below it, and once it sits among the appointment documents, changing it is a visible act.
The fifth step is the arithmetic, and Ferrybank now earns its place. The method is four moves and a calendar. First, align the references, bringing every container from every task plan into one register and checking each against the detailed responsibility matrix for gaps and duplicates. Second, chain across the teams, connecting every dependency that names another team's container to the date that container is planned in the plan that owns it. Third, add the two review windows between the task team delivery date and the milestone. Fourth, read the whole set against the milestone, checking every planned issue date against the delivery date and every duration against the time its owning team has.
Here are the three plans, nineteen containers in all. Marrow and Vale's architecture team has six containers and starts first, on Monday the seventh of September, because it issues the spatial coordination model that both other teams wait on, and that model arrives on Friday the second of October. Kirkwall has eight containers and starts on Monday the fifth of October. Penhale has five containers and cannot start until Monday the twelfth of October, because its appointment began five weeks after Stage 3 did. Every plan appears to deliver by the twentieth of November. Read one at a time, all three are fine.
Chain them, and the first conflict appears. Penhale's building services coordination model is planned for issue on Friday the thirteenth of November, and it waits on Kirkwall's superstructure model, which Kirkwall's plan shows arriving on Friday the twentieth. A container due on the thirteenth depends on one that does not exist until the twentieth. Two further containers, the structural and the architectural coordination check reports, wait on Penhale's model in turn, so the collision spreads into both other teams. No single plan could have shown this. Each plan is right, and only the combination is wrong.
This first conflict can be resolved inside the delivery team. Kirkwall's plan records that a second engineer becomes available from Monday the twenty-sixth of October, so its substructure and superstructure models can be produced in parallel rather than in sequence. The superstructure model comes forward to Friday the sixth of November and the collision disappears. Notice why that route existed. Kirkwall stated its resource assumption in the plan, exactly as the third step asked. A team that had kept the same assumption in someone's head would have offered the lead nothing to work with.
Before the second conflict, separate the two kinds, because confusing them is the costliest error in the subject. The first kind is a sequencing problem. There is enough time in total, but it has been allocated in the wrong order, and the delivery team can usually fix it by resequencing, parallel working or moving a dependency. The outcome is a change notified to a task team. The second kind is a shortage of time. The work does not fit the window however it is arranged, because the constraint is the size of the window rather than the order inside it. The delivery team cannot fix that on its own.
Now the second conflict. Penhale's appointment begins on Monday the twelfth of October. Count the working days from then to Friday the twentieth of November and you get thirty. Penhale's own plan gives the coordination model a production duration of thirty-five working days. Its planned date said the thirteenth of November. Its stated duration said something else, and this is why the second step insisted on durations being stated, because a date can simply be typed in and a duration has to be worked out. Even starting on the first morning, with no wait for any predecessor, the earliest the model can be finished is Friday the twenty-seventh of November. Five working days short. Resolving the first conflict does not help, because the dependency was never the binding constraint. The duration belongs to Penhale and cannot be cut without its agreement. The appointment start has already happened. And the milestone belongs to the council.
So the test for whether a milestone can be met is three questions, asked of every container in this order. Do its predecessors arrive before it needs them? Does its own duration fit between the earliest date it can start and the date it is due? Does the owning team have that many working days inside its appointment at all? The third is the question people skip, and it is the only one that catches a late appointment, which is the most common source of the second kind of conflict. The order matters because asking the third question first tempts you to adjust durations until the sums work, and adjusting a duration changes a team's commitment without asking the team. Asked last, it arrives after everything the delivery team can legitimately do has been tried, so whatever remains is unresolvable rather than inconvenient.
That resolves the fifth step. One calculation produced two conflicts. The first stays inside the delivery team as a notified change. The second has become a fact about the milestone, and somebody outside the delivery team has to be told.
The sixth step is what clause five point four point five asks once the master plan is established, which is three duties, in an order where each assumes the last. Baseline the plan. Inform each task team and notify any change its plan needs. And inform the appointing party of any risk or issue that could affect the information delivery milestones.
Baselining is often treated as filing, and it is not. A baseline fixes two things. It fixes the deliverables, meaning which containers the delivery team has committed to for this milestone, so that a container added later shows as a change. And it fixes the dates, so that a date that moves afterwards moves against a recorded position, and variance becomes a quantity someone can be asked about rather than an impression. Because the master plan is an appointment document under change control, the appointing party can see that it changed.
The notification to a task team decides whether the plan stays cooperative. It is not a criticism. Kirkwall's plan was correct in every respect, and became a problem only beside Penhale's, which is a fact about the aggregation. A workable notification carries four things. The container affected, by its identifier. The change required, stated as a new date or sequence rather than an outcome. The reason, naming the other team's container that caused the conflict. And the date by which the updated plan is needed back. Kirkwall then updates its own plan, which is exactly why maintenance was a requirement in the first step and not a courtesy.
The third duty is where the whole subject lands. The clause does not say inform the appointing party once a milestone has been missed. It does not say inform it if the delivery team judges the matter serious enough. The trigger is a risk or issue that could affect a milestone, and on Ferrybank that condition was met on the sixteenth of October, the day the aggregation was completed. Notice that the duty protects the lead appointed party itself. A lead appointed party that reports a five-day shortfall in October is managing an appointment competently. One that discovers the same five days in December has failed to manage it. The facts are identical. Only the timing differs, and the timing was the lead's to control.
So go back to the choice from the opening, because both options look like plans. The quiet reschedule re-dates Penhale's model to the twenty-seventh of November so the register looks consistent, tells Penhale to compress a duration it has already committed to, and absorbs Marrow and Vale's own review window to make the sums close. The council sees a master plan that meets its milestone, because that is what it was shown. The shortfall comes back in December, with fewer options and spent credibility. The compliant response records the model at its real duration and its real earliest completion. Penhale keeps a duration it can defend, so the number stays reliable the next time anyone uses it. Marrow and Vale informs the council of a risk to the Stage 3 milestone. And the council decides, because moving a milestone is its decision and nobody else's.
What keeps a master plan true afterwards is a rhythm with four triggers. Weekly, compare planned issue dates against what has reached the common data environment. On every task plan update, because a revised duration or a new author changes an input to the arithmetic. On every change to a milestone, which moves both review windows and means the third question of the achievability test has to be asked again. And on every appointment change, because a late appointment is where the unfixable conflicts come from. Under ISO 19650-3 the operational phase carries the same duties, and there the lead also adds or updates risks in the delivery team's risk register.
A forward look, labelled as such. Draft revisions of Parts 1 and 2 were open for public comment from the tenth of March to the third of May 2026. A draft revision of Part 3 followed, open for comment until July 2026. The drafts propose a single information management process across delivery and operation, and they propose renaming the BIM execution plan as the information production plan. None of them is in force, and no publication date has been announced. The 2018 clauses are the ones to plan, teach and audit against.
That resolves the sixth step. Baseline so that change is visible, notify so that each task team can re-commit, and inform the appointing party at the moment the arithmetic shows a risk, not at the moment a milestone is missed.
The episode is drawing together, so back to the sixteenth of October. The question was how many task team plans become one master plan whose dates hold, and what the lead appointed party must do when they cannot. A master plan is only as sound as the weakest task plan it gathers, so the answer starts below it. Each task team keeps its own plan for the whole appointment, six fields for every container, with the duration and the author stated. It sets its dates backwards from the milestone through two review windows it does not own. The lead appointed party aggregates those plans rather than writing its own, chains the dependencies, adds the two windows and tests every container with three questions in order. A sequencing conflict it resolves inside the delivery team and notifies. A shortage of time it reports to the appointing party as soon as the arithmetic shows it, because only the appointing party owns the milestone.
Three checks are worth making on every plan you touch. When you read a delivery plan, check for the duration and the author in every row, because a plan with dates and no durations cannot be recalculated and a container with no author has not been planned. When you aggregate, check the length of each appointment last, and never adjust a team's duration to make the sums close. And when the arithmetic fails, check the date on the report to the appointing party, because the duty is met by a report written in October, not by an apology written in December.
For reading, start with clauses five point four point four, five point four point five and five point four point six of ISO 19650-2, with its UK National Annex, then ISO 7817-1 for the level of information need column, PD 19650-0, withdrawn guidance now read as history, for the move from the 1192 suite, ISO 19650-3 for the operational variant, and Guidance Part A for who produces each resource. The episodes on the BIM execution plan, the delivery phase in ISO 19650 Part 2 and the common data environment sit closest to this one. The next episode, Roles and the Responsibility Matrix, is about who answers for each information management task, and how you prove it.
From me, Dr Sepehr Abrishami, and from Reimagining AEC, one piece of arithmetic to do. On the next delivery plan you write, receive or study, find a date that nobody has tested against a duration, and test it yourself before a milestone does it for you.