Architecture is a set of decisions
Enterprise architecture decides what an organisation changes, in what order and to which standards, so separate projects add up to one coherent estate instead of a pile of exceptions.
TOGAF (The Open Group Architecture Framework) is a method for planning change across a whole organisation. This page explains how it works: the ADM (Architecture Development Method), building blocks, reuse, principles, gap analysis, governance and the ArchiMate language. It also shows how each part maps onto the designs, roadmaps and reviews of real infrastructure programmes.
TOGAF (The Open Group Architecture Framework) is an open standard for EA (enterprise architecture): how an organisation plans change across its business, data, applications and technology, and keeps projects pointing the same way while they deliver. It splits the enterprise into four architecture domains. A real change touches all of them.
Swipe sideways to see the whole diagram.
One change touches every domain. A goal such as getting plant data into the cloud starts as a business driver. It needs data owners and flows, new or changed applications and interfaces, and networks and hosting to carry them. Limits found lower down, such as a plant network that cannot be changed during production, flow back up and reshape the plan.
| Part | What it holds |
|---|---|
| Publisher | The Open Group, a vendor neutral consortium. TOGAF 1.0 appeared in 1995. The TOGAF Standard, 10th Edition, was launched on 25 April 2022. |
| Fundamental Content | Six documents: Introduction and Core Concepts; Architecture Development Method; ADM Techniques; Applying the ADM; Architecture Content; Enterprise Architecture Capability and Governance. |
| Series Guides | Guidance by topic that changes faster than the core, such as business architecture, security architecture, agile architecture, the digital enterprise and microservices. |
| Modelling language | ArchiMate, also from The Open Group. TOGAF says what to produce; ArchiMate gives a common notation for drawing it. See ArchiMate below. |
| Certification | Two levels for people: TOGAF Enterprise Architecture Foundation, then TOGAF Enterprise Architecture Practitioner. |
Enterprise architecture decides what an organisation changes, in what order and to which standards, so separate projects add up to one coherent estate instead of a pile of exceptions.
TOGAF describes a way of working and the content it produces. It expects to be tailored. Applied whole and unchanged, it produces paperwork faster than decisions.
As a technical design authority I already set principles, write the vision and high level design, plan phased transitions and review designs. TOGAF gives those steps shared names and a shared order.
The ADM (Architecture Development Method) is the core of TOGAF. A Preliminary phase sets up the practice. Then phases A to H take one piece of change from vision through design, planning and delivery, and back again. Requirements Management sits in the middle, because every phase adds requirements and depends on them. Step through the cycle, then open any phase in detail.
Swipe sideways to see the whole diagram.
The Preliminary phase prepares the organisation to do architecture at all. It agrees which parts of the enterprise are in scope, the principles that will guide designs, how TOGAF is tailored, who governs, and which tools and repository hold the work. It is revisited when the practice changes.
Phase A starts when a sponsor sends a Request for Architecture Work. I confirm goals, drivers and constraints, identify stakeholders and their concerns, assess readiness for change and sketch the baseline and target at a high level. The sponsor signs a Statement of Architecture Work before detailed design starts.
Phase B describes the business baseline and target: capabilities, value streams, organisation and processes. Gap analysis between the two shows what must change. For infrastructure work I keep it short but never skip it, because it tells me which capabilities the network must protect first.
Phase C covers data and applications, in either order. Data architecture describes the main data entities, their owners and how data moves. Application architecture describes the applications and their interfaces. For plant work this is where historians, MES (manufacturing execution systems) and cloud platforms appear.
Phase D describes the technology baseline and target: platforms, networks, hosting, cloud and the standards that govern them. Most of my design work sits here: WAN (wide area network) and campus, data centre fabrics, security zones and cloud connectivity, each traced back to the application and business needs above.
Phase E is the first view of delivery. I consolidate the gaps from B, C and D, group them into work packages, decide whether to build, buy or reuse, and define transition architectures: stable states the estate can safely stop at. The result is a first roadmap.
Phase F turns the draft roadmap into a plan the organisation can fund and staff. I agree costs, benefits, risks and dependencies with programme and portfolio owners, sequence the work packages and check that each transition delivers business value. The architecture documents are finalised here.
Phase G watches delivery. Each project signs an Architecture Contract that states what it will build and to which standards. I review designs and builds for compliance, approve dispensations with an end date, and raise change requests when delivery shows the architecture was wrong.
Phase H keeps the architecture current after delivery. I watch new business needs, technology and lifecycle dates, and assess each change request. Then I decide whether it is a simple update, a partial rework, or big enough to start a new cycle with a fresh Request for Architecture Work.
Requirements Management is not a step in the sequence. It runs alongside every phase. New and changed requirements go into one repository, and each change gets a Requirements Impact Assessment, so a late requirement in Phase G is checked against the vision, the design and the plan.
Phase H can start a new cycle, and any phase can send work back to an earlier one. Most real work loops through B, C and D several times before the target settles.
A requirement found during build is still handled properly: it goes to the repository and gets an impact assessment, rather than a quiet change to one design.
Each phase has defined outputs, such as the Statement of Architecture Work in Phase A or the Architecture Contract in Phase G, so everyone knows what has been agreed and where it is written down.
TOGAF expects the ADM to be repeated and tailored, never run once from start to finish. It groups the phases into four iteration cycles, describes three levels of architecture, and scopes each piece of work by breadth, depth, time and domain. Choose a cycle to see which phases it repeats.
Swipe sideways to see the whole diagram.
Architecture Capability. The Preliminary phase and Phase A, repeated when the practice itself needs to change: a new board, new principles, or a big programme that needs its own scope and sponsor. I run this loop before anything else on a new engagement.
Architecture Development. Phases B, C and D, cycled until the domains agree with each other. A technology choice in D often changes an application decision in C, so I expect two or three passes rather than one.
Transition Planning. Phases E and F, repeated as costs and dependencies become clear. Each pass refines the work packages and the transition architectures until the plan can be funded.
Architecture Governance. Phases G and H run for as long as delivery runs. Compliance reviews and change requests loop between them, and a big enough change triggers a new cycle from Phase A.
| Level | Scope | Typical horizon | Who uses it | Where my work sits |
|---|---|---|---|---|
| Strategic architecture | The whole enterprise | Several years | Executives and the Architecture Board | The direction my designs must fit, such as cloud first or zero trust |
| Segment architecture | One business area, portfolio or programme | One to three years | Programme and portfolio leads | An OT (operational technology) maturity roadmap across manufacturing sites, or a WAN refresh across a region |
| Capability architecture | One capability or project | Months to a year | Project teams and design authority | An AI (artificial intelligence) ready data hall, or the network and security for one migration |
Horizons are indicative. TOGAF describes the three levels; the horizons depend on the organisation.
Which parts of the enterprise are in scope: one site, one business unit, or the whole group, including partners and suppliers.
How much detail each domain needs. A strategy needs a page per domain; a capability architecture needs enough for a low level design to start.
How far ahead the target sits, and how many transition architectures lie between today and that target.
Which of business, data, application and technology must be worked on. Infrastructure work is mostly technology, but never only technology.
The content framework defines what the ADM produces. Deliverables are the formal outputs that stakeholders sign off. Artefacts are the catalogues, matrices and diagrams inside them. Building blocks are the reusable pieces the architecture is made from. Step through one building block, from a business need to a built solution.
Swipe sideways to see the whole diagram.
A requirement arrives from the business: every manufacturing site needs secure, resilient links to the data centre and the cloud. It says what is needed, not which product to buy.
I describe the need as an ABB (architecture building block): a set of functions and qualities with no product attached. Two diverse paths, encrypted transport, central policy and direct cloud access that is still inspected.
SBBs (solution building blocks) are real products or services that could deliver the ABB. Here there are three options: two SD-WAN (software defined WAN) products and a managed MPLS (Multiprotocol Label Switching) service.
Each option is scored against the ABB, the principles and the standards: fit, cost, lifecycle, skills and supplier risk. The managed MPLS service fails the cloud access requirement. Both SD-WAN products fit; the first matches the skills the operations team already has.
The chosen SBB is specified in the low level design and built. Because the ABB came first, the decision record shows why this product won and what any replacement must still do.
Both building blocks go into the architecture repository. The next site, or the next acquisition, starts from the approved pattern instead of a fresh evaluation.
A deliverable is formally reviewed, agreed and signed off by its stakeholders. It is a contract, so it changes through change control.
Catalogues, matrices and diagrams are the parts people read. I build deliverables mostly from artefacts, which keeps the prose short.
A good building block has a clear boundary and interfaces, so it can be reused or replaced. ABBs say what is needed; SBBs say what delivers it.
| Artefact type | What it is | Examples from network and infrastructure work |
|---|---|---|
| Catalogue | A list of things of one type | Application portfolio, technology standards, sites and circuits, principles |
| Matrix | The relationships between two types | Applications to data, applications to technology, stakeholders to concerns, flows to firewall zones |
| Diagram | A picture for one audience and purpose | Context, network zones, data flows, resilience, the migration roadmap |
| Deliverable | Phase | What it fixes |
|---|---|---|
| Request for Architecture Work | Before A | The sponsor's request: the problem, the goals and the constraints |
| Statement of Architecture Work | A | The agreed scope, approach, plan and sign off for the architecture work |
| Architecture Vision | A | A short view of the baseline and the target, and the value of getting there |
| Architecture Definition Document | B to D | The architecture itself: baseline, target and gaps in every domain |
| Architecture Requirements Specification | B to D | The requirements a solution must meet to conform to the architecture |
| Architecture Roadmap | E, F | The work packages and transition architectures, in order |
| Implementation and Migration Plan | E, F | How and when the roadmap is delivered, with costs and resources |
| Architecture Contract | G | The agreement between the architecture function and a delivery team |
| Compliance Assessment | G | The result of reviewing a project against the architecture |
| Change Request | H | A proposed change to the architecture, with its impact assessed |
| Requirements Impact Assessment | Any | What a new or changed requirement does to the work so far |
The Enterprise Continuum is a way of classifying architectures and solutions from generic to specific, so teams reuse before they build. The architecture repository is where those assets, and the decisions about them, are kept. Both only work if someone owns them.
Swipe sideways to see the whole diagram.
Reuse before you build. I look left before designing anything: is there a common pattern, an industry model or an agreed standard that already answers the question? Each organisation specific design is a specialisation of those. When a project produces something better, it moves left as a new reusable pattern.
The types of thing the practice records, such as capabilities, applications, data, platforms and requirements, and how they relate.
How the practice itself runs: roles, the board, processes, templates and tools.
Baseline, target and transition architectures, at strategic, segment and capability level.
The products, versions and patterns the enterprise has agreed to use, each with a lifecycle status, for example trial, active, deprecated or obsolete.
Reusable patterns, templates and reference models, including industry models such as TM Forum's.
Decisions, compliance reviews, dispensations and their expiry dates.
Every architecture requirement, with its source, owner and the designs that meet it.
What is actually deployed, ideally fed from the CMDB (configuration management database) and the network source of truth rather than typed in.
| Tool | What it is good for |
|---|---|
| Sparx Systems Enterprise Architect | Detailed models in ArchiMate and UML (Unified Modeling Language), with a large repository |
| Bizzdesign, SAP LeanIX, Ardoq | Hosted EA platforms with portfolio views, surveys and roadmaps that business owners can read |
| Archi | A free, open source ArchiMate modelling tool. Good for small teams and for learning |
| ServiceNow CMDB and application portfolio management | The deployed estate and its owners, which keeps the solutions part of the repository honest |
| A network source of truth, such as NetBox | Sites, devices, circuits and addressing, linked to the designs that created them |
Tool names are examples of what I meet in clients. The repository is a practice, not a product.
Principles are the standing rules that guide every design. TOGAF gives each one four parts: a name, a statement, a rationale and its implications. A good principle is short, can be tested in a review, and has consequences someone will feel. They are agreed in the Preliminary phase and confirmed in Phase A.
Example principle
| Weak principle | Why it fails | Stronger version |
|---|---|---|
| "We use best of breed." | Cannot be tested, and every project can claim it. | "We choose from the approved standards list; anything else needs a dispensation." |
| "Security is everyone's job." | True, but implies nothing a reviewer can check. | "Every flow between zones is listed, owned and inspected." |
| "Use the cloud." | A preference with no reason, so it breaks at the first exception. | "Cloud first for new applications, unless latency, data residency or cost rules it out, with the reason recorded." |
| "Keep it simple." | No one disagrees, so it never settles an argument. | "No more than two routing protocols and one SD-WAN platform in the estate." |
Ten to twenty principles is plenty, and they should outlast any one programme. If they change every quarter, they are really policies or standards.
A principle that never stops anyone doing anything is decoration. The implications line is where the cost shows.
At design review I quote principles, not opinions. A designer who disagrees argues with the principle, in the open, through the board.
Gap analysis compares the baseline with the target, building block by building block. Each gap becomes work. Transition architectures are the stable states in between, and each one should be worth stopping at. Step through an illustrative network roadmap, then try the matrix.
Swipe sideways to see the whole diagram.
The baseline is the estate as it runs today, described only as deeply as the decision needs: MPLS to every site, firewalls only in the data centre, flat plant networks and changes made by hand.
The target comes from Phases B to D: two carriers per site under SD-WAN, a cloud security service for users, zoned plant networks and changes delivered through a pipeline. I agree the target before planning the route.
Gap analysis compares the two, building block by building block. Every difference becomes a gap: something to build, change or retire. The tool below runs the same comparison.
A transition architecture is a stable state the estate can stop at safely if money or attention runs out. The first one proves the pattern: SD-WAN at pilot sites, one plant zoned and a source of truth started.
The second transition scales what worked: every site on SD-WAN, MPLS contracts ended, half the plants zoned and every site built from a template. Each transition delivers value on its own.
The last step reaches the target. Phase H then watches for the next change, because by the time a target is reached the business has usually moved on.
TOGAF's gap analysis puts baseline building blocks in rows and target building blocks in columns. A baseline block can be kept, enhanced, replaced or eliminated. Any target block that nothing in the baseline turns into is new. Change what happens to each block and watch the gaps, and the warnings, change.
| Baseline | SD-WAN | Cloud security | Zoned plants | Automation | Telemetry | Kept or enhanced | Eliminated |
|---|---|---|---|---|---|---|---|
| MPLS WAN | Replaced | ||||||
| Data centre firewalls | Replaced | ||||||
| Flat plant networks | Replaced | ||||||
| Manual configuration | Replaced | ||||||
| Branch routers | Eliminated | ||||||
| SNMP monitoring | Enhanced | ||||||
| New | New |
Each replaced, enhanced, eliminated or new block is a gap, and each gap becomes part of a work package in Phase E.
7 gaps to plan. 4 replaced, 1 enhanced, 1 eliminated and 1 new. Check every eliminated block: something may still depend on it.
TOGAF tells the architect to check every eliminated block: was it removed on purpose, or was a function simply forgotten? Retired branch routers might still be the only place a site's voice gateway or its DHCP (Dynamic Host Configuration Protocol) relay runs.
Funding stops, priorities change and suppliers slip. Each transition architecture must be safe, supportable and worth having if the programme pauses there.
Every gap has people, process and supplier sides: new skills, new support contracts, changed runbooks. I list them with the technical work so they are costed too.
Every stakeholder has concerns. TOGAF, and the international standard for architecture description, ISO/IEC/IEEE 42010, handle them with viewpoints and views: a viewpoint says how to show something to one audience, and a view applies it to this architecture. Choose an audience to see the same design drawn for them.
Swipe sideways to see the whole diagram.
Sponsor view. The CFO (chief financial officer) and the business sponsor care about cost, value, risk and the order of work. They get one page: the outcome, the transitions, what each one costs and what could stop production.
Security view. The CISO (chief information security officer) cares about zones, trust and every flow between them. The same design is drawn as zones, with each permitted flow listed and inspected, and plant traffic leaving only through the industrial DMZ.
Operations view. Operations care about who owns what, what is monitored, which alerts fire and what the runbook says. The same design gains a monitoring path from every site and an owner for every part.
Engineering view. Engineers need interfaces, protocols and numbers: routing, VLANs (virtual local area networks), address ranges and private links. This view leads straight into the low level design.
| Stakeholder | Main concerns | What I give them |
|---|---|---|
| Sponsor and finance | Cost, value, risk, timing | A one page vision and a costed roadmap |
| Chief information security officer | Zones, identity, data protection, compliance | Zone diagrams, a flow matrix and the risk register |
| Plant and site managers | Production safety, downtime, change windows | Site impact, outage plans and the change calendar |
| Operations and the service desk | Ownership, monitoring, support, runbooks | Support model, monitoring design and runbooks |
| Engineers and suppliers | Interfaces, standards, configuration | High and low level designs, standards and templates |
| Architecture Board | Fit with principles and standards, reuse | The design, the decisions and any dispensations requested |
TOGAF's stakeholder management technique sorts people by their power over the work and their interest in it. I redraw the map at the start of each phase, because both change.
Phase G and the governance part of the standard describe how architecture stays real during delivery: an Architecture Board, Architecture Contracts, compliance reviews and dispensations. TOGAF also defines how closely an implementation can follow its specification, from irrelevant to fully conformant.
Swipe sideways to see the whole diagram.
The project submits its design with a compliance checklist: which principles, standards and requirements it meets, and where it departs from them. I ask for this before the low level design is finished, while change is still cheap.
The Architecture Board, or a design authority acting for it, reviews the design. Questions go back and forth. The board checks fit with principles, standards, the target architecture and other projects.
A compliant design proceeds. The project and the architecture function sign an Architecture Contract: what will be built, to which standards, and how compliance will be checked again before go live.
A justified exception gets a dispensation: approved, but with conditions, a named owner and an end date. A supplier lead time or a legacy dependency is a fair reason; preference is not.
A non conformant design goes back with written findings. Most rework is small when the review is early. Late reviews are where governance gets its bad name.
Every decision, condition and dispensation goes into the governance log. Phase H reviews dispensations as they expire, so exceptions do not quietly become the standard.
| Level | What it means |
|---|---|
| Irrelevant | The implementation has nothing in common with the architecture specification, so conformance does not arise. |
| Consistent | Some features in common, built as specified. But some specified features are missing, and there are extra features the specification does not cover. |
| Compliant | Some specified features are missing, but everything built is covered by the specification and built as specified. |
| Conformant | Every specified feature is built as specified, but some extra features are built that do not follow the specification. |
| Fully conformant | A full match: every specified feature is built as specified, and nothing extra. |
| Non conformant | Any of the above where some specified feature is built in a way that does not follow the specification. |
The six compliance levels described in TOGAF's architecture compliance guidance, in my words.
Senior architects for each domain, security, operations and a business representative, with a chair who can say no. A board that cannot refuse is a mailing list.
I review at high level design, at low level design and before go live. An early review costs a conversation; a late one costs a rebuild.
Every exception has an owner, a reason, a compensating control and a date. Open dispensations past their date are reported to the board each month.
My own design authority method, with decision records, factory builds and supplier reviews, is on the design authority page. This section is the TOGAF view of the same job.
Architecture governance meets project governance at the stage gates on my PRINCE2 page.
TOGAF says what to produce. ArchiMate, also from The Open Group, is a modelling language for drawing it consistently, with defined elements and relationships across motivation, strategy, business, applications, technology and change. Choose a domain to see one small model built up.
Swipe sideways to see the whole diagram.
Motivation. Drivers, goals, outcomes, principles and requirements explain why the architecture exists. Here a driver, plant data that nobody can see, influences a goal of live plant data, and a requirement that OT stays isolated helps realise it.
Strategy and business. A capability, production insight, is realised by a process: reviewing line output. A business actor, the plant manager, is assigned to that process. Strategy elements such as capabilities and value streams sit above the business domain.
Application. An application component, the data platform, serves the business process and accesses a data object, line telemetry. In ArchiMate 4, behaviour such as process, function, service and event comes from one common set of elements, used in every domain.
Technology. A technology node, the edge gateway, serves the data platform, and a communication network, the plant network, connects it to the line. This is where my network designs meet the rest of the model.
Implementation and migration. A work package, the plant pilot, realises the change to the data platform and reaches a plateau: a stable state, which matches a TOGAF transition architecture.
| Domain | Main elements in ArchiMate 4 |
|---|---|
| Common | Role, collaboration and path; service, process, function and event; grouping and location |
| Motivation | Stakeholder, driver, assessment, goal, outcome, principle, requirement, meaning and value |
| Strategy | Resource, capability, value stream and course of action |
| Business | Business actor, business interface, business object and product |
| Application | Application component, application interface and data object |
| Technology | Node, device, system software, technology interface, communication network, distribution network, equipment, facility, artifact and material |
| Implementation and migration | Work package, deliverable and plateau |
From the ArchiMate 4 Specification pocket guide, published by The Open Group in April 2026. Element names are ArchiMate's own.
| Group | Relationships | Example |
|---|---|---|
| Structural | Composition, aggregation, assignment, realisation | A work package realises a plateau |
| Dependency | Serving, access, influence, association | A data platform serves a business process |
| Dynamic | Triggering, flow | A line alarm triggers a maintenance process |
| Other | Specialisation, junctions and multiplicity | An edge gateway is a specialisation of a node |
Because every element has a defined meaning, the sponsor's one page view and the engineer's detail can come from the same model and stay consistent.
ArchiMate 3.2 was published in October 2022. ArchiMate 4 followed in April 2026, keeping the core ideas while reorganising the language into domains with a common set of elements. Tools and models move across at different speeds, so I agree the version up front.
Archi is free and open source. Sparx Systems Enterprise Architect, Bizzdesign and similar platforms add repositories, collaboration and reporting.
Most infrastructure work happens at programme and project level, where the documents are high and low level designs, migration plans and design approvals. TOGAF's names map neatly onto them. The method also has to live alongside agile teams and cloud platforms with frameworks of their own.
Swipe sideways to see the whole diagram.
Just enough, just in time. TOGAF has a Series Guide on agile architecture. In practice the vision, target and principles change slowly, while the runway of approved patterns grows a little ahead of the teams. Decisions are short written records, and what sprints learn flows back into them.
| TOGAF name | What I usually call it | Where it lives |
|---|---|---|
| Request for Architecture Work | A demand or project brief | The portfolio or demand tool |
| Statement of Architecture Work | The scope of work for the design | Signed off by the sponsor |
| Architecture Vision | A solution outline: the one page vision | The design repository |
| Architecture Definition Document | The HLD (high level design) | The design repository, reviewed by the board |
| Architecture Requirements Specification | The requirements catalogue, with NFRs (non functional requirements) | An appendix to the HLD, or a requirements tool |
| Below TOGAF's level of detail | The LLD (low level design) and build sheets | The engineering repository, generated from the source of truth where possible |
| Architecture Roadmap and Implementation and Migration Plan | The migration plan, in waves | The programme plan |
| Architecture Contract | Design approval with conditions | The decision log |
| Change Request | A design change request | The change process |
Most of my documents are high and low level designs. The HLD carries the architecture: context, principles, target, gaps and decisions. The LLD turns one solution building block into something an engineer can build.
Cloud providers publish adoption frameworks, such as the Microsoft Cloud Adoption Framework for Azure and the Cloud Adoption Framework from AWS (Amazon Web Services). They cover one platform in depth; TOGAF sits above them and decides where each platform fits.
Agile teams still need a target, principles and guardrails. I keep those short and stable, and let detailed design happen in the teams, close to the work.
TOGAF expects tailoring. Choose the size of the change, the delivery style and how regulated the organisation is, and see how I would scale the method.
A programme, agile, heavy regulation.
Enterprise architecture fails in familiar ways, and almost never because the framework was wrong. Choose a symptom to see where in the cycle it usually starts, what I see, and what I change. My general method for finding faults, technical or not, is on the troubleshooting page.
Swipe sideways to see the whole diagram.
Documents nobody reads. The architecture was written to satisfy the framework, not the people who must act on it. I cut it to a one page vision, a view for each audience and decision records, and I test each document by asking a stakeholder what it told them.
Projects route around the board. Governance is slow or late, so teams avoid it. I move reviews to the high level design, commit to a turnaround, publish approved patterns that need no review, and make the board the fastest route to a yes.
Endless current state work. The baseline became the project. I time box it, describe only what a decision depends on, and agree the target early so the baseline work has a purpose.
A target with no route. The architecture stopped at Phase D. Without Phases E and F nobody can fund it. I group the gaps into work packages and transition architectures, cost them with the programme, and make each transition useful on its own.
Exceptions that never expire. Dispensations were granted without an end date or an owner. Each one now has both, with a compensating control, and the board sees every expired dispensation each month until it is closed or renewed on purpose.
A repository nobody trusts. It was filled once by hand and never maintained. I give each area an owner, feed deployed facts from the CMDB and the network source of truth, and remove what nobody uses.
Framework for its own sake. TOGAF was adopted whole, so the team produces artefacts instead of decisions. I tailor it in the Preliminary phase: start from the decisions the organisation must make, and produce only what those decisions need.
Nobody agrees what success is. Phase A was skipped. I go back to it: confirm the goals, the measures, what is out of scope and who decides, and get the Statement of Architecture Work signed before more design.
| Signal | What it tells me | Healthy, indicative |
|---|---|---|
| Review turnaround | Whether governance helps or blocks | Most reviews answered within ten working days |
| Reviews before build | Whether projects use governance at all | Nearly every in scope project reviewed at high level design |
| Expired dispensations | Whether exceptions are controlled | None open past their date without a decision |
| Pattern reuse | Whether the repository is useful | Most new designs start from an approved pattern |
| Requirements traced | Whether designs answer real needs | Every target building block traces to a requirement or driver |
| Stakeholder feedback | Whether documents help anyone | Sponsors can state the target and the next transition |
Targets are indicative. I agree real ones with the board in the Preliminary phase.
Architecture work starts with a request from a sponsor. If it goes well, it ends with a roadmap that projects follow and a practice that keeps it current. I follow the same path whether the scope is one capability or a whole segment: agree the outcome, find out what really exists, design the target, plan the transitions and govern delivery.
Swipe sideways to see the whole diagram.
I start with the sponsor's request and turn it into a signed Statement of Architecture Work: the problem, the outcome, what is in and out of scope, and the measures of success. I map the stakeholders and record who decides in a RACI (responsible, accountable, consulted, informed) chart.
I turn drivers and goals into requirements: what the target must do, and how well, with a number wherever one can be agreed. Principles come from the Preliminary phase or are agreed now. Each requirement has a source and an owner, and goes into one repository with an impact process for changes.
I describe the baseline only as deeply as the decisions need. I pull inventories from the CMDB, the application portfolio, cloud accounts and the network source of truth, then check them with the owners in workshops. Capability and maturity assessments show where the real gaps are.
I design the target across business, data, applications and technology, with security in each, and compare it with the baseline building block by building block. Options are scored against the principles and requirements. The result is an Architecture Definition Document and a high level design the board can approve.
I group the gaps into work packages and define transition architectures, each a stable state worth stopping at. With the programme and finance I cost, sequence and fund them. The roadmap and the migration plan become the programme's own plan, not a separate architecture document.
During delivery each project signs an Architecture Contract and is reviewed at high level design, low level design and before go live. Dispensations get an owner and an end date. Decisions go into the governance log, so later teams can see what was agreed and why.
After delivery I keep the architecture current. I assess change requests, track product end of life dates and new business needs, and update the standards and the repository. When a change is big enough, it becomes a new Request for Architecture Work and the cycle starts again.
What the architecture engagement must deliver.
How well it must do it, with a number wherever one can be agreed.
Numbers are typical starting points. I agree real ones with the sponsor and the board.
| Tool or source | What it tells me |
|---|---|
| SAP LeanIX or Ardoq | The application portfolio: owners, lifecycle and costs |
| ServiceNow CMDB and application portfolio management | Deployed services, dependencies and support groups |
| Sparx Systems Enterprise Architect, Bizzdesign or Archi | Models in ArchiMate: capabilities, applications, technology and the links between them |
| NetBox or another network source of truth | Sites, devices, circuits and addressing |
| Azure Resource Graph and AWS Config | What actually runs in each cloud account |
| Capability and maturity questionnaires | Where the business sees its biggest gaps |
When an architecture engagement goes off track, the cause is usually missing agreement, not a missing artefact. I work through it in order.
| Layer | What I check |
|---|---|
| Business | Capabilities and outcomes agreed with their owners, with a measure for each. |
| Data | An owner for every data set that crosses a boundary, and its classification applied. |
| Application | An accurate portfolio, every interface listed, and lifecycle dates known. |
| Technology | Current standards, platforms in support, and failure domains understood. |
| Security | Zones, identity and flows drawn in every domain, not added at the end. |
| Governance | Decisions logged, contracts signed, and dispensations dated and owned. |
My general method for finding faults is on the troubleshooting page, and my design authority method is on the design authority page.
An architecture is only useful while people use it. I agree how the practice will run after the first cycle.
A standing board with a chair, terms of reference, a monthly cadence and a fast route for urgent decisions.
Weekly or fortnightly design reviews for active programmes, with published turnaround times.
A named owner for each area of the repository, with feeds from the CMDB and the source of truth.
An approved patterns library that teams can use without a review, updated through the board.
A short record of every significant decision: context, options, choice and consequences.
Monthly figures on reviews, turnaround, dispensations and reuse, shared with the sponsors.
Architectures, standards and the frameworks themselves change. I plan for each.
These are the projects closest to this page. They use the skills TOGAF describes: agreeing principles and standards, writing the vision and the high level design, planning change as phased transitions, and governing delivery as a design authority.
Led network and security transformation through AVEVA's reverse merger, modernising every site worldwide. Set the design authority standards adopted by every regional delivery team, and planned the rollout as a phased, site by site transformation roadmap.
For Dyson, assessed OT maturity across manufacturing sites, identified capability gaps and shaped a phased IT/OT transformation roadmap covering Purdue Model zoning, OT cybersecurity and a UNS (Unified Namespace) for plant data.
Produced high level designs for Virgin Media O2 across RAN (radio access network), core, transport, and OSS and BSS (operations and business support systems), using TM Forum frameworks. The work fed architecture into its autonomous network and service orchestration roadmap. More on the TM Forum page.
As Digital Transformation Architect and Technical Design Authority, led discovery, due diligence and architecture for network infrastructure automation aligned to TM Forum and MEF (formerly the Metro Ethernet Forum) standards, in support of Vodafone's digital customer platforms.
Delivered the EMEA (Europe, the Middle East and Africa) Network Refresh programme: audit, HLD and LLD design and implementation of LAN (local area network), WAN, SD-WAN and data centre infrastructure. Advised on change governance and architecture standards, and kept the estate compliant with MiFID II (the Markets in Financial Instruments Directive) throughout.
Lead architect on network, cloud and data centre modernisation programmes from discovery to operational handover across several client environments, including Weetabix, Dyson, Virgin Media O2 and Network Rail.
Related credentials: TM Forum Architecture, PRINCE2 Practitioner, and ITIL v3 and v4. See all credentials
Every page follows the same idea: how it works, the choices that matter, and where I have used it. The foundations pages explain the basics the others build on.