Under the name of IT programme, organisations run undertakings that do not follow the same rules. One moves forward from milestone to milestone, split into work packages, each with its own scope, budget and sometimes contract. Another brings together long-lived teams that replan together at a fixed cadence, with no scope set beyond the current cycle. A last one combines the two. What distinguishes them is an IT programme management model: the way of planning, of measuring progress, of allocating decisions, and what the programme expects of its teams.
We compare these three models here on these four dimensions. They are models in common use, which we describe without adopting the definition of a reference framework. The distinction between project and programme appears in our article on the difference between programme and project, the allocation of decisions by level in the one on programme governance, the overall timetable in the one on the master schedule.
The models of programme conduct in their ordinary form
The programme driven by milestones and work packages splits the objective into bounded work packages: a common solution, a data migration, deployments by site. Each work package has a plan, a budget and an end date, and the programme links them through milestones where a decision is taken. The logic recalls that of PRINCE2, a project method in which the steering committee authorises one stage at a time: the programme commits to a stage, observes its result, then decides on the next.
The programme organised in trains or value streams starts from long-lived teams. The SAFe framework publishes a detailed form of it. According to the public part of its site (framework.scaledagile.com, read on 12 November 2026), the Agile Release Train is a long-lived team of agile teams, which incrementally develops and delivers one or more solutions in a development value stream, and generally numbers 50 to 125 people (page with no update date displayed). It moves forward in Planning Intervals of 8 to 12 weeks (page updated on 25 February 2025), opened by a PI planning event that generally lasts two days (page updated on 15 March 2026). The wording is ours.
The hybrid programme brings the two regimes together under a single direction. A platform run by a train sits alongside a data migration or a deployment subject to an external date, run by work packages. The two regimes meet at common milestones: integration, cutover, opening of a site.
Planning, measuring, deciding under each IT programme management model
| Milestones and work packages | Trains or value streams | Hybrid | |
|---|---|---|---|
| What it plans | A scope, work packages and milestone dates, set at launch and revised at each stage | A cadence and a capacity; the content is decided cycle by cycle | Dated common milestones, and a cadence for the components that have one |
| What it measures | The variance from a baseline: dates met, budget consumed, scope delivered | What is demonstrated on the assembled system, and the achievement of the cycle’s objectives | Two measures side by side, reconciled on committed costs, risks and dependencies |
| Who decides | The programme committee at each milestone, the programme director in between, within written tolerances | Content owners and business owners present at each planning event | A common body for the shared milestones, each component for its own operation |
| What it assumes of the teams | Teams formed for a work package, sometimes at a supplier, and a need stable enough to be set | Long-lived, cross-functional teams able to assemble the system at each iteration | Written interfaces between components and an office able to reconcile the two measures |
Planning separates the first two models most clearly. The milestone model commits to a scope and a date, then manages the variances. The train model commits to a capacity and a rhythm. SAFe’s site thus classifies the objectives of an interval into committed and uncommitted objectives (page updated on 18 November 2025, according to its metadata).
Measurement follows planning. In the first model, progress is read against the baseline plan. In the second, it is read from what works: the framework’s site presents the System Demo, held at the end of each iteration, as an objective measure of progress towards the interval’s objectives (page updated on 18 March 2026). Our article on the agile PMO describes what this change does to the work of the project office.
Decision-making changes place. SAFe’s site gives the Business Owners primary business and technical responsibility for return on investment, governance and compliance, and has them take part in the train’s events (page updated on 12 March 2025). Funding then goes through the portfolio, whose logic is presented in our entry on lean portfolio management: a budget allocated to value streams rather than to projects.
The situations each model suits best
The milestone model suits when the result has a set form and a date that does not depend on the programme: a regulatory deadline, the end of a hosting contract, a single cutover that no splitting can avoid. It also suits when work packages are entrusted on a fixed-fee basis, since the supplier commits to a scope and a date; our comparison between fixed fee and time and materials sets out the consequences.
The train model suits a solution expected to evolve with no end date, carried by enough interdependent teams for common planning to have a purpose, in an organisation that agrees to fund a capacity. The conditions for adopting SAFe are examined in our article on SAFe and scaled agile project management.
The hybrid model answers programmes that mix a long-lived product and finite operations. Its risk is maintaining two schedules that only meet at the end. The common milestones are therefore set from launch, with, for each one, what each component must contribute to it.
A poorly chosen model can be spotted by certain signs. Work packages replanned at every committee may indicate a need too unstable to be set. A train whose objectives are read back as firm commitments may indicate a leadership that in reality expects a date and a scope.
The work of direction changes with the model
The model sets the nature of the work of direction. With milestones, it holds the baseline, the tolerances and the preparation of stage decisions. With trains, it ensures that each cycle’s objectives serve the expected effect, and carries what goes beyond a train: external dependencies, deadlines, benefits. In the hybrid, it does both, with a programme office able to consolidate information of different kinds. Our page on setting up a PMO presents how this office is formed, the one on programme management the way we support a programme through to its benefits.
The choice is made component by component. We suggest filling in three boxes for each: what is fixed (date, scope, capacity), what can move, and who decides on it. A component whose date and scope are fixed falls under milestones; a component of which only the capacity is fixed falls under a train. A component that cannot put anything in the first box needs framing before any model.