What a decade of Phoenix tells us about the public service

By Karl Salgo, Associate and Executive Advisor

Published on March 3, 2026 in the Ottawa Citizen

There is much Prime Minister Mark Carney can learn from the pay system fiasco in his goal to make government more efficient.

The Phoenix pay system is still impacting public servants 10 years after it was launched. Photo by JEAN LEVAC /POSTMEDIA

Ten years after the rollout of the federal government’s notoriously unsuccessful Phoenix pay system, the enduring mystery is less why it failed than why the government launched an initiative whose flaws were recognized before it even started and then persisted with it for years afterwards.

Part of the answer lies with a public sector culture that Prime Minister Mark Carney might want to bear in mind as he contemplates a “quiet revolution” in government operations.

As the prime minister plans to improve government efficiency by 15 percent, including through much-needed structural reform and the use of technology, he may want to note some lessons from Phoenix.

First, government is complex – theoretically sound solutions don’t always work and should be tested in advance.

Second, internal capacity isn’t necessarily just fat and shouldn’t be gutted prematurely. And last, effective oversight and risk management must focus more on outcomes and trust that good optics will follow.

Phoenix was supposed to save more than $70 million annually by replacing a crumbling legacy system but has cost Canadian taxpayers $5 billion and counting since implementation, in addition to leaving widespread and sometimes traumatic hardship in its wake. Hundreds of thousands of backlogged cases continue to dog the system even as the government moves slowly towards its replacement program, Dayforce.

Why did Phoenix fail? Public service unions will tell you that public sector employment agreements are too varied and complex for the kind of off-the-shelf, one-size-fits-all software supplied by the external contractor, IBM, and that the government made emerging problems virtually unfixable by gutting internal payroll capacity before the system was even launched. While public service unions are never keen on outsourcing, this perspective is largely true.

The program was almost certainly launched prematurely in essentially an everything-everywhere-all-at-once approach.

A single-department pilot planned for 2015 was cancelled and warnings from both IBM and the pay centre that the system wasn’t ready were largely ignored. Technically, there were two stages, but the second followed within weeks of the first and ignored already emerging problems.

By 2017-2018 the Canada’s auditor general was already describing Phoenix as an “incomprehensible failure,” decrying a lack of advance testing and oversight.

For a long time, officials desperately threw money at ever-worsening problems. Only after several years and in the face of a parliamentary report calling the system an “international embarrassment” did they admit outright defeat and begin the protracted search for a replacement. There were some systemic and cultural reasons for this.

First, in deciding to launch an untested system, public service executives were faced with an expectation of meeting annual savings targets within a budget framework that had been set seven years earlier. As the auditor general later observed, the senior public service has an obedient culture.

Public servants don’t like to be seen as obstructionist, and they evidently preferred to compromise functionality rather than push back on timing and costs. (Ottawa commuters may discern parallels in the history of our light rail transit system.)

The obedience issue is worsened by the public sector’s approach to risk, which focuses too much on avoiding blame and bad optics and too little on actual outcomes. In such a culture, there’s a great temptation to believe that problems can be managed with the right narrative. In fairness, the government did commission two external studies in advance, only one of which suggested that the risks might outweigh the benefits.

I suspect that the supporting analysis may also have been too theoretical. In addition to adopting a commercial off-the-shelf system, Phoenix consolidated pay functions in a new centralized pay centre located in Miramichi. At the time, back-office consolidations were in vogue, based on expected economies of scale that looked reasonable on paper but often didn’t materialize.

Officials may have convinced themselves that it was just a matter of time before the changes paid off and thus threw good money after bad, a cognitive bias that economists call the sunk cost fallacy. In effect, the government spent billions trying to avoid wasting millions. Admittedly, no quick fix was available, as the incremental adoption of Dayforce suggests.

There were also failures in oversight and accountability, as the auditor general pointed out. While people joke about bureaucrats establishing committees to oversee committees, Phoenix should have had intensive oversight and reporting beyond the officials directly responsible for the project.

These things said, Phoenix would almost certainly have been axed sooner if it served the public rather than the public service. The government is not a harsh employer but it is an impersonal one, and it doesn’t face the same accountability for harm to bureaucrats that it would face for other citizens.

But as Phoenix goes to show, a failed initiative can bring harm to both.