Keystone is a structured methodology for delivering ERP and digital transformation programmes. Six phases, twenty stages, two real board gates, role distinctions most methodologies blur, an eight-level testing model, RACI by binding decision, and a library of pre-drafted artefact templates — refined across thirty-plus programmes and openly published.
This Command Centre is the methodology turned into an interactive reference. Walk it tab by tab, drill into stages, expand topics, copy what's useful into your programme. Nothing here is gated.
Click any panel to expand. About 30 seconds each.
Keystone is built across four assets, all openly available:
The Story page explains why each piece exists.
Keystone is openly published. Anyone may use, adapt, or teach it. There is no licence, no signup, no gate, no paywall. Templates are downloadable in editable formats. The methodology is free.
Three reasons it's free:
Andy Peel, who built the methodology, sells time on programmes that benefit from someone who's seen this shape of problem before. The methodology is the proof of judgement; the time is the asset.
The Command Centre uses a small set of consistent interaction patterns. Once you've spotted them once, the whole site reads the same way:
If you're new to the site: open /method/index.html for the written overview, walk back here to drill into specifics, and download what's useful from /downloads.html as you go.
Keystone is a generic ERP delivery methodology. It has been refined across thirty-plus programmes — the deepest of those on Microsoft Dynamics 365 (F&O, SCM, CE) — and where the methodology gives concrete examples, vocabulary or vendor specifics, those typically lean D365.
The structure itself — six phases, twenty stages, the gate scheme, role distinctions, the eight-level testing model, RACI — adapts to any ERP platform: SAP S/4HANA, Oracle Cloud ERP, Workday, NetSuite or other. The phases sequence the same way, the gates do the same job, the role distinctions hold, the testing model applies.
Where the Command Centre includes platform-specific content — cost ranges, service tiers, functional summaries on the Vendor & Platform tab — it is included primarily for reference. Real cost figures, qualifying criteria and functional positioning shift continuously and should be confirmed with your SI or directly with the vendor before any commitment in a Full Business Case, an SOW, or a board paper.
Honest positioning matters more than vendor-neutral pretence. The methodology speaks D365 most fluently because the practitioner who built it has run more D365 programmes than any other platform. The structure translates; the examples lean.
Total timeline 18–24 months end-to-end. The Gates & Events strip at the top of the chart is the master annotation: large gold diamonds are the two real board gates (Funding Envelope & Benchmark Costs (S6) — proceed, Solution Design & Full Business Case (S12) — commit to build), the amber diamond is the Executive Sponsor's go-live go/no-go, small blue diamonds are phase checkpoints (forum reviews confirming exit/entry — not board gates), and the gold star marks Go-Live. Vertical guides drop from each marker through the phase rows so you can read what stage a gate sits on.
The default Method timeline above shows a single-wave delivery — one programme, one go-live. At PLC scale and across multi-entity, multi-region or multi-brand estates, delivery is more often multi-wave: a global template built once, deployed in successive waves to regions, brands, business units or factories. The Method bends to wave delivery in a defined way without breaking. Multi-wave is the norm at this scale, not the exception.
Run once, programme-wide
Repeat per wave (S13–17)
Wave can mean: regions (EMEA → AMER → APAC), brands within a group (Brand A → Brand B → Brand C), business units (UK trading division → service division → finance shared service), or functions (Finance & HR → Supply Chain → Sales/CRM). The shape of the diagram holds for all of them. Discovery (S11) surveys the whole estate — every BU, brand or function — even though Wave 1 only deploys to one slice. The Solution Design & Full Business Case (S12) locks Wave 1 firm at ±10–15% and Waves 2–3 indicative, with the blue re-baseline gate ◆ at the start of each subsequent wave to refresh ROM and reconfirm sponsorship before that wave's S13 starts.
For the canonical statement of how stages adapt to multi-wave delivery — and when not to use it — see _CANONICAL_REFERENCE.md → Delivery patterns — single-wave vs multi-wave. The Discovery (S11) and Solution Design & Full Business Case (S12) cards in the Stages tab now carry the same multi-wave note inline.
The business case matures as accuracy improves. There are exactly two real board gates: Funding Envelope & Benchmark Costs (S6) — proceed and Solution Design & Full Business Case (S12) — commit to build. Anything else called a "gate" is decoration.
Benefits Map, baselines, ROI Driver Matrix. Benefit Owners named. The benefits half of the eventual business case.
Benchmark costs, board-approved, programme initiation. First real board gate.
Rough Order of Magnitude pricing from selected SI. Confidence narrows but not yet committed.
Firm/capped SI build & test pricing. Final investment decision before build begins. Second real board gate.
Common mistake: calling Market Engagement & RFI (S7), SI Selection (S9) or Build & Configuration (S13) a "gate" in exec material. S7 and S9 are checkpoints — useful but not binary go/no-go. S13 is build execution; if you treat it as a gate, you've already committed in error.
The accountable body for the programme. Owns the funding decision at Funding Envelope & Benchmark Costs (S6) and the commit-to-build at Solution Design & Full Business Case (S12). Authorises handover to Deploy.
Operating governance. Tracks progress against plan, reviews risks, approves change requests, and owns scope/cost/time trade-off decisions within the approved envelope.
The technical conscience of the build. Approves design decisions, polices scope creep, signs off integration architecture, and owns P2-with-workaround sign-off at go-live (with the relevant Process Owner).
All three are active simultaneously from the end of SI Selection (S9) through Hypercare & Stabilisation (S17) — that's the highest-load period for programme governance. Plan accordingly: secure attendees' diaries early.
Day-to-day operating rhythm. Set up at Programme Setup & Mobilisation (S10), signed by Client and SI, referenced in the SOW. Setting cadence reactively at Build & Configuration (S13) when disputes appear is a common mistake.
A workstream is a deliberately-scoped slice of programme delivery with its own Project Manager. Programmes structure them by discipline (Functional, Data, Integration, Test, Cutover, Change), by module (D365 F&O, D365 CE, SAP FI), by business process (Order-to-Cash, Procure-to-Pay), by business function (Finance, Supply Chain, Procurement), or hybrid. The structure is decided at Programme Setup & Mobilisation (S10).
| Forum | Cadence | Stage span | Chair | Purpose |
|---|---|---|---|---|
| Programme Status | Weekly | S10–S17 | Programme Manager | Status, milestones, RAID, escalations across all workstreams |
| Project review | Weekly | S10–S17 | Project Manager (per workstream) | Formal workstream-level review — deliverables, milestones, RAID, dependencies. Feeds the Programme Status meeting above. |
| Workstream stand-ups | Daily or three times a week per stream | S10–S17 | Workstream Lead | Operational coordination — Functional, Technical, Data, Integration, Test, Cutover |
| Design Authority | Bi-weekly | S9–S17 | Solution Architect (Client) co-chairs with SI Solution Architect | Design decisions, change requests, deviation requests |
| Workstream coordination | Bi-weekly | S10–S17 | Programme Manager | Cross-workstream sequencing, dependency clearance |
| Steering Committee | Monthly | S10–S18 | Executive Sponsor or designate | Operating governance — progress, risk, formal change approvals over DA threshold |
| Executive Sponsor Group | As required + every phase gate | S0–S19 | Executive Sponsor | Programme commitment; gate decisions |
| Risk & Issue review | Weekly (often part of Programme Status) | S10–S17 | PMO Lead | Drives RAID register; escalates to Steering at threshold |
The Method assumes Hybrid delivery with EDUF (Enough Design Up Front). Pre-Programme & Selection are structured. Programme Setup through Solution Design (S10–S12) are EDUF — time-boxed, milestone-driven, iterative within. Build (S13) is Agile sprints. Testing (S14) reverts to Waterfall — sequenced SIT → UAT → NFT → Pre-UAT → Dress Rehearsal. Deploy (S15–S17) is Waterfall.
| Ceremony | Cadence | Attendees | Purpose |
|---|---|---|---|
| Sprint Planning | Start of each 2–3 week sprint | Functional Lead, Process Owners, SI dev team, Client Test Manager | Agree backlog for sprint, Definition of Done |
| Daily Stand-up | Daily, 15 min | Sprint team | Coordination, blockers |
| Sprint Review (show-and-tell) | End of sprint | Process Owners, Functional Lead, SI team | Demo working configuration; capture feedback |
| Sprint Retrospective | End of sprint, internal | Sprint team only | Continuous improvement |
| Backlog Refinement | Mid-sprint | Functional Lead + SI Functional Lead | Groom upcoming items |
| Release Planning | Across multiple sprints | Programme Manager, Functional Leads, SI PM | Sequence sprints into releases |
The trigger for raising a Change Request is not size in days. It is whether the change touches a contracted deliverable, the SDD, scope, cost, timeline or benefit. Threshold and process authored at Programme Setup & Mobilisation (S10) in the Change Control Procedure; signed by both Client and SI; referenced in the SOW.
| Level | What it is | Process | CR form? | Approver |
|---|---|---|---|---|
| 1. Configuration discretion | Minor tweaks within the agreed SDD's intent — workflow approver lists, lookup values, screen layout adjustments | Functional Lead decides; logged in sprint backlog | No | Functional Lead |
| 2. Design clarification | Requirement was ambiguous; decision needed but consistent with SDD | Functional Lead + Process Owner decide; logged in Design Decisions Log | No | Functional Lead + Process Owner |
| 3. Design deviation | Clear divergence from the signed SDD or a contracted requirement | Drafted as Change Request; reviewed at Design Authority | Yes | Design Authority |
| 4. Scope / commercial change | Adds, removes or materially changes scope, cost, timeline or benefits | CR → DA technical approval → Steering commercial approval; ESG if material to BC; may trigger per-wave re-baseline | Yes | DA + Steering (+ ESG if material) |
A Change Request is required when any one of the following applies:
Anything below all four = sprint-level discretion.
Templates the PM walks into Programme Setup & Mobilisation (S10) with: Programme Status Report · RAID Register · Design Decisions Log · Change Request form · Sprint Review pack · Design Authority pack. Pre-drafted, pre-signed where applicable, in Supporting Documentation/Cadence and Change Control/.
Common error: staffing Project Managers in Pre-Programme or Selection. There are no workstreams yet — they appear from Programme Setup & Mobilisation (S10).
The role timeline below shows roles in their canonical singular form. In practice several roles multiply with programme scale. Click to expand for the sizing pattern.
A single Business Architect can credibly carry one functional area; they cannot credibly carry six. A single SI Functional Lead can hold one workstream end-to-end on a small programme; they cannot hold every named sub-process within Finance (AP, AR, Assets, GL, Treasury, Tax) on a 1,500-user programme. Sizing isn't just headcount — it's who can hold a process / module / sub-process domain credibly under load.
| Role | Small (<100) | Medium (100–1,000) | Large (1,000+) |
|---|---|---|---|
| Business Architect | 1 | 1–3 (per major area) | 1 per functional area (4–6) |
| Solution Architect (Client) | 1 | 1 + secondary-platform specialist | Lead + module-specific SAs |
| SI Solution Architect | 1 | 1 + module specialists | Lead + multiple module SAs |
| SI Functional Lead | 1 per workstream | 1 per workstream | 1 per workstream |
| SI Functional Consultants | 0–1 per workstream | 1–2 per workstream | Multiple per workstream (AP/AR/Assets/GL named) |
| SI Technical Lead | 1 | 1 | Often split: Integration + Infrastructure |
| Project Manager | 1 (often combined with PgM) | 1 per workstream | 1 per workstream |
| Process Owners | 1 per process stream | 1 per process stream | 1 per stream + deputies on highest-volume processes |
| Test Analysts (Client + SI) | Light, often shared | Per workstream | Per workstream + dedicated NFT |
| Site Readiness Lead | — | — | 1 per site or per wave (multi-site only) |
When more than one of a role exists, name a Lead. Multiple Business Architects → Lead BA carries cross-domain Accountable duties on RACI; individual BAs are Accountable for domain-specific decisions. Same pattern: Lead Functional Consultant within a workstream Functional Lead's team; Lead SA where multiple module SAs sit; Lead Project Manager on the largest programmes where several PMs need coordination beyond what the Programme Manager carries directly.
Common mistake: SI bids on small-programme sizing with a "one Functional Consultant per workstream" model when the programme is actually medium-to-large. The named consultant becomes the bottleneck mid-build, change requests pile up to add specialists, and the Full Business Case at S12 no longer holds. Sizing the team honestly at Programme Setup & Mobilisation (S10) — and naming Leads where roles multiply — is part of the mobilisation deliverable, not a discovery later.
Same role data as the Roles tab, cut by phase instead of by Gantt. Use this when you want to see who is in the room during a given phase. Toggle to By Role for an alphabetical list with full detail.
One card per binding decision or major deliverable across the lifecycle — not a 300-row task grid. Each card names a single Accountable (the role that owns the call), the Responsible roles doing the work, and the wider Consulted and Informed set. Click any card for the why-it-matters, evidence required, and common mistakes.
Data hygiene runs in BAU before the SI joins, coordinated by the Client-side Data Lead with one Data Owner per master-data domain; data migration kicks in once Discovery profiles the cleansed master data. The handover sits at Programme Setup & Mobilisation (S10) — Data Owners cleansing in parallel, SI Data Migration Lead joining with the Data Lead as Client-side counterpart. Below is the stage-by-stage activity for each half of the thread.
Three concepts worth grounding before walking the stage-by-stage activity below. Click any to expand.
Master Data = the relatively static reference records: customers, suppliers, products, employees, chart of accounts, cost centres. This is what Data Hygiene works on from Value Definition & Case for Change (S2) onwards — cleansing, deduplication, hierarchy fixes, reference-data rationalisation. Target-agnostic: a duplicate customer is a duplicate customer regardless of which ERP wins, which is why the work can start before Selection completes.
Transactional Data = the operational records flowing through the business: sales orders, invoices, GL postings, stock movements, journal entries, payments. Migrated late in the cutover sequence and usually only the open / unfinished records at cutover — the rest archived or aggregated as opening balances. The cutover sequence itself runs master first, then transactional, then reconciliation.
Many Client-side stakeholders think "data migration" is one bucket. It isn't — the two halves have different owners, different timelines, different cleansing demands.
Extract → Transform → Load.
Two clean dry runs at full volume, with reconciliation reports, are non-negotiable. The second dry run at full volume in Testing (S14) is where most defects surface — skipping it pushes them into go-live.
ETL activity is layered, not single-shot. There isn't one moment when "the ETL starts." Six distinct kinds of run across the lifecycle, each teaching you something different about where the data is breaking.
Master first, Transactional layered on. Smoke tests, Dry Run 1 and the first half of Dry Run 2 migrate Master Data only. Once Master Data lands cleanly into the target's reference tables, Transactional Data ETLs are layered on top — sales orders against valid customer/product references, GL postings against the chart of accounts. Master is the smaller and more stable population (test the pipeline first); Transactional integrity depends on Master being correct in the target.
Typically 4 to 6 ETL iterations across S13 into S14 as defects surface, get fixed, and the next run reveals the next layer. Where the ETL pipeline is automated, runs can continue iteratively until no errors surface — modern migration tooling supports this. Manual ETL processes constrain the iteration count and reliably leave defects in flight at cutover.
Completion deadlines. Data Hygiene must be substantially complete by the start of Discovery (S11) — SI profiling depends on cleansed master data, not work-in-progress. Migration completes at cutover (S16) with post-go-live validation in Hypercare & Stabilisation (S17).
Errors during ETL split into four types with different ownership. Lock this split into the SOW at Programme Setup & Mobilisation (S10).
Warning: do not let the SI fix data quality issues. The Data Lead drives Data Owners to fix data hygiene defects in the source system, then re-runs the ETL. SI billing day rates to clean Client master data is one of the most reliably expensive sub-patterns in data migration — and it leaves the Client carrying dirty data into BAU because the SI fixed the symptom, not the cause.
SIs implementing a given platform (D365 F&O, D365 CE, SAP S/4HANA, Workday) have done data migration dozens of times. They have proven templates: object catalogues per module, mapping templates per master-data domain, reconciliation templates, dry-run runbooks, defect-classification taxonomies. Ask for them on day one of Programme Setup & Mobilisation (S10). Don't start blank.
The Keystone artefact library carries worked examples — the Customers Data Migration Mapping (Excel) and Customers Data Migration Specification (Word) on the Supporting Documents page — as a structured starting point the Client walks into the first SI workshop with.
Warning: SIs that won't share their templates are a warning signal — either they don't have them (concerning), or they want to bill for redoing the same thinking (also concerning). The Client owns the methodology; the SI brings proven accelerators that fit it.
A focused three-stage workstream from Solution Design & Full Business Case (S12) through Testing (S14). The catalogue is built at S12, interfaces are built and unit-tested at Build & Configuration (S13), and SIT plus end-to-end integration testing closes the workstream at S14.
Three integration patterns cover most ERP programmes. The choice is a strategic decision — locked at Programme Setup & Mobilisation (S10), reflected in the Solution Design at S12, and impossible to change cheaply once the catalogue is built. Click any to expand.
Single switch. Every integration goes live on Day 1. All finance and SCM interfaces — ledger postings, supplier invoicing, customer billing, stock movements, banking, tax — are cut over together at the same cutover event.
Suits: simpler estates; single-platform replacements; programmes with an urgent or regulatory-driven business case; organisations where keeping two systems in parallel is operationally untenable.
Plus: shortest legacy retirement (no parallel running); cleanest commercial position with the SI (firm pricing for one cutover); single set of dry runs; benefits land fastest.
Warning: highest single-event risk. Every integration must be tested and ready simultaneously. Defects cascade across multiple interfaces at once. The Cutover Go/No-Go (S15) is binary across the entire estate — one weak link can stop the whole switch.
Multiple cutover events. Brands, entities, regions or factories migrate over time. Temporary integrations bridge the legacy systems and the new platform during the transition, keeping both in sync as each unit comes online. Permanent integrations are built once for the new estate and replace the temporary ones as legacy retires.
Suits: multi-brand, multi-entity, multi-region or multi-factory rollouts; estates that cannot all migrate simultaneously due to scale, risk appetite or operational windows; programmes where benefits per wave justify the parallel cost.
Plus: lower per-event risk (one wave at a time); learnings from each wave reduce defects in the next; commercial framework supports per-wave re-baselining; reversibility per wave is realistic.
Relationship to multi-wave delivery. Phased integration is about how interfaces are sequenced; multi-wave is about how the programme delivers. They typically coexist — a multi-wave programme nearly always uses temporary integrations bridging legacy and new while waves migrate, with permanent integrations retiring the bridges as legacy winds down. Big Bang is incompatible with multi-wave; Greenfield is only compatible if each wave is genuinely greenfield.
Warning: the most expensive trap is letting temporary integrations become permanent because nobody owns retirement. Lock the temp-integration retirement plan in the Solution Design at S12 — named owner, target retirement date per integration, named successor permanent integration. Carrying two systems and the bridges between them for years is one of the most reliably under-estimated costs in multi-wave programmes.
No legacy integrations to retire, no historical data to migrate beyond opening balances. Suits a new business unit standing up, an M&A integration onto the parent's platform, a brand-new operation, or a divestment establishing standalone systems.
Plus: the cleanest possible cutover — no parallel running, no temp integrations, no legacy data quality issues to inherit. Highest control over data quality from day one. Fastest realisation of benefits because there is no compromise required for backward compatibility.
Suits: standing up a new entity; carve-outs from M&A; new geographies or factories with no inherited systems; replacing a discrete legacy that has no upstream dependencies.
Warning: Greenfield can be a smokescreen for not addressing the legacy estate — standing up new clean while the legacy mess persists alongside, deferred indefinitely. Be honest in the business case about what gets retired vs. what stays running. Otherwise the “greenfield” programme delivers benefits on paper while operations carry both estates.
A six-stage workstream from Programme Setup & Mobilisation (S10) through Cutover Planning (S15). Environments first, then security architecture, then performance and security build, then the full NFT cycle, exiting with an NFT certificate at S15.
Every cost, threshold and qualifying criterion in this tab is an illustrative list-price order of magnitude, dated May 2026. Pricing changes frequently. Real list-to-net discount can be 30–60% at scale, varies by user type, module mix, volume tier, multi-year commitment and region. Verify current figures directly with the vendor or with your SI before committing to anything in a Full Business Case, an SOW, or a board paper.
These figures exist to help frame the conversation, not to quote in one.
Cloud ERP introduces the vendor (Microsoft, SAP, Oracle, Workday) as a third commercial party alongside the Client and the SI — with prerequisites, controlled environment lifecycles, and paid service tiers. On-prem keeps the relationship Client-and-SI, with the Client owning environments and infrastructure. The deployment-model decision lands at Solution Confirmation (S8); the vendor service-tier decision lands at Programme Setup & Mobilisation (S10); the cost flows into Solution Design & Full Business Case (S12). Get the order wrong and the vendor commercial conversation arrives at Cutover Planning (S15) under time pressure.
Per-user licence is one cost lever — TCO is the conversation. Total cost of ownership over five years also includes implementation (SI cost from the Full Business Case at S12, typically one and a half to three times the licence cost over five years for a mid-to-large programme), platform run cost (subscription or on-prem hosting), internal team cost (PM, BA, Process Owners, Data Lead, Test Manager, Change Lead, Hypercare), hypercare and stabilisation, and optimisation, future upgrades and scope expansion. Compare platforms on TCO, not the list-price headline.
The deployment model isn't just an infrastructure choice — it changes who controls environments, when cutover windows can run, how code reaches production, and where operational responsibility sits after go-live.
Microsoft Lifecycle Services (LCS) provisions production environments and gates code deployment for Dynamics 365 F&O / SCM. CE is cloud-first; legacy AX 2012 is on-prem. Vendor accelerator tier here is the most consequential of the four platforms because Microsoft funds FastTrack for qualifying programmes — if the SI doesn't engage it, the Client misses out.
SI doesn't engage Microsoft FastTrack despite the Client qualifying. Microsoft would have funded it for free, but no one made the call. Client ends up paying for accelerated provisioning effort the vendor would have provided at no cost.
What drives your cost
SAP Cloud ALM is the deployment-services backbone for cloud S/4HANA; SAP Activate is the methodology that applies to both cloud and on-prem paths. The on-prem path keeps environment lifecycle Client-controlled, which materially changes Cutover Lead authority at Cutover Planning (S15).
Programme picks cloud S/4HANA on cost grounds without modelling the loss of Cutover Lead control. Cutover window options narrow to what SAP can support — the Client discovers this when the first rehearsal is being scheduled, not when the model was chosen.
What drives your cost
Oracle Cloud ERP is cloud-only; legacy E-Business Suite remains the on-prem path for organisations not yet on cloud. Pod provisioning windows and the quarterly release calendar are the dominant constraints when planning cutover and major test cycles.
Test plan ignores the quarterly release calendar. UAT runs straight through a pod refresh window and loses test data; the rebuild cost shows up as a slip rather than a planning miss.
What drives your cost
Workday is cloud-only with no on-prem option. Tenant lifecycle is managed by Workday, with the configuration migrator handling tenant-to-tenant moves. Acceleration runs through certified delivery partners rather than a Workday-funded programme like FastTrack.
Programme treats the two Workday major releases per year as background noise. The release window collides with cutover and the tenant configuration that worked yesterday needs revalidation today — another rehearsal cycle the plan didn't carry.
Workday is typically priced on employee headcount + module mix rather than purely per active user. HCM is priced per employee (counts all employees); Financials is priced per Full user.
What drives your cost
Clients legitimately ask: we've paid the SI to deliver — why are we now being asked to engage with the vendor too? The honest answer: the SI's job includes navigating the vendor relationship — knowing your eligibility for free accelerated tiers (especially Microsoft FastTrack) and securing them; aligning test exit criteria to vendor prerequisites at Programme Setup & Mobilisation (S10); owning vendor submissions through the lifecycle. The SI should not drop you in it. But vendor knowledge sits with the vendor — even the best SI doesn't have direct routing to vendor product engineering. The methodology's question isn't “SI or vendor”; it's “did the SI factor vendor service properly into the FBC at Solution Design & Full Business Case (S12), and was the Client's eligibility for free or paid tiers exercised in time?”
Identify the audience before drafting or reviewing. Exec content is short, decision-oriented, jargon-free. Programme/team content carries detail, RACI, sequencing.
Describe what you or the team are working on. The finder maps it to the most likely stage in the method.
Pick any stage to see its exit-criteria checklist. Tick items as evidence is in — state is saved in your browser between visits. Funding Envelope (S6) and Solution Design / FBC (S12) are the two real board gates and have firmer thresholds. The chip shows percent complete; chips turn green when fully complete.
Tick items as evidence is in. Funding Envelope (S6) and Solution Design / FBC (S12) are binary board gates — sponsor signs only when all items are evidenced or formally accepted.