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

Transformation management and PMO

Why do IT projects fail? A reading through the signs of drift

In brief

An IT project fails, under the convention adopted here, when it misses its deadline, its cost, its scope or the expected benefit beyond the variance its sponsor had accepted. Rather than a list of causes, Management Square offers a reading method here: the observable signs of drift at each stage of the project, what each one can reveal and the response it calls for.

Page contents

To say that an IT project has failed is to apply a convention. Here we adopt the following one: a project fails when it misses, beyond the variance its sponsor had accepted, one of the commitments made at its launch, namely a deadline, a cost, a scope and an expected benefit. To the question “why do IT projects fail”, this article therefore answers neither with a list of causes nor with a ranking, and it relies on no statistics.

It offers a reading method. The variance on a commitment may be observed late, sometimes after go-live. Signs may precede it, visible in the project’s documents and bodies: at framing, in governance, in the schedule and the scope, then during acceptance testing. For each, we offer a reading, what it can reveal, and a response.

A project’s commitments, and the moment the variance is observed

Each commitment has its reference, and its time.

Commitment Reference Moment the variance is observed
Deadline The go-live date recorded in the launch file, and its approved revisions At each update of the forecast, then on the planned date
Cost The authorised budget, contingencies included At each re-estimate of the remaining spend
Scope What the project must deliver, as accepted by the sponsor At each request to add or remove, then at acceptance testing
Expected benefit The effect announced to justify the spending, with its baseline After go-live

Other conventions add the quality of what is delivered or user satisfaction. By choice, ours leaves outside failure two cases decided with full knowledge of the facts: the project stopped because the remaining cost exceeded what it could still bring in, and the project one of whose commitments was revised with the sponsor’s agreement.

When the acceptable variance has never been written down, the finding of failure becomes a matter of judgement.

Commitments made at framing before they were grounded

The first signs can be read in the launch file. A go-live date appears in it, coming from an external deadline or an announcement, although the work has not been estimated by the people who will do it. The sign may indicate that this date is a constraint, to which scope or resources will have to bend. It calls for an explicit statement of the degree of certainty of each deadline, which our article on the IT project roadmap details.

Another sign: the expected benefit was written by the project team, without a user department having taken ownership of it. The fourth commitment then risks being neither met nor observed. This sign calls for a designation before the build: the person who will report on the effect obtained, and the baseline value against which it will be compared.

A last sign at framing: the plan rests on assumptions that nobody has listed, such as the availability of key users, the state of the data to be migrated or the behaviour of a neighbouring system. If proven wrong, they may take the form of delays. The response is a list, and a trial for those whose failure would cost the most, following the approach described in our article on managing complex projects.

In committee, decisions that do not come

Governance leaves written traces, in which the signs can be noted: the minutes of the steering committee record no decision, the same request for arbitration appears on the agenda of several meetings in a row, the sponsor delegates their seat to someone who does not have the power to decide.

These signs may reveal that the project is waiting. Part of the delay that will later appear in the schedule may form here, in the time between a request and its answer.

The response requires no additional resources: a deadline written opposite each decision requested, with what overrunning it costs. The meeting format that makes this possible is described in our article on the project review.

A schedule that slips, a scope that swells

During the build, the schedule may show several signs. The end date slips at each update, by a period close to the interval between two updates. Tasks remain “in progress” from one report to the next. The remaining work is obtained by subtraction, initial budget minus consumed, without re-estimation. Milestones are declared reached with reservations that nobody lifts.

These signs may reveal that the schedule no longer describes the real work. They call for a re-estimate of the remaining work by those who produce it, based on what is finished and verified. When several projects show these signs, the measurement of progress needs reviewing across the organisation, the subject of our page on setting up a PMO.

The scope has its own: requests come in without anything going out, without the date or the budget moving, and without a written trace. In an agile approach, the sign takes another form: the backlog grows faster than it empties, and its order changes with the latest requester. In both cases, the sign may reveal that nobody exercises the power to refuse; in Scrum, ordering the backlog falls to the product owner. This sign calls for a rule: every entry comes with what it displaces, in date, in budget or in content.

Acceptance testing, where accumulated variances become visible

It is at acceptance testing that scope is observed, and that the delays of earlier phases may be passed on. Several signs can be noted there. Its start is postponed while the go-live date is maintained. Defects are opened faster than they are closed. Blocking defects are requalified without having been fixed. The users responsible for testing are not released from their day-to-day work. Load tests and data migration are pushed back to the last few days.

These signs may reveal a trade-off made without being stated: the date is kept at the expense of what is delivered, in the form of defects that users will discover and of technical debt. They call for go-live criteria written before acceptance testing begins, and an explicit decision by whoever signs it off, the project owner (maîtrise d’ouvrage) in the vocabulary recalled in our entry MOA, MOE, AMOA.

A last sign concerns the expected benefit. The old system has no shutdown date, and nobody is responsible for checking usage after go-live. The project can then meet its deadline, its cost and its scope, and miss the fourth commitment. Preparing adoption falls under change management.

Why IT projects fail, a question to investigate on the evidence

These signs give direction, without establishing the cause. A schedule that slips may come from an estimate that was too short, a pending decision, a skill withdrawn or a technical solution that does not hold. For a given project, the “why” is established by checking each explanation at its source: documents, measurements, cross-checked accounts.

Isolated signs, to which the committee obtains answers, are dealt with in routine monitoring. If they accumulate at several stages of the project, or if the explanations gathered do not agree, internal monitoring may no longer be enough to decide between them. A project audit addresses this situation: our page presents this independent review, from the engagement letter to the findings presentation.

The “why” that comes out of it holds for this project, at this date. It says nothing about projects in general, and it is on this basis that the sponsor decides to continue, reframe, suspend or stop.

Insights

Further reading

All insights
A programme's master schedule, built around decisions, dependencies and margins

A master schedule is, in common usage, the overall schedule of a programme or a large project, reduced to decision milestones, dependencies between projects and the margins that protect them. Management Square proposes here a construction in five stages, from scope to the update rule, and clarifies what separates it from a project's detailed schedule and who maintains it.

Read the analysis
Product owner or product manager, two functions to distinguish before appointing anyone

In the Scrum Guide, the product owner is the person accountable for the value of a product and for its Product Backlog; product manager is a job title, absent from that guide, which in practice refers to the person in charge of a product's trajectory. Management Square compares product owner and product manager here (horizon, scope, counterparts, decisions) and describes the possible set-ups.

Read the analysis
In programme governance, who decides, at what level and on what information?

Programme governance refers to the rules that set who decides what in a programme, at what level, on what information and within what timescale. Management Square proposes here an allocation of decisions, following the questions of an executive committee about to fund a programme: inventory of decisions, delegation thresholds, information for the board, disputes between projects, closure.

Read the analysis

FAQ

Frequently asked questions

Another question? Ask us

When can an IT project be said to have failed?

The answer depends on the convention applied. The one we adopt compares the project with four commitments made at its launch, namely a deadline, a cost, a scope and an expected benefit, and speaks of failure when one of them is missed beyond the variance the sponsor had accepted. Other conventions add the quality delivered or user satisfaction.

Is there one main cause of IT project failure?

We do not name one. Ranking causes would require a common measure, applied to comparable projects, which this article does not claim to hold. For a given project, on the other hand, the cause can be established: you work back from the signs observed to the decisions taken or left pending, checking each explanation against the documents and with several people.

Is a project stopped before completion a failure?

Not under the convention adopted here, when the stop was decided with full knowledge of the facts. A sponsor who stops a project because the remaining cost exceeds what the project can still bring in is taking a management decision. Failure rather describes the project that nobody stopped or reframed, and that misses its commitments without a decision having accepted it.

Can a project delivered on time and on budget have failed?

Yes, under the convention adopted here, which counts the expected benefit among a project's commitments, alongside deadline, cost and scope. A tool can be in service on the planned date without being used, or be used while the old system remains open. This can only be observed after go-live, which means having named, before launch, the person who will report on it and the baseline against which the result will be compared.

Does an agile approach protect a project from drift?

Not by itself, in our reading, which is a line of reasoning and rests on no measurement: above all it changes the form of the signs of drift. Frequent releases can make progress easier to observe, provided they are put into users' hands. The signs to read can then be a backlog that grows faster than it empties, an order of priority that changes with the latest requester, finished increments that are not put into service.

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.