Skip to content
Technology transformation expertise firm · ARDIANE Group Innovate, secure, succeed

Transformation management and PMO

Best practices for managing complex projects, in the order in which to apply them

In brief

A complex project is, by working convention, a project whose outcome depends on dependencies, stakeholders and uncertainties that its plan cannot fix in advance, whatever its size. Management Square proposes here an order for applying best practices in managing complex projects, from the written qualification at scoping through to setting the steering rhythm.

Page contents

A complex project, according to the working convention used in this article, is a project whose outcome depends on elements that its plan cannot fix in advance. These elements are of three kinds: dependencies on what the project does not control, stakeholders whose expectations diverge, uncertainty about the need or the solution. Size does not enter into this definition. In managing complex projects, best practices follow from these three factors.

We propose applying them in this order: qualify what makes the project complex, test early the assumptions on which the rest depends, reduce dependencies and then keep hold of those that remain, give stakeholders a decision process, and finally set the steering to the speed at which the project learns. Each produces a specific document and has an owner.

What distinguishes a complex project from a merely large one

A large project has many tasks, teams and budget lines. If each task is known, if the links between them can be described and if experts can say in advance what will produce the outcome, it remains predictable. The planning and coordination effort grows with size, without the way of running it changing.

The Cynefin framework, created by Dave Snowden, provides a similar reference point. Its presentation on the cynefin.io wiki distinguishes three kinds of systems, including ordered systems, where cause-and-effect relationships are clear or can be discovered through analysis, and complex systems, which can only be understood by interacting with them. Transposed to a project, this distinction leads us to a practical rule: in the first case, more analysis improves the plan; in the second, some answers only come from a trial.

Factor Large, predictable project Complex project Sign to look for
Dependencies Numerous, internal to the project or set by contract On teams, systems or decisions that the project does not control Milestones waiting for an external delivery or decision
Stakeholders Numerous, with a client who decides Several divisions with divergent objectives, with no designated arbiter The people whose agreement is required to accept the outcome
Uncertainty The need and the solution are known, execution remains The need, the solution or both only become clear by trying The plan’s assumptions that no trial has confirmed

A project can be complex through just one of these factors, and with a small team.

A written qualification before the first schedule

The first practice consists of identifying, at scoping, what makes this project complex. It produces a short note: the external dependencies with, for each, a contact; the people whose agreement is required, with the objective each pursues; the unverified assumptions, ranked according to what it would cost if they proved wrong.

The project manager writes this note and the sponsor validates it, because it sets what follows. A project that is complex only through its technical uncertainty does not need the arrangements suited to a project disputed between four divisions. The note therefore serves to choose, among the practices that follow, those that apply and with what intensity.

It is reread at each decision point, because complexity shifts: an uncertainty is removed by a trial, a dependency appears with a new supplier.

Testing early what no one yet knows

When uncertainty dominates, the order of work is set by what it teaches. The assumptions whose disproof would cost the most come first, each with a trial in proportion: a mock-up submitted to real users, a prototype of the least known interface, a load test, a pilot on a reduced scope. The project manager answers for this list of trials and their results. It complements the register described in our article on project risk management.

Commitment follows the same logic. The project commits to dates and a budget for the tranche that trials have made predictable, and announces a range for the rest. Each tranche ends with a decision point at which the sponsor continues, redirects or stops, on criteria written at the start of the tranche. This practice produces a plan in tranches, for which the sponsor and the steering committee answer since they authorise each one. How to display these degrees of certainty is detailed in our article on the roadmap of an IT project.

Fewer dependencies, and dependencies kept in hand

A dependency is reduced before it is tracked. Several means fall under the design of the project: agreeing early on a stable interface with the neighbouring system, so that each team moves forward without waiting for the other; testing against a simulator as long as that system has not been delivered; integrating components at short intervals, without waiting for the end of the journey. Another falls under organisation: assigning full time the skills on which the project depends most, on the model of a project-based organisation. The project’s technical lead answers for the interfaces and the integration calendar.

The dependencies that remain are kept in hand one by one: the person who commits to delivering, what they deliver, the deadline beyond which the project falls behind, and a confirmation requested again at each steering cycle. The project manager answers for them.

It happens that the project splits into several projects that depend on one another to achieve the same outcome. The object to be run has then changed in nature, and the instruments become those of programme management.

A decision process for stakeholders

A project that is complex through its stakeholders brings together divisions none of which commands the others, and whose requirements may contradict one another. The appropriate practice is a decision map. For the scope, for the priority between two incompatible requirements and for acceptance of the outcome, it names the person who decides, those they consult, their response time and the recourse in the event of disagreement.

The sponsor answers for this map, because it is up to the sponsor to obtain the commitment of their peers. The project manager prepares the requests and keeps the record of decisions made. The map separates two roles that a workshop can confuse: contributing to the expression of the need, and having the power to accept or refuse the outcome.

Complex project management best practices, in proportion

The faster the project changes, the shorter the steering cycle. At each project review, the project manager’s report adds to progress what the project has learnt since the previous one: assumptions confirmed or disproved, dependencies confirmed or slipped, decisions pending. On a project that is both large and complex, keeping these documents can be entrusted to a project management office, or PMO, under the authority of the project manager. Before a heavy commitment, a project audit makes it possible to check that the assumptions taken for granted have indeed been confirmed by a trial.

These practices have a cost: the time of the sponsor, experts and users. Applied to a large but predictable project, they weigh it down without teaching it anything. Conversely, an end-to-end detailed plan, laid over a complex project, displays precise dates that nothing supports.

The initial qualification serves as a setting, in both directions. When trials have removed the assumptions and dependencies are confirmed, the next decision point can record the return to a simpler way of running the project.

Insights

Further reading

All insights
A stalling transformation: the diagnosis to make for a successful digital transformation

A successful digital transformation depends essentially on five management mechanisms: governance that decides, a portfolio held to real capacity, arbitrations made on time, adoption prepared with the business and measured benefits. Management Square offers here the diagnosis of a stalling transformation: from the signal observed to the failing mechanism, then to the remedies, in increasing order of cost.

Read the analysis
Should you adopt SAFe for agile project management at scale?

SAFe is a framework that has several agile teams, grouped into release trains, plan and deliver together at a fixed cadence. Management Square sets out here the conditions under which SAFe suits agile project management at scale, the lighter alternatives and the situations where it is better to refrain from it.

Read the analysis
At group scale, the questions raised by the digital transformation of large enterprises

The digital transformation of large enterprises stands out by the number of its decision-makers, information systems and project portfolios: each entity can have its own. Management Square addresses here the questions a group executive committee must settle: the level of each decision, shared services or autonomy, legacy systems, a saturated portfolio, arbitration between divisions.

Read the analysis

FAQ

Frequently asked questions

Another question? Ask us

What is a complex project?

By working convention, a complex project is a project whose outcome depends on elements that its plan cannot fix in advance: dependencies on what it does not control, stakeholders with divergent expectations, uncertainty about the need or the solution. Other conventions exist: some organisations classify their projects by budget, duration or criticality. This one has the advantage of designating what changes in the way the project is run.

What is the difference between a complicated project and a complex project?

According to a customary distinction, a complicated project involves many elements whose behaviour experts can predict through analysis, whereas a complex project involves elements whose interaction is only known by trying. A practical rule follows: the first is run with a detailed plan, the second with trials and closely spaced decision points. The Cynefin framework, created by Dave Snowden, makes a similar distinction between ordered systems, where cause-and-effect relationships are clear or can be discovered through analysis, and complex systems, which can only be understood by interacting with them.

Are agile methods enough to run a complex project?

If, by convention, three complexity factors are used (dependencies, stakeholders, uncertainty), they address the third. Working in short iterations, with deliveries submitted to users, suits uncertainty about the need or the solution. On its own, it does not resolve dependencies on other teams or other systems, nor arbitration between divisions with divergent objectives, which require their own practices.

When should a complex project be run as a programme?

When it is made up of several projects, each with its own manager and deliveries, which depend on one another to produce an outcome that none would achieve alone. Programme management then takes charge of the overall trajectory, the dependencies between projects and the arbitrations. A complex project that keeps a single team and a single manager remains a project.

How do you estimate the timescale of a complex project?

In tranches. The project commits to a date for the part that trials have made predictable, and gives a range for the rest, narrowed at each decision point. A single date announced at launch for the whole rests on assumptions that no one has yet verified, and the sponsor must know which ones.

A project to bring to completion?

A first conversation is enough to understand your challenge, identify the right people and tell you frankly what we can bring.