When Nobody Agrees on the Numbers

How to build a single source of truth when every system tells a different story

The WARN Act notification was due in 48 hours. Federal law requires employers to notify workers 60 days before a mass layoff affecting 100 or more employees. We needed to file the exact number of affected positions.

The CFO’s report said 285. HR’s system showed 310. The departmental spreadsheets, when we added them up, totaled 265.

We couldn’t file three different numbers. We couldn’t file an approximate number. We needed one number that would withstand legal scrutiny - and we had two days to get it.

This is the moment when data reconciliation stops being an academic exercise and becomes an operational emergency.

Why the Numbers Never Match

Every large organization has this problem. The specific numbers differ, but the pattern is universal.

Timing differences create the first gap. HR processed terminations on the last day of the month. Finance reflected them on the pay date. Departments marked them the moment someone stopped showing up. On any given date, the three systems showed different totals because they updated at different frequencies.

In our case, the HR system had already processed 8 position eliminations that wouldn’t show in Finance until the next pay run. Meanwhile, 19 positions showed as “active” in HR but had already been marked “eliminated” in departmental planning spreadsheets.

Definition differences create the second gap. Is a contractor an employee? HR said no - contractors aren’t in the HRIS. The departments said functionally yes - contractors were doing essential work and would be affected. For WARN Act purposes, the answer depended on specific legal tests about control and integration that neither system captured.

What about someone on leave? Technically employed, not actually working. Part-time workers? One employee, or 0.5 FTE? Grant-funded positions? Technically employed by the university, but funded by external sources with their own rules.

Scope differences create the third gap. The CFO’s number included only core operations. HR included all legal entities. Departments included everyone who reported to them, regardless of which entity technically employed them.

We had subsidiaries with separate EINs. We had affiliated entities with their own boards. We had jointly-employed positions where two entities each claimed the person. The “same organization” had different boundaries depending on who was measuring.

The 48-Hour Reconciliation

We couldn’t wait for perfect systems. We needed a defensible number in two days. Here’s what we did:

Hour 1-4: Establish the governing definition. Before touching any data, we met with legal counsel to understand exactly what WARN required. The law specifies “employees” under its own definition, which doesn’t match HR’s definition or Finance’s definition or the departments’ definition. We needed to apply WARN’s definition to our data, not vice versa.

Hour 5-12: Pull all sources. We extracted complete datasets from HR (active employee records), Finance (payroll records), and every department (their planning spreadsheets). We also pulled the legal entity structure, which would matter for determining which notifications to file.

Hour 13-24: Build the crosswalk. This was the tedious, essential work. For every position that appeared in any source, we documented: Where does this appear? What status does each source show? Why are they different?

Most of the 310 vs. 285 vs. 265 gap explained itself once we did this work:

• 12 positions were in HR but not Finance because they were new hires not yet on payroll
• 8 positions were in Finance but marked “terminated” in HR due to timing
• 28 positions were in departmental spreadsheets but weren’t in HR because they were contractors
• 9 positions appeared in multiple departmental spreadsheets (double-counting)

Hour 25-36: Apply the legal definition. With the crosswalk complete, we could apply WARN’s specific tests. The 28 contractors were evaluated individually - 10 met the “employee” definition under WARN, 18 didn’t. The joint-employment situations were resolved by legal determination. The timing issues were resolved by using a common “as-of” date.

Hour 37-44: Validate with stakeholders. We shared our number with each source owner, showing exactly how their number reconciled to ours. This wasn’t about agreement - it was about ensuring we hadn’t made errors in the crosswalk.

Hour 45-48: Document and file. We filed 289 positions. More importantly, we documented exactly how we got to that number, which reconciliation decisions we made, and why. When questions came later - and they did - we had a defensible audit trail.

The Broader Lesson

Most organizations never face a WARN filing. But every organization faces decisions that depend on accurate numbers, and accurate numbers require reconciliation work that no system does automatically.

The principles that saved us in 48 hours apply to ongoing data governance:

Establish the question before touching the data. “How many employees do we have?” is unanswerable without specifying: by what definition? As of what date? Including which entities? For what purpose? Different purposes require different answers.

Build crosswalks between systems. Every discrepancy has an explanation. Finding those explanations is tedious, unglamorous work - but it’s the only way to understand what your data actually represents.

Create a single source of truth for decisions. You don’t need one system that’s right about everything. You need clear agreement about which system is authoritative for which purposes. Payroll might be authoritative for compensation decisions. HR might be authoritative for headcount reporting. The departmental data might be authoritative for workload planning.

Invest in ongoing reconciliation. A one-time reconciliation degrades immediately. Systems diverge. New errors accumulate. Organizations with good data have regular reconciliation processes, clear ownership, and consequences for data errors.

The Hidden Cost of Data Chaos

We met our WARN deadline. But the 48-hour scramble revealed a deeper problem: the organization had been making workforce decisions for years based on numbers that didn’t agree with each other.

When the CFO said “we have 285 positions in this unit,” he believed it. When HR said “we have 310,” they believed it. Both were acting on their best understanding. Neither knew their number was incomplete.

How many decisions had been made on incomplete data? How many headcount targets were set using one definition and evaluated using another? How many budgets were built on numbers that didn’t match reality?

We’ll never know. But the WARN crisis created organizational will to fix the problem. Three months later, we had a single authoritative source for position data, with defined ownership, regular reconciliation, and clear documentation.

The crisis was expensive. The solution was worth it.

Data quality isn’t about technology. It’s about governance - clear ownership, regular reconciliation, and consequences for getting it wrong. The organizations that invest in this discipline make better decisions. The ones that don’t make decisions on numbers that might not be real.