In a separation, the date is a contract. Five calls decide the rest.
The sector produced four structural separations in 19 months, and each one is an HR, finance and payroll program with a legal close date attached before anybody scoped it. I ran the HR and payroll separation of AMC Networks off Cablevision: $10 million, five workstreams, two years, live on time with nobody missing a paycheck. Walk the five calls, find your seat, then run the two-minute check.
Five calls from a separation that landed, from the chair each was made in
AMC Networks, off Cablevision: Workday HCM and payroll stood up standalone, six legacy systems retired against the date, first ordinary pay cycle clean. The evidence beside each call comes from the sector's own SEC filings, which name these failure modes in writing. Pick a decision.
Week one of the AMC separation. The date came out of an agreement between two corporations, which meant it behaved like a contractual obligation, not a project milestone. Nobody was renegotiating it, and everyone understood that from the start.
Plan every workstream backward from an immovable date, and make the scope the variable instead. What does not fit gets deliberately deferred, with a documented fallback, not silently carried.
A fixed date changes the job from schedule management to scope triage, and teams that treat it as a normal milestone burn months discovering that. The sector's current wave proves the pattern: one spun-off company disclosed in its own filing that it began incurring systems build costs in the second quarter of 2025 for a January 2026 distribution. Roughly seven months of pre-close build against a hard legal date, and it still carried a services tail of up to two years afterward. Ask any separation CIO when their build started relative to close. The honest answers are all short.
Which scope items are deliberately deferred past Day One, and where is each one's documented fallback?
Your counterparty is the company you are leaving, and it is reorganizing itself while you ask it for data, access, and support. Goodwill is real. It is also finite, and it decays as the date approaches and their people head for the door.
Build a matrix, application by application and department by department: what access continues past the date and until when, what data gets pulled before the cutoff and by whom, and who at the parent answers afterward. Secured early, in writing, with a name on every line.
The matrix produced things no status meeting would have found. HR kept four read-only administrative users on the parent's legacy system for five months past the split, and that end date was not ours to set: it was bounded by the parent's own license. Recruiting had to extract candidate records before access closed, because afterward the data was simply gone. Payroll carried an open question the whole way on whether historical check data could convert at all, with a documented fallback if it could not. None of that is technology work. All of it decided whether the program landed.
What have we secured from the parent in writing, and what are we still holding on goodwill?
Every separation plan celebrates Day One. But a separating company's own risk disclosure names two disruption windows: at separation, and again at transition services expiry, when the parent stops providing and everything deferred has to run on your own platform with no fallback behind it.
Treat the services exit as a second go-live with its own plan, its own parallel payroll cycles, and its own named team. Resourced now, not discovered later.
A team that spends everything getting to Day One arrives at the services exit exhausted, with the hardest deferred items still open and the safety net withdrawn. I have seen the second window treated as an administrative formality more than once, and it never is. The dependency math will not let you sequence around it either: a typical transition exit carries more than 1,500 interdependent design decisions, which is why ERP, payroll, service management and identity land as one program. Two cutovers means two parallel run cycles, and the second one is the one nobody staffs.
Where in the plan is the cutover at services exit, and who exactly is resourced for it?
The transition services agreement everyone quotes in steering is the master agreement. The constraints that decide your sequence live in the Schedules, service by service, written by the party you are leaving.
Read the Schedules before the plan, and negotiate the exit mechanics per service: end dates, step-ups, and discontinuance, line by line.
In one executed media agreement, the term runs until the end of the Services, meaning duration lives in the Schedules and the first expiry is the real deadline. Discontinuing a service takes 30 days notice, unless wind-down would reasonably take the seller longer, as the seller determines, so the seller controls your exit tempo. Invoices arrive quarterly in arrears with 30-day terms, so cash lags service by up to five months and the burn is invisible during the quarter when overspend gets established. Extension step-ups run 15% to 25%, capped at 50% of the initial term, and negotiating them per service rather than agreement-wide is the win most buyers never know was on the table. Liability is capped near fees received, so the seller's exposure on a botched payroll cutover is roughly what you paid them. Your protection is program discipline, not the contract.
Which service expires first in the Schedules, and is the extension step-up negotiated per service?
Media runs two payroll estates by design. Corporate staff on Workday, SAP or ADP. Cast and crew with specialist production payroll vendors who carry the guild rate tables, the residuals engines, and the loan-out machinery. Somebody on every program proposes unifying them.
Hold the boundary. Corporate HCM is the system of record for corporate people. The interface to production is a GL feed and a cost-object mapping, never a shared worker record.
The guild rate layer is not configuration a commercial payroll team can absorb. A pension contribution of 9.75% instead of 11.25% depending on which season of the show it is. A health rate that moves $0.56 or $0.86 per hour based on whether the employer has crossed $15 million in its own historical residual contributions. Triple time after 15 elapsed hours and double time after 12 worked hours, two different clocks in the same agreement, with rest-period penalties accruing in tenths of an hour. Those tables also do not sit still: the rate calendar is already legislated through 2030. A program that pulls crew into corporate HCM inherits all of it, and fails on it.
What exactly crosses the boundary between corporate and production payroll, and who owns that interface?
What those five mean for the chair you sit in
Separation steering committees are crowded and pages get forwarded. 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 cutovers and an expiring platform
The second disruption window, at services exit, is in the filings and rarely in the plan. Meanwhile the platform decision has a clock inside it: SAP ECC mainstream maintenance ends 2027-12-31, so a NewCo that lifts and shifts ECC is buying a system with a known expiry. The sector's best public precedent is a sequencing one: one major media company ran five HCM go-lives in five years and consolidated Workday first in every deal, deferring historical data past Day One and backfilling later.
The program plan contains exactly one cutover, and the TSA exit appears as a line item rather than a phase.
Where is the second cutover in the plan, and who is resourced for it?
The costs that live outside the program budget
One-time separation costs run roughly 1% to 5% of the divested business's revenue, reaching 13% in complex carve-outs. Then the aftermath: 51% of divesting companies saw profitability drop more than 3.3 points after year one, and most take close to three years to recover EBITDA margin. Your program gets judged inside that window. The dual-run cost has a filing-grade name, duplicated information technology infrastructure, and transition billing arrives quarterly in arrears, so the burn is invisible exactly when overspend gets established.
The business case carries no dual-run line and no stranded-cost estimate for the remaining business.
What would twelve more months of the three services we are most likely to extend cost, per service?
A rate calendar that runs to 2030, and new custody duties
The guild reset is already legislated: writer health contributions step from 13% to 16.25% and then 16.75%, contribution caps climb from $250,000 toward $400,000 across the term, the performer pension merger targets 2028-01-01 with a conditional 1% increase behind it, and the crafts agreement expires 2027-07-31, likely mid-stabilization. Newer still: AI consent records and digital replica assets carry protections that survive a transfer of ownership, and a separation is exactly that. California now also wants quarterly loan-out reporting, first period ended 2026-03-31.
Guild rates were loaded once as static tables, with no owner for the change calendar.
Who controls the digital replica and consent records on Day One, and does the TSA cover their custody?
The separation timeline the filings describe
Decision three, drawn out. The published pattern from the sector's own disclosures: a compressed pre-close build, a Day One everyone plans for, a services tail with the meter running, and a second cutover most programs meet unstaffed.
The two media agreements on the public record ran 13 months and up to two years. The deeper the shared-services entanglement, the longer the tail, and a network group carved out of a parent's payroll sits at the deep end.
Four items already on the calendar for 2026 through 2028
Guild rate calendars are legislated years ahead, which is the one planning advantage this sector offers. Use it. Every date below is a payroll configuration event with a name attached.
Motion picture payroll services companies file quarterly with EDD, first period ended 2026-03-31, late after the last day of the following month.
The next negotiation means the next rate tables, work rules, and AI consent provisions. A go-live near it changes payroll configuration mid-stabilization.
Extended maintenance runs to 2030 for a fee. A NewCo that lifts and shifts ECC is buying a platform with a published expiry date.
Target date, with a 1% contribution increase on 2028-07-01 conditional on the merger completing. A benefits deadline with a condition attached, so configure both branches.
Five questions I would ask your separation program this week
Answerable from memory, scored on this page, nothing captured and nothing emailed. If you are mid-separation, you will know every answer. That is rather the point.
These five are the start of the instrument. A full review also covers the deferred-scope register, the historical data conversion questions, replica and consent record custody, and the stranded-cost picture. Or skip the tooling and book the program review.
Standing up a company against a date nobody can move?
Pre-close, mid-transition, or staring at a services exit with deferred items still open. I sell no software and staff no builds, and in a separation the date is set by contract anyway. My value is telling you early what will not make it, while there is still time to act on it.