Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Reimagining AEC

Reimagining AEC artwork

Information Delivery Manuals in ISO 29481

Nov 10, 202642 minAudio
ShareXLinkedInFacebookWhatsAppEmail

On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks how an information exchange can be specified precisely enough that every party can agree to it and software can check it. A hypothetical college teaching block, where a cost consultant receives room data it cannot trust, carries the argument through the three components of an information delivery manual under ISO 29481-1:2025, the use case and its header, interaction maps, transactions and process maps, exchange requirements built from information units and constraints, the six standard project phases and the standards around the manual, and the step from natural language to exchange view definitions, Industry Foundation Classes and the buildingSMART Information Delivery Specification. Chapters 0:00 Opening: the room data a cost consultant cannot trust 3:09 Why specify delivery, the three components and what the third edition changed 7:21 The use case description and the common header 10:27 The business context: interaction maps, transactions and process maps 18:49 The exchange requirement: preconditions, information units, constraints and the level of information need boundary 27:11 Project phases, the phase mapping and the standards around the manual 33:03 From natural language to machine readable: exchange view definitions, Industry Foundation Classes and the Information Delivery Specification 38:56 Close After listening, you should be able to: Explain why information delivery has to be specified rather than assumed, name the three components clause 5.1 of ISO 29481-1 requires in order, and state what the 2025 edition changed from the withdrawn 2016 text. Write a use case description with a complete common header and the five required statements, and test an aim as a purpose rather than a deliverable. Analyse a business context with an interaction map and a process map, name the initiator and executor of each transaction, expand a transaction into its messages, and identify where the exchange requirement attaches. Specify an exchange requirement as preconditions, nested information units with unambiguous descriptions, and type and rule constraints, and state where it stops and the level of information need framework of ISO 7817-1 begins. Map the six standard project phases of ISO 22263 to a local plan of work in one of the three permitted patterns, and route a practitioner's question to the standard that governs it, distinguishing an exchange requirement from an exchange information requirement under ISO 19650. Describe what an exchange view definition does and does not do, state the current positions of ISO 16739-1:2024 and the buildingSMART Information Delivery Specification version 1.0, and explain why open information exchange is more than open file transfer. References British Standards Institution. (2018). BS EN ISO 19650-1:2018 Organization and digitization of information about buildings and civil engineering works, including building information modelling (BIM). Information management using building information modelling. Part 1: Concepts and principles. BSI. British Standards Institution. (2018). BS EN ISO 19650-2:2018 Organization and digitization of information about buildings and civil engineering works, including building information modelling (BIM). Information management using building information modelling. Part 2: Delivery phase of the assets (incorporating corrigendum February 2021, including UK National Annex). BSI. British Standards Institution. (2022). BS EN ISO 19650-4:2022 Organization and digitization of information about buildings and civil engineering works, including building information modelling (BIM). Information management using building information modelling. Part 4: Information exchange (including UK National Annex). BSI. British Standards Institution. (2024). BS EN ISO 7817-1:2024 Building information modelling. Level of information need. Part 1: Concepts and principles. BSI. British Standards Institution. (2026). PAS 1958:2026 Built environment data and information standards landscape. Guide. BSI. buildingSMART International. (2024). Information Delivery Specification (IDS) version 1.0. https://www.buildingsmart.org/standards/bsi-standards/information-delivery-specification-ids/ buildingSMART International. (2024, June 4). Information Delivery Specification (IDS) v1.0 is approved as a final standard. https://www.buildingsmart.org/?p=30065 International Organization for Standardization. (2008). ISO 22263:2008 Organization of information about construction works. Framework for management of project information. ISO. International Organization for Standardization. (2012). ISO 29481-2:2012 Building information models. Information delivery manual. Part 2: Interaction framework. ISO. (Withdrawn, cited as history.) International Organization for Standardization. (2013). ISO/IEC 19510:2013 Information technology. Object Management Group Business Process Model and Notation. ISO. International Organization for Standardization. (2016). ISO 29481-1:2016 Building information models. Information delivery manual. Part 1: Methodology and format. ISO. (Withdrawn, cited as history.) International Organization for Standardization. (2018). ISO 16739-1:2018 Industry Foundation Classes (IFC) for data sharing in the construction and facility management industries. Part 1: Data schema. ISO. (Withdrawn, cited as history.) International Organization for Standardization. (2020). ISO 21597-1:2020 Information container for linked document delivery. Exchange specification. Part 1: Container. ISO. International Organization for Standardization. (2022). ISO 29481-3:2022 Building information models. Information delivery manual. Part 3: Data schema. ISO. International Organization for Standardization. (2024). ISO 16739-1:2024 Industry Foundation Classes (IFC) for data sharing in the construction and facility management industries. Part 1: Data schema. ISO. International Organization for Standardization. (2025). ISO 29481-1:2025 Building information models. Information delivery manual. Part 1: Methodology and format. ISO. International Organization for Standardization. (2025). ISO 29481-2:2025 Building information models. Information delivery manual. Part 2: Interaction framework. ISO. Royal Institute of British Architects. (2020). RIBA plan of work 2020 overview. RIBA. Reimagining AEC is Dr Sepehr Abrishami's own podcast. For related subjects, hear Information Management in ISO 19650 Part 1: Concepts and Principles.

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show.

Information on a project goes wrong most often at the moment it changes hands. I am Dr Sepehr Abrishami, and in six years of architectural practice before I moved into research, much of what my office drew went on to the people who would cost and build it, so in a season of Reimagining AEC on the ISO standards that handover gets an episode of its own. This episode is about information delivery manuals, as ISO 29481 Part 1 sets out their methodology and format.

Here is the handover going wrong, in a case that is hypothetical. A further education college is building a new teaching block inside the Northgate Quarter regeneration area. The college estates team is the client. An architectural design consultant holds the model, and a cost consultant has to prepare the cost plan for the design stage. The cost consultant asks for room data, and the request looks trivial, because every room is in the model. Then the data arrives. The areas exist, but different authors measured them to different faces of the walls. Each room's function is recorded three times, in its name, in a property and in a schedule, and the three do not agree. Nobody has written down which value governs, because nobody was ever asked to. So the cost consultant faces a choice with no good answer. It can price from numbers it does not trust, or it can send the data back and wait for a re-issue while the date for the cost plan stays where it is. And the same conversation failed to happen on the college's two previous projects. The information was in the model the whole time. What was missing was a statement of what was needed, from whom, in what form, and at what point.

So the question for this episode is how you specify an information exchange precisely enough that every party can agree to it and software can check it. ISO 29481-1 answers with a document it calls an information delivery manual. By the end you will be able to write one for the college. You will state the purpose before the information, name who asks and who answers, prescribe the information as units with rules that cannot be read two ways, and keep that requirement separate from the file format and the software that will test it.

We will take it in six steps. First, why delivery has to be specified at all, and what the current edition changed. Then the use case, which now comes first. Then the business context, drawn in two ways. Then the exchange requirement, down to its information units and constraints. Then the phases of a project and the standards around the manual. And last, how a natural language specification becomes something a machine can read, which is where an old question about two sets of matching initials gets its answer. The college stays with us throughout.

The first step is the case for specifying, because a client who cannot make that case will never give anyone the time to do the work. An appointment usually names a deliverable, such as a model at a stage, a schedule or a report. It rarely names the information inside that deliverable, item by item. The producing party makes reasonable choices. The receiving party makes different reasonable assumptions. Neither is careless, and the gap stays invisible until someone tries to use the information and finds that the one field they needed was never filled. The cost is not the missing field. It is the request, the wait, the re-issue and the loss of confidence that follows. So the alternative to specifying delivery is not flexibility. It is rework, arriving later.

The standard's remedy is the information delivery manual, which it defines as a specification of a use case, expressed using business context maps and exchange requirements. A use case is a purpose someone is trying to achieve. The business context is the roles involved and how work passes between them. And the exchange requirements state the information itself. Clause five point one says a manual carries three components, in this order. A description of the use case. An analysis of its business context. And the information needed to support it, set out as one or more exchange requirements. The order is the method. Purpose first, then who does what, and only then the information.

The manual is written in natural language, on purpose. The standard calls it a manual precisely because every stakeholder can understand it. A specification written only in a schema can be validated by software, but the client whose need it claims to capture cannot check it. Plain language can be read, argued over, corrected and agreed. The machine-readable forms come afterwards, derived from the agreed text. They implement the specification. They do not replace it.

Now the edition, because much of what is written about this standard describes a version that has gone. Part 1 has had three editions, in 2010, 2016 and 2025, and the 2025 edition superseded the 2016 text, which the UK adopted in 2017 and which is now withdrawn. Its foreword lists three changes. It covers infrastructure as well as buildings. Its terminology is aligned with related standards, in practice ISO 19650, so the two families now share words such as information container and information model. And its text was revised for readability, which understates what happened, because the framework was restructured. In the second edition a manual comprised the maps and the exchange requirements, and the use case sat outside them as context. In the third, the use case is a component in its own right, and it comes first. The maps survive, but as one way of representing the business context rather than as components in themselves.

Several terms went with the restructure. Transaction map and model view definition are no longer defined terms. An exchange requirement is now defined against a use case rather than against a business requirement at a phase. And one sentence disappeared from the definitions, the note that said a manual could be referred to as an information delivery specification. Hold on to that missing sentence, because the last step of this episode turns on it. None of this invalidates a manual written before 2025. What has moved is the framework you measure new work against, and the vocabulary of an inherited document usually gives its edition away. If it treats transaction maps as components, or says a manual can be called an information delivery specification, it was written against the withdrawn text.

So the first step resolves. The college never lacked information. It lacked a specification, and the current edition says that a specification starts with a purpose.

The second step is the use case description, the first component, and it begins with something people forget. Clause five point two requires every component, the use case, the maps and the exchange requirements alike, to carry a common header. The header holds identifiers, which are a full title, a short title and a globally unique identifier. It also holds authoring information, which is the authors and a change log of creation and every later change, with the responsible author and the date. The unique identifier exists because components are meant to be assembled and reused across projects, which is the whole point of the method. And the change log, the field most often left out, is what lets an appointment name the version it relied on. A header filled in mechanically fails just as surely as an empty one. Identical full and short titles, an identifier copied from a sibling component, a change log created at issue and never touched again. It passes a glance and fails at the first revision.

Beneath the header, the use case carries a short overview and five required statements, which are the aim, the scope, the uses, the regional context and the relevant standard project phases. Each has a typical failure. An aim that restates the deliverable. A scope with no boundary, so the work outgrows the resource. Uses left out, so the requirement can never be argued about. A regional context assumed until it turns out to be wrong. And phases left vague, which the fifth step deals with.

There is a three part test for any aim. Does it name a purpose rather than a deliverable. Would a reader outside the project understand it. And could another project reuse it without editing. Producing a room schedule is a deliverable. Enabling a cost plan to be prepared from design stage room data is a purpose, and it passes all three tests. An aim that fails any of them is a project note, not a use case.

For the college, the aim is enabling a cost plan to be prepared from design stage room data. The scope names its exclusions explicitly, because an exclusion is worth more than a silence. The uses field names two uses, not one, because the same data serves the cost plan and the college's own check of area against its brief. Keep that second use in mind, because the process map in the next step proves it. The regional context is the United Kingdom, with a pointer to the local plan of work. That resolves the second step. A use case with a complete header and five honest statements can be found, versioned and reused, and reuse is the reason the standard asks for all of it.

The third step is the business context, the second component, and it is where responsibility stops being vague. Clause five point four requires a description of the business context, and then says it may carry one or more business context maps. So the analysis is required, and the drawings are the recommended way of doing it. Two kinds are recommended, interaction maps and process maps, used in any combination. Clause four point five point one sets out the choice. Where the information belongs to an activity inside a business process, draw a process map. Where it belongs to a transaction between actors, and the concern is that agreed protocols of communication are in place, draw an interaction map.

An interaction map rests on four definitions. An actor is a person, an organisation or an organisational unit, such as a department. A role is the function an actor performs at a point in time, and the standard adds that a role is determined by action and outcome, not by profession or trade. That is why a map shows roles rather than job titles. An interaction is a communicative action between two roles. And a transaction is a sequence of those interactions in which information moves as messages. The notation comes from the design and engineering methodology for organisations, usually shortened to DEMO.

One rule carries the whole map. Every transaction has one initiator and one executor. The initiator starts it and is the role that needs the result. The executor is answerable for seeing it through. Notice what follows. The initiator is not necessarily the role that produces anything. When a cost consultant needs room data from a designer, the cost consultant initiates and the designer executes, even though all the production work sits with the designer. People tend to draw that the wrong way round the first time, and getting it right is what makes the map worth drawing.

Annex D says that before anything is drawn, you define the scope of interest, meaning which parts of the use case belong in the model. Transactions that cross that boundary are border transactions, with an external role at one end and an internal role at the other. Annex D also filters what belongs on the map at all. Only essential transactions go on it, the ones that create or change what it calls a unique organisational fact, so look for verbs such as deciding, judging, paying or manufacturing. Sending a letter is not one of them. It carries information inside a transaction that is. Apply that filter and most maps get smaller.

Now walk through the college map. Start with the boundary, drawn around the production of information for the teaching block. Inside it sit two roles, the design consultant and the cost consultant. Outside it sits the college estates team, because the college commissions the information rather than producing it. Three transactions connect the three roles, and two of them cross the boundary. In the design commission, the college initiates and the design consultant executes. In the cost commission, the college initiates and the cost consultant executes. The third transaction runs inside the boundary, and it is the room data exchange, in which the cost consultant initiates and the design consultant executes. Read the map back and it answers what the appointments left open. The room data belongs to a transaction of its own, so it needs specifying in its own right rather than being assumed inside the design commission. The cost consultant is the initiator, so the cost consultant must say what it needs, which is exactly the conversation that never happened on the earlier projects. And the room data exchange has to complete before the cost commission can.

An interaction map tells you that two roles have a transaction, not what passes between them or in what order. Clause four point five point three and Annex D fill that in by expanding the transaction into a sequence diagram, usually drawn in the unified modelling language. This is where the second edition's transaction map went. The label was dropped and the work survived. DEMO holds that every transaction follows a standard pattern of four messages, which are request, promise, declare and accept, and the standard treats that pattern as a good default while permitting free modelling of the sequence.

Expand the college's room data transaction. The cost consultant requests room and space data for cost planning, and the request names the purpose. The design consultant promises it by the stage date, which is a commitment rather than an intention. The design consultant then declares it delivered. And the cost consultant accepts it, or returns it with a reason, so the route back is designed in rather than argued about later. Three of those messages coordinate the work. Only the declare message carries the room data, and the exchange requirement attaches to that one. So a four step transaction usually yields one exchange requirement, not four. Interaction maps and their transactions also have a machine-readable form, the interaction framework, specified in ISO 29481-2, whose second edition of December 2025 replaced the 2012 text.

The process map is the other drawing. Clause four point five point two says it describes how activities flow inside one business process, which role each actor plays, and the information required, consumed and produced along the way. Wherever an activity in one role needs input that another role produces, there is an exchange, and it gives rise to an exchange requirement. The recommended notation is business process model and notation, published as ISO/IEC 19510, and Annex C recommends, without mandating, a subset of twenty-two of its symbols. Two rules matter most. Every actor gets a lane, because the lane is how the map identifies the actor. If a role has no lane, any exchange that role was going to make has quietly left the analysis, which is the commonest way an exchange requirement goes missing. And Annex C states the rule that links this step to the next. An exchange requirement appears in a process map as a data object.

Walk the college through it. There are three lanes, one for each role. The process starts when the brief is issued. In the design consultant's lane, the room and space data is produced, and it emerges as a data object, which is the exchange requirement. Follow the flow out of that object and it reaches two activities in two different lanes. In the cost consultant's lane, the stage cost plan is prepared. In the college's lane, area is checked against the brief. One exchange requirement, two consumers, and that is the evidence for the two uses named in the use case.

Set the two drawings against each other. The interaction map showed who is answerable for each transaction, and that the client sits outside the scope of interest. The process map showed the order of work, and that two activities consume the same data. Neither contains the other. Together they justify the exchange requirement, which is what clause five point one means when it calls the analysis of the business context a discovery process, one whose purpose is to show how each requirement was derived. Annex D adds one check worth the time, that the map holds every transaction needed within scope, because that is what catches a missing exchange.

That resolves the third step. Mapping the business context turns a vague sense of who needs what into named roles, a named initiator for every exchange, and one message, or one data object, where the requirement attaches.

The fourth step opens that data object. The exchange requirement is the third component, and it is the part a design team works to. The current definition calls it a defined collection of information units required to fulfil a use case. Clause four point six says it connects process to data, describing information one actor develops and delivers so that another can carry out a downstream activity, and it must be written with a clear grasp of what the downstream actor needs. That rules out writing one from the producer's side alone. The phase has not gone from the method. Clause five point five point one places each exchange requirement within one business process at one project phase.

The same clause asks for very little as a minimum. An exchange requirement shall include at least one information unit and the common header, including a name that indicates its purpose. Almost everything that makes it useful sits in the optional list, which allows sub exchange requirements, descriptive notes, information constraints and a reference link to external schema elements. The name does more work than it looks, because it is how a map points at the requirement and how someone searching a library finds it. A name such as stage three room schedule ties a requirement to one project and one moment. A name such as room data for cost planning still makes sense on the next project. Sub requirements solve a problem the college meets at once. Suppose a services engineer joins and needs the room data plus occupancy, which the cost consultant does not need. A core requirement with a sub requirement for each consumer serves both.

Preconditions come first. Clause five point five point one says a precondition is itself an exchange requirement that has to be complete before the current one can run, and its example widens that to the resources or competencies a team member needs. The college needs both kinds. The design commission has to be in place and the storey structure fixed, because room identifiers cannot be stable until the storeys are. The naming convention has to be agreed and issued, or two authors will diverge. The measurement rule for area has to be agreed, which is the direct fix for the failure the college started with. And a competent person has to be available to produce the data.

Then come the information units. An information unit is a description of a piece of information, and the standard's examples run from an inspection report to a window identifier to a room depth, so the scale is yours to choose. Clause five point five point two sets the rule that makes the method work. Each exchange requirement shall be prescribed as a set of information units. If you cannot express what you need as a set of units, you do not yet know what you need. The same clause adds two rules that work as a pair. Each unit deals with one concept only. And units are typically nested in tabular form, with a unit identifying a real-world object and its required unique identifier, beneath it units for that object's attributes, and beneath those, where needed, units for its component parts.

The college uses two of those levels. Building, storey and room are objects, each with its identifier. Room name, function and net internal area are attributes of the room. Function is now one unit holding one value, which ends the three-way disagreement between a name, a property and a schedule. There is no component part level, because the requirement is about spaces, not the walls that enclose them, and knowing where to stop is part of the skill. The building identifier is not defined as a number in a field. It is the unique reference in the college's estate register, which tells the producer where to find it. Ask for room data and the consumer gets whatever the producer judged relevant. Ask for room name, function and area, and the consumer gets three values per room, joined by nothing. Ask for a room object carrying an identifier, with its attributes beneath it, and the consumer gets a joined set it can check. Two rooms sharing a name is ordinary. Two sharing an identifier is an error, and only the identifier makes it findable.

Clause five point five point two also asks for descriptions explicit enough not to be confused with neighbouring concepts. Room area, or area in square metres, leaves the measurement rule to the reader, and two experienced people will supply two different rules, which is exactly what the college's earlier projects received. A working description states the rule. The floor area of the room, measured to the internal face of the enclosing elements, excluding the area of internal columns. Two authors reading that produce the same number. A unit that names a file type, a schedule or an export setting fixes how the information arrives without saying what it contains. Then a delivery in another format with exactly the right content satisfies nothing, while the right format with the wrong content passes.

Information constraints, at clause five point five point three, do two jobs. The first is the data type, such as text, a number, a two-dimensional graphic or a three-dimensional model, extending into units of measurement, precision and permitted values, which a machine can check once the requirement is implemented. The second is the context, rules and restrictions on using the unit, and the standard's examples are all rules with force. A precast element carrying at least three identifiers. A room serving a given purpose being no smaller than a stated area. A panel no larger than a normal truck can deliver. Each records something somebody learned at a cost, so the drafting question is what went wrong last time and which constraint would have caught it. At the college, the room identifier constraint asserts uniqueness within the building and points to the naming convention. The net internal area constraint repeats the measurement rule, deliberately, because the description tells a reader what the unit is, and the constraint tells a producer how to arrive at the value.

One boundary has to be drawn here, because the same people write two closely related documents at the same point in a project. The exchange requirement specifies which information units are needed to fulfil a use case, between named roles. The level of information need framework, in ISO 7817-1, specifies how much information is needed, for a purpose, at a milestone, across geometrical information, alphanumerical information and documentation. Which units, against how much. The introduction to ISO 29481-1 presents that framework as a fuller basis for deciding what the units should hold, so it informs the units and does not replace the requirement. Do not drop a level of information need setting into an exchange requirement as if it were a constraint, because it is neither a data type nor a rule of use. Reference across rather than restating, and record in each document what it does not cover. The episode Level of Information Need in ISO 7817-1 takes that framework much further.

So the fourth step resolves. The cost consultant at the college now holds a requirement it can test, with preconditions, nested units that each hold one concept, descriptions that cannot be read two ways, and constraints that carry the rule.

The fifth step places the manual in time and among its neighbours. Clause five point three point one requires the use case to state the standard project phases, and clause five point three point two explains why. If every manual used its own local vocabulary for phases, nobody could search a library of manuals or reuse one written elsewhere. So the standard adopts a common framework from ISO 22263, which it describes as one of the simplest project life cycle classifications. It has six phases. Inception establishes the need. Brief tests feasibility and gains financial approval. Design identifies and then fixes the major elements. Production finalises the deliverables, builds, and hands over as planned. Maintenance operates and maintains. And Demolition decommissions and disposes. ISO 22263 dates from 2008 and is current, confirmed at review in 2022, which is a reminder that age is not a measure of currency.

Clause five point three point three lets you work in local phases, such as the Plan of Work of the Royal Institute of British Architects, the RIBA Plan of Work, in the UK, on one condition. A direct mapping shall be maintained between the standard phases and the local ones, which means written down and kept current. Any mapping takes one of three shapes. One standard phase to one local phase. One standard phase decomposed into several local ones. Or several standard phases onto one broader local stage. What is excluded is partial overlap, where a local stage starts part way through one standard phase and ends part way through another, because a requirement tagged to a standard phase would then land in two local stages and belong properly to neither.

Mapped to the RIBA Plan of Work, two rows need judgement rather than a lookup. Production decomposes into stage five, Manufacturing and Construction, and stage six, Handover, because the phase includes handing over the product as planned. Someone who mapped Handover to Maintenance instead would not be obviously wrong, which is exactly why the mapping has to be recorded. And Demolition has no stage at all, because the Plan of Work ends at Use. An unmapped phase is a legitimate result, provided you say so. Forcing it into the nearest stage is worse, because a forced mapping looks complete and misleads the next reader. For the college, whose requirement belongs to the Design phase, the recorded mapping travels with the manual rather than living in one person's head.

Now the neighbours. The useful way to hold the surrounding standards is by the question each answers. Why information is needed, by whom and under what process, which is the ISO 19650 series. Which information units are needed, and at what point, which is this standard. How much each unit carries, which is ISO 7817-1. And how the information is structured, packaged and delivered, which is a family of its own, from Industry Foundation Classes in ISO 16739-1 to the container series ISO 21597, with Parts 2 and 3 of ISO 29481 holding machine-readable forms of the manual itself.

One relationship gets more care in the introduction of ISO 29481-1 than any other. An exchange requirement in a manual is generic, written for actor types, stating what a client type needs from a designer type for a given business process. An exchange information requirement under ISO 19650 is specific to one appointment on one project. So one exchange information requirement can draw on several exchange requirements. The two families share their vocabulary on purpose, and the price is that exchange requirement and exchange information requirement sound alike and mean different things.

PAS 1958, a publicly available specification that the British Standards Institution published in 2026 as a guide to how the data and information standards of the built environment relate, still uses the term transaction map, which the current edition no longer defines. Where guidance and the standard differ, the standard governs, and you flag the difference rather than quietly resolving it.

Run the college's questions through that map and a dispute starts to dissolve. Whether the design consultant must deliver room areas at all is a why question, settled through the appointment and its exchange information requirements. Which items must come with each area is a which question, answered by the manual. How much geometric detail each room carries is a how much question, for the level of information need framework. And how the delivered data is checked and authorised is a how question, under ISO 19650-4, which the episode Information Exchange in ISO 19650 Part 4 covers in depth. When two standards seem to apply equally, the question is nearly always two questions, and separating them is the work. That resolves the fifth step. A manual tagged to standard phases with a recorded mapping can travel between regions, and a question named by its kind finds its standard.

The sixth step goes from natural language to something a machine can read. Clause six describes the principles and scope of technical implementations without specifying them. Its overriding intention is that the specification is converted into a machine-readable form that carries the common metadata, which means the titles, identifier, authors and change log from the second step travel with it.

The central term is the exchange view definition, a computer-interpretable representation of one or more data models, or parts of them, that exists to fulfil an exchange requirement. It replaced model view definition, the industry's name for a view of a building model in Industry Foundation Classes, because manuals now span the whole life of an asset and many business contexts, and an exchange may not involve a building model at all. The older term is still live in software documentation and certification schemes, so expect to hear both. An exchange view definition delimits the part of a data model that an exchange needs, and may narrow the values its attributes can take. It does not change the data model's types, and it does not define the content, which belongs to the delivery itself. Clause six point three says that the requirement and the definition, presented together, may be used to validate a delivery. The requirement says what must be included. The definition says what the exchange must be able to carry.

The schema most definitions are written against is Industry Foundation Classes, published as ISO 16739-1. Its 2024 edition carries schema version four point three point two point zero, replacing the 2018 edition at four point zero point two point one, and buildingSMART International has an extension of version four point three and a refactoring of the four series in development. What the schema contributes is semantic precision. A wall arrives as a wall with properties and relationships, not as lines a person has to interpret. And the version a standard publishes is not the version your software supports, because implementation and certification take time. So specify the information rather than the version, keep any schema reference in the optional reference link, and where a project must fix a version, record it in the appointment with a review point, not in a manual meant to outlast it. A requirement that rooms be delivered as spaces in the current schema version ages the moment the version changes. A requirement that each room carry an identifier, a name, a function and a net internal area, with the measurement rule as a constraint, lasts.

Now the missing sentence from the first step. The buildingSMART Information Delivery Specification is a computer-interpretable specification for defining information requirements, used mainly with Industry Foundation Classes. It is published by buildingSMART International, a member organisation rather than a standards body. Version one point zero was approved on the first of June 2024 as that organisation's official standard, and earlier versions were development releases. It is not at present an ISO or a European standard. People ask whether it is the same thing as an information delivery manual. It is not, and the sentence that once let a manual share its name belongs to a withdrawn edition. The current edition moved the relationship into its introduction. There it observes that delivery is often assumed to mean an information model in Industry Foundation Classes, conforming to an information delivery specification that satisfies the exchange requirements. Then it significantly broadens that assumption, to any information format that can satisfy the requirements, in whole or in part. So the manual is the requirement, readable by every party and therefore something they can agree to. The buildingSMART specification is one way of testing whether that requirement was met. A rule set with no manual behind it cannot be reviewed, because when someone asks to relax a rule, nobody can say what it was protecting.

Clause four point seven names three routes for delivery, and the third surprises people. The first is attaching information to a digital message or, more commonly, linking to containers held in a shared repository. The second is bundling related containers into one file, under the ISO 21597 series, so that the relationships between documents stay explicit. The third is defining queries against shared databases. A query returns current information instead of a copy that goes stale once it is sent, and it has to satisfy the same exchange requirement as any other route. The requirement is the constant. The route is the variable. That is why open information exchange is not the same as open file transfer. A delivery in a perfectly open format that answers the wrong question satisfies nothing.

For the college, the units map onto the spatial concepts of the open schema, so an exchange view definition, or a set of checks written in the buildingSMART specification, could filter a delivered model down to exactly those units and test it. None of that changes the requirement. That resolves the sixth step. The specification comes first, and the implementation follows it.

The college's room data gives the episode its answer. The question was how to specify an information exchange precisely enough that every party can agree to it and software can check it. The answer is a manual in three components, written in natural language. A use case states the purpose, under a header that lets it be found, versioned and reused. A business context names the roles, with an initiator and an executor for every transaction, and shows where the requirement attaches. And an exchange requirement prescribes the information as nested units, each holding one concept, with descriptions and constraints that cannot be read two ways, behind stated preconditions and tagged to a mapped phase. Software checks it only after people have agreed what it says. At the college, the cost consultant would initiate the exchange, the area rule would be agreed before anyone measured a room, each room's function would be held once, and the re-issue would never be needed.

Two habits matter most, and a third protects them. Before writing any requirement, name who needs the information and for what purpose, because the role that needs the result is the one that must say what it needs. When you specify, prescribe units rather than formats, so that a stranger working only from your requirement would produce the values you expected. And the habit that protects both: when anyone presents a checking file or a schema version as the requirement, ask for the manual behind it.

For reading, start with ISO 29481-1 of 2025, especially clauses four and five, Annexes A, C and D, and its introduction on the relationship with ISO 19650. Read the buildingSMART documentation for the Information Delivery Specification from its published source, and keep ISO 22263 and the RIBA Plan of Work 2020 beside you for the phase mapping. The episode Information Management in ISO 19650 Part 1: Concepts and Principles sets out the exchange information requirement that a manual feeds. The next episode, Management Systems for Information, is about giving information management a place in the quality, risk and safety systems an organisation is already certified to.

Pick the exchange that fails most often in your own practice or research, find the role that needs its result, and have it write down what it needs before another file is sent. That is my starting point, as Dr Sepehr Abrishami, and the reason Reimagining AEC gave the handover an episode of its own.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Roles and the Responsibility Matrix

    Nov 10, 202633 minAudio

    On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks which organisation has failed when an information management task on a project goes undone, and how anyone could prove it in writing. A hypothetical council...

    Credit the people on this episode

  2. The Delivery Plans

    Nov 10, 202635 minAudio

    On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks how many task team plans become one master plan whose dates hold, and what a lead appointed party must do when they cannot. A hypothetical community hub,...

    Credit the people on this episode

  3. The BIM Execution Plan

    Nov 10, 202636 minAudio

    On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks how one BIM execution plan can serve both the tender and the appointment. A hypothetical community hub for a district council, whose tender cites a withdrawn...

    Credit the people on this episode

  4. The UK BIM Framework and the Move from PAS 1192

    Nov 10, 202634 minAudio

    On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks what the UK BIM Framework is and is not, and how a legacy document set maps onto the standards in force today. A hypothetical council's information pack,...

    Credit the people on this episode

  5. Management Systems for Information

    Nov 10, 202648 minAudio

    On Reimagining AEC, his own podcast, Dr Sepehr Abrishami asks where information management should sit inside an organisation's quality, risk and safety systems, and what changes when it sits there. A hypothetical design...

    Credit the people on this episode