REPORT TO REVEAL IF PHOENIX FIXES ARE WORKING

Union chief wants government to consult with its own staff experts on a new system

When Auditor General Michael Ferguson tables his latest report this Tuesday, much attention will focus on his analysis of the Phoenix pay system — in particular, whether the fixes being applied to this epic information technology disaster will actually work.

On recent evidence, the answer seems likely to be “no.” The federal government’s pay system was struggling last month with a backlog of some 265,000 transactions, not counting a nearly equivalent number of administrative queries. The upshot: There’s been zero progress since spring.

This, despite millions of dollars in overtime expenses for privatesector contractors, and the hiring of more than 200 additional pay advisers at satellite offices in Winnipeg, Shawinigan, Montreal and Gatineau. The $310-million project to modernize and consolidate multiple pay systems across the federal government is now nearly $200 million over budget with no end in sight to the repair bills.

Debi Daviau, president of the Professional Institute of the Public Service of Canada, earlier this week suggested it will take at least three more years to get Phoenix to steady state. The current approach, she added, may never succeed.

Daviau’s prescription: “We need a pay system that works, and we have the people to build it,” she said.

Daviau made clear in a followup interview that she was not proposing that her union take over the job of fixing Phoenix. While PIPSC has 57,000 members, many of them experts in information technology, fewer than 30 are directly involved in the Phoenix project.

Daviau says she merely wants the managers at Public Services to consult with the experts on the ground — in her union’s case, the software jocks. These are the folks who understand the PeopleSoft technology that underpins most of the government’s pay transactions. And they are telling Daviau that building a new pay system on the foundation of PeopleSoft’s latest technology platform (version 9.2) can be done in as little as a year.

There’s much more to it, of course, and Daviau understands that. The underlying technology merely captures the pay data. Processing it requires hundreds of employees trained on a system with more than 80,000 rules governing the application of pay in all its variants, from maternity leave to overtime, across 27 collective agreements. Members of the Public Service Alliance of Canada, the largest federal government union, do much of the processing at a centralized pay centre in Miramichi, N.B.

No matter the complexity, designing a pay system isn’t rocket science. What’s required is careful testing from the ground up, bargaining unit by bargaining unit, department by department.

Managers need to pay attention to how the changes are affecting those on the front lines. In the case of Phoenix, those in charge simply did not listen, a number of government-commissioned studies have suggested. A key consideration for Ferguson is whether this attitude has changed and, if not, what should be done about it.

The peculiar deafness of federal managers was captured neatly in the “lessons learned” study published earlier this year by Treasury Board. Management consultants at Goss Gilroy Inc. spent nearly five months interviewing key players involved in the development and rollout of Phoenix, which began in February 2016.

“We heard that, given that pay transactions were carried out by compensation advisers who are lower-ranking employees,” the consultants noted in their final report, “decision-makers may have underestimated their role and undervalued their expertise.”

Gartner Inc., the consulting group hired in 2015 to assess the risks associated with Phoenix, discovered a profound communications gap between the pay system’s project managers (who operated out of the department of Public Services) and the other federal departments being asked to join the Phoenix pay system.

Gartner noted the testing of the Phoenix system by project managers produced a 90 per cent pass rate while testing by their counterparts in the departments maintained it was closer to 50 per cent. “This is resulting in departmental stakeholders having a low level of confidence in the Phoenix system quality,” Gartner concluded.

This was just days before launch.

Goss Gilroy consultants, who examined the pre-launch testing of the Phoenix system, confirmed Gartner’s assessment. Just weeks before Phoenix went live for 120,000 federal employees, tests of the system revealed four types of defects, including some related to poor integration between old and new pay systems. “There were a large number of defects that were major,” the consultants reported, “and a large number of defects that had no planned fix date.”

The pay advisers, who had been given “workarounds” for each of the problems, knew very well the risks being assumed. Some voiced their concerns to management.

Yet the project went ahead anyway. Gartner concluded the top bureaucrats were anxious to get going. “Although most departments expressed uncertainty as to whether the (Phoenix) system has been thoroughly tested,” the Gartner report noted, “most also felt it will be substantially correct. Almost all expressed the opinion they are as prepared as possible for the go-live.”

Gartner did not address the question of appropriate skills in a bureaucracy where only a small minority have experience managing large information technology projects. This may have been a factor in downplaying risks.

Budgets were also an issue. The government was anxious to book savings for the pay system as a whole. Automating the process was expected to save $70 million plus annually. Hundreds of pay advisers across government were trimmed starting in 2014 to make way for what was expected to be a superefficient operation at the centralized pay centre in Miramichi.

Above all, the Phoenix project highlights the danger of a topdown approach typical of large government projects.

“An important driver for the (Phoenix) project management team was meeting established launch dates,” the report by Goss Gilroy noted, “hence any feedback that might slow down or delay progress was unwelcome and resisted.”

During nearly two years of crisis management has anything changed? Let’s hope Ferguson offers some useful insight into that this Tuesday. jbagnall@postmedia.com