Every other module in the suite can have a soft launch. Payroll cannot. One in five payrolls in the United States carries an error, and each fix costs about $291. Two payroll problems are enough to send roughly half of employees looking for another job. Payday still lands on the same date either way. That is the job.
This guide is written for the people who carry that risk: the VP of Payroll and the Payroll Director who run it, and the CHRO and CFO who answer for it. Payroll reports through HR in most organizations and pays through Finance, so it lives on a seam that neither side fully owns. It is the deep dive the HCM module guide hands off to. We walk the ten modules one by one: the key items, the takeaway, and what to watch. It is system agnostic, illustrated with Workday Payroll, Oracle Cloud Payroll, and ADP where a real example helps.
Payroll has no grace period. It is the one system judged on a single run in front of every employee. Own the seam between HR and Finance, prove it with a clean parallel, and let the first payday be boring.
1. What is really on the line
Payroll is judged on a single event. The run either pays everyone correctly and on time, or it does not, and everyone knows within the hour. The first live payroll is the real go-live, the same way the first close is the real go-live for finance.
The error data is sobering. An EY study of 508 U.S. organizations found the average company makes 15 corrections every pay period, that one in five payrolls contains an error, and that each error costs about $291 to fix. The two largest sources are the time feed and the absence feed. The cost runs well past operational. After two payroll problems, about 49% of employees say they would start looking for a new job, and about 24% would leave after one. Gartner still expects more than 70% of ERP initiatives to fall short of their goals by 2027, and poor data quality alone costs organizations about $12.9 million a year. Payroll is where those failures meet the workforce in person.
Name one accountable owner for the payroll seam between HR and Finance before design starts. Most payroll failures are unowned handoffs, not bad software.
2. Own the operating model before you configure anything
The first real decision has nothing to do with a setting. It is who operates the run and who owns the compliance work. Run it in house and you keep the staff, the configuration control, and the liability. Move to managed or outsourced payroll and a provider operates it for you, which trades some control for a specialist's compliance depth. The two are separable from the software choice. ADP GlobalView, for example, lets you process the payroll yourself or have ADP run it on the same platform.
Score the decision on jurisdiction count, internal payroll depth, and compliance risk, not on the sticker price of the software. Outsourcing moves the work, not the legal responsibility.
3. Ten modules, and the one decision that decides each
Each module has a handful of decisions that decide whether it works, and almost every one ends at a seam. Read the ten as a pay stub, top to bottom. The map first, the detail after.
Payroll Foundation
The pay group binds a population to a frequency, a calendar, and a legal entity, and everything downstream inherits it. Biweekly is the most common frequency in the U.S. at 43%, and several states set a minimum, so one national calendar will not keep a multi state workforce compliant. Workday models this with pay groups and period schedules; Oracle uses payroll definitions and consolidation groups.
Earnings and Deductions
The gross to net engine is the hardest part of the build. Every earning and deduction is a pay component with taxability flags, and accumulators build the bases the rest of the calc reads. One mis flagged code silently corrupts thousands of checks. Imputed income, like employer group term life over $50,000, is the classic trap: a non cash earning that still has to be taxed for Social Security and Medicare.
Tax
Multi state withholding is a location problem before it is a math problem. Reciprocity agreements and the convenience of the employer rule in Connecticut, Delaware, Nebraska, New York, and Pennsylvania can double tax a remote worker. Local taxes add another layer. Decide a native tax engine or a third party like Vertex on purpose, not by default.
Garnishments and Involuntary Deductions
Federal Title III caps ordinary garnishment at the lesser of 25% of disposable pay or the amount above 30 times the federal minimum wage. Child support runs to 50 to 65%, and where a state rule is more protective, the lower garnishment governs. There is no room for hand calculation, so configure the caps and order priority in the system.
Time to Payroll
The first of the two feeds that break payroll, and the single most common source of error. Every hour type, regular, overtime, shift differential, and holiday, needs a pay code mapped exactly to an earning, and time and payroll have to share one overtime and workweek rule set. Late approvals and retro edits arriving after the deadline are the usual failure.
Absence to Payroll
The second feed, and the second most common source of error. Leave has to reconcile on the balance and on the paid versus unpaid treatment, including the termination payout. Workday calculates the balance with the same engine as payroll and can pay it out at termination. Unpaid leave coded paid overstates the check.
Retro and Off Cycle
Retro pay recalculates prior periods for backdated raises and late time, and the engine tracks when the event is recognized separately from when it pays out. Off cycle runs bypass the stable controls and change funding and tax treatment. High off cycle and correction volume is a symptom, so track it and fix the feed upstream.
Banking and Payments
Direct deposit is near universal at 92.65% of workers. Pay cards cannot be the only option under Regulation E. Earned wage access is the part changing under you, and it is a payroll deduction and a multi state compliance obligation, not a perk. More on this below.
GL Posting and Costing
Every run produces the money movement and the accounting. Costing splits one person's pay across cost centers, projects, grants, and funds, and in Workday the allocation has to total exactly 100% or the run will not post. The payroll to GL map is where Finance and Payroll meet, and where the close slips.
Global Strategy and Year End
Global payroll is a portfolio: a native engine, a partner, or an aggregator, chosen country by country, with the HCM kept as the system of record. Year end is the annual test. W-2s go to the SSA by January 31, penalties land per return and again per employee statement, and multi state unemployment means a different taxable wage base in every state.
Freeze the earning and deduction catalog early. Every new code needs an owner, a taxability decision, and a test case before it enters the system.
Name one owner. On one program, finance and procurement data loads kept failing during conversion. The cause was not the software. No single person owned the validation rules across the modules, so every team assumed another had it. We named one owner, put a date on each item, and the failures stopped. The fix cost nothing.
4. Payments look solved until earned wage access arrives
Payments look solved because direct deposit is near universal. The parts that still need decisions are cards and earned wage access, or EWA. EWA lets a worker draw wages they have already earned before payday, repaid by payroll deduction. It is not small: about 10 million workers accessed roughly $31.9 billion this way in 2022, and about 8% of workers already receive it while 28% want it.
The regulation is a live 2026 issue. States are legislating fast, Nevada first in 2023 and roughly a dozen by mid 2025, and they split on the core question. California and Connecticut treat most EWA as credit, while a December 2025 federal advisory opinion held that covered, employer partnered EWA is not credit under the Truth in Lending Act. For payroll, an employer partnered EWA program is a deduction the interface handles every cycle and a compliance obligation in every state where employees work.
Decide your EWA position now. If you offer it, treat it as a payroll deduction and a multi state licensing question, and pick the employer partnered model to keep it out of credit territory.
Fifty on paper, ninety in reality. Discovery listed 50 integrations. The real number was closer to 90 once we mapped every handshake to banks, tax agencies, benefit carriers, and downstream apps. Payments and remittances are where the integration footprint hides.
5. The payroll-to-GL map translates between two different data models
The payroll to GL map is a translation between two data models. Payroll thinks in pay components and workers; the general ledger thinks in accounts, cost centers, and entities. When the definitions differ, a reconciliation can tie at the top line and be wrong underneath, which is a recurring source of close delays. Period end adds accruals, because a biweekly period rarely lines up with a monthly close, and a tax nuance trips people up: for federal employment tax the liability arises when wages are paid, not when the period ends.
Build the GL mapping and the per account reconciliation control with Finance in the room, not as a downstream export nobody owns. The control is simple: beginning liability plus additions minus payments equals expected ending.
6. Three delivery models, and most enterprises end up with two
Global payroll never resolves to one system. There are three delivery models, and most enterprises use all three at once. Match each country to the model on depth and risk, and keep the HCM as the single system of record.
The vendors sit in different places. Workday runs native payroll for the U.S., Canada, the UK, and France, and reaches more than 180 countries through certified partners under Global Payroll Connect. Oracle offers native cloud payroll for more than 46 countries. ADP runs GlobalView for 40 plus countries and adds the Celergo aggregator for about 140.
Keep the HCM as the single system of record, choose native, partner, or aggregator country by country, and plan year end before go live. Penalties land per form and per statement.
7. What AI actually does in payroll today
The honest win from AI in payroll today is catching a bad run before it pays. ADP Assist flags anomalies in payroll data before an error happens. SAP's Joule can already explain a pay statement to an employee. Workday has a Payroll Agent in early access, and Oracle has announced a Payroll Run Analyst Agent that still carries a future product disclaimer. Read the vendor speed claims as vendor reported.
Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027 on cost, unclear value, and weak controls. For payroll that caution matters more than most, because the output is money and compliance.
Buy anomaly detection now, and put every agent behind the same approval and audit you give a person who can move pay.
8. Prove it before you cut over
Payroll go-live is judged on one run, so proving it beforehand is the whole game. Parallel testing, running the new system beside the old and reconciling the results, is how you know it will land. The bar from the field is at least three mock runs, with under 2% error on the final one and Finance reconciling to the cent. Change management decides whether people use what you built: initiatives with strong change management are about seven times more likely to meet objectives, 88% versus 13%.
Two more things belong on the 2026 checklist. Pay transparency laws in a growing list of states now pull pay range and comp data into the compliance perimeter, including internal moves. And the goal is a boring first run, not a good one.
Do not go live without a clean parallel. Reconcile to the cent, three cycles, before you cut over.
One role stalled the test. An end to end testing cycle once stopped for a full week because one security role exposed compensation data far wider than intended. Every test script that touched pay had to pause. The build was fine. The design decision was not. Treat pay data security as a design task, not a go live task.
The red flags to run before every gate review
The vendors are capable and the implementers are skilled. What they cannot do is decide how your payroll should run, or own the seam between HR and Finance for you.
Payday does not move
A payroll program is a sequence of decisions that live on the seam between HR and Finance. Which operating model, which pay groups, which taxability flags, which interfaces reconciled, which delivery model per country, and proven with which parallel. Make those calls early, own the seam, and treat the first live run as the real go-live. Get it right and the best outcome is that nobody notices. That is the win.
Related: this is the payroll deep dive the HCM and HR modules guide hands off to. See also implementing finance ERP modules for the ledger this run posts into, and implementing supply chain ERP modules for the cutover discipline a payroll go-live depends on.