Five decisions set the outcome. Most get made by accident.
On a $25 million multi-hospital Workday program, the calls that decided the result were not software calls. They were made in rooms where the software was working fine, by people with partial visibility. Walk the five below, find your seat, then run the two-minute check on your own program.
Five calls from one program, from the chair each was made in
Workday across HCM, payroll, finance and supply chain, a separate workforce management track, two integrators. I was hired directly by the client and owned it from selection through hypercare. Pick a decision.
Second round of the RFP. Clinical scheduling depth and ERP depth almost never live in the same firm, and the shortlist proved it: the ERP houses handwaved nurse scheduling, the workforce firms handwaved the general ledger.
Two integrators, and one named person on our side owning the interface between them. Not an owner on each side. One owner of the thing in the middle.
Terminations and off-cycle changes crossing between the HR system and timekeeping in the wrong order is a published root cause in health system stabilization work: pay rule errors, wrong accruals, overpayments. Around month seven a reasonable batch-timing change on one side pushed termination events past the timekeeping cutoff on the other. One person could see both halves. Neither status report could.
Who owns the interface, by name? Not who owns each side of it.
Steering wanted one date. One weekend, everything on. A single date is easier to communicate, easier to staff, easier to defend to a board.
Big bang, and I would do it again. But HR went live first, and finance and supply chain landed about a month behind it.
Managers learning to approve time and the same managers learning to approve a requisition are two different adoption curves. Land both on one Monday and the command center spends week one triaging confusion instead of defects. By the second wave the time questions had burned off and the tickets were mostly real. The cost was a few weeks of dual running, at the documented 20% to 40% premium parallel operation carries.
If you are arguing for a big bang in a hospital, are you also arguing for the gap inside it?
Two partners already reporting status on their own work. A third firm whose only deliverable is an unwelcome opinion is a hard line item to defend, until it earns its fee.
Fund independent quality and risk assurance alongside both integrators, reporting to the steering committee rather than to the program.
Testing tracked green on both partner reports and the counts were honest. The assurance read looked at coverage instead: the pay combinations one nurse hits in one week, a night differential on a charge premium on callback against a contract minimum, had been tested one at a time and never together. The date moved. KLAS found the same thing from the buyer side in July 2026: firms that push back on a poor decision score higher than firms seen as accommodating.
Who in your governance is paid to tell you no, and who do they report to?
Kickoff. Several bargaining units, each contract on its own cycle, a plan running past two years. Nobody had put the two calendars on the same page.
Every contract expiration date onto the program plan in month one, next to the go-live date. Then build and test the retro scenario before anyone needed it.
Five or six bargaining units across a 24 to 36 month program means at least one contract settles inside the window. That is a calendar, not a risk assessment. Retro pulls fringe and benefit fund contributions back through the period, and people who left before ratification may still be owed, so terminated records have to be reachable and correct. The alternative is that settlement landing two weeks after go-live, on a payroll system nobody trusts yet, in front of a union that just finished negotiating.
Lay your CBA expiration dates next to the go-live date. It takes an hour.
A design session redrawing cost centers, worktags, pay component groupings. Nine capable people. Not one from reimbursement.
Ask who owns the Medicare cost report Worksheet S-3 mapping. Then require paid hours by cost center to reconcile before and after cutover.
The wage index is built from Worksheet S-3 Parts II and III: paid hours and salaries by cost center, adjusted for reclassifications on Worksheet A-6. Your general ledger structure decides those inputs, and CMS works from data roughly four years old. Cutover, cost report, audit, payment year: by the time the damage lands, the integrator demobilized three years earlier and nobody connects the two events. I have asked this question in every health system design phase since. I have never once got a confident answer on the first pass.
Who owns the Worksheet S-3 mapping, and how do we prove paid hours reconcile across cutover?
What those five mean for the chair you sit in
Programs get forwarded around a buying committee. This is the screen to send: each seat's sharpest exposure, the early sign it is live, and the one question worth asking this quarter.
Two programs, one bench
Epic sits in 43.7% of acute care hospitals, and the dozen people who can explain charge capture or the float pool budget are named on the EHR work and the ERP work at once. Meanwhile Workforce Central on premises dies March 31, 2027, a reimplementation rather than an upgrade, and Epic began shipping its own ERP modules in June 2026 with no SI channel.
Two partner status reports, and no line item anywhere for the interface between them.
Who owns the seam between HCM and workforce management, by name?
Money leaking on a delay fuse
A cutover redraws the cost centers feeding Worksheet S-3, and the payment year it damages arrives roughly four years later. Nearer term: health systems carry 1,200 or more GPO and local agreements, with under 45% of spend in the ERP item master and under 65% on a valid active contract. The module lands on that gap. It does not close it.
Nobody can state today's contract-linked spend percentage. Unmeasured before the build means measured by the go-live.
Who owns the Worksheet S-3 mapping, and what is our contract-linked spend today?
Nurse pay meets a union and a regulator
One nurse, one week, a dozen earnings: differentials at 5% to 20%, on call at $2 to $5 an hour, callback against a four-hour minimum. Nearly all belong in the overtime regular rate, and DOL Fact Sheet 54 calls the omission a common healthcare error, doubled by liquidated damages. A late CBA settlement then pulls retro, fringe and benefit funds back through the same period. Since January 2026, per-shift staffing evidence is also a Joint Commission artifact that must survive migration.
Hypercare exit written in weeks instead of completed payroll cycles that exercise a full nursing rotation.
How many parallel payroll cycles are planned, and do they cover a full self-scheduling rotation?
The slowest, most expensive defect in the program
Decision five, drawn out. Paid hours by cost center feed the Medicare wage index on a four-year lag, so a mapping broken at cutover bills you in a payment year nobody connects to it.
Four items already on the calendar for 2026 and 2027
None of these appear on a program schedule, and every one of them lands on the same organization that is trying to go live. Verify each against your own plan before the next steering meeting.
Per-shift, per-unit documentation that staffing matched acuity. The evidence comes out of scheduling and timekeeping, and it has to survive migration with its history.
A reimbursement deadline that appears in no ERP deck anywhere, run by people who are usually not in your design sessions.
Engineering already stopped in December 2025. Moving off it is a reimplementation, not an upgrade, and the accreditation evidence has to come with you.
Credentialing and cost accounting mid-2027, supply chain and financials by end of 2027. Epic staffs its own implementations, so there is no SI channel for any of it.
Five of the ten questions I ask in week one
Answerable from memory, scored on this page, nothing captured and nothing emailed. These five are the ones that predict the go-live better than the plan does.
These five are half the instrument. The full ten-question diagnostic goes deeper on EHR concurrency, the item master baseline, your workforce management path, hypercare exit and the worker model, and it is just as ungated. Or skip the tooling and book the program review.
Which decision are you sitting in right now?
Pre-SOW, mid-build, or stabilizing after a rough go-live. I sell no software and staff no builds, so the questions on this page get asked out loud. Tell me where the program is and I will tell you what I see.