How long does an ERP freeze window last?

Longer than a single cutover. A global brewer migrating dozens of ERPs country by country stacks freeze after freeze, each with a hypercare tail. With SAP ECC mainstream support ending in 2027 and paid extended maintenance only to 2030, most large brewers face years, not weekends.

Why does an S/4HANA migration freeze non-ERP systems like loyalty or delivery?

Because they reconcile to the ledger. Loyalty points are liabilities, returnable deposits are cash, and delivery confirmations trigger invoicing. Their data must tie out to the financial core, so the change board freezes them while that core is being rebuilt and revalidated.

Can you avoid the ERP freeze window without skipping the upgrade?

You cannot skip the 2027 deadline, but you can shrink the blast radius. Route commerce, loyalty and delivery through an event bus instead of point-to-point ERP interfaces, so a migration touches adapters, not every downstream system, and each layer keeps its own release cadence.

Frozen by your own upgrade: the ERP freeze window

A multi-year ERP migration freezes far more than the ERP. Program offices impose a change freeze on every system that touches the financial core — order capture, stock, loyalty, returnables, delivery — because each interface adds regression risk to the cutover. Those systems inherit the ERP's calendar, so your route-to-market roadmap stops moving for the length of the migration, not the length of a weekend. That is the ERP freeze window.

What exactly is an ERP freeze window?

A change freeze is an ordinary program control. During a cutover, teams stop unrelated changes so that when something breaks they can isolate the cause instead of debugging a moving target. The trouble is scope. On a modern estate the ERP is not an island: it is the transactional and financial core that order-to-cash, pricing, master data, stock, loyalty accruals and returnable deposits all post against. To protect the cutover, the freeze expands to cover everything holding a data contract with the ERP. Your commerce and loyalty layer does not get frozen because it is risky. It gets frozen because it is adjacent.

Why does freezing the ERP freeze everything downstream?

Because dependency runs downhill. Loyalty points are a liability, returnable deposits are cash, and delivery confirmations trigger invoicing — every one of those reconciles to the ledger. You cannot let the numbers that feed the general ledger drift while the general ledger itself is being rebuilt and revalidated field by field. So the change advisory board draws the freeze boundary around the whole route-to-market stack, not because those teams are careless but because their data has to tie out to a target that is, for the duration, unstable by design. The decision is locally rational and globally expensive: to protect the correctness of the books, the program switches off the systems that actually grow the business. Nobody signs a memo saying “stop innovating in trade”; they sign a memo protecting the cutover, and the effect is the same.

How long does the window really last?

Longer than the plan admits, and the calendar is not optional. SAP mainstream maintenance for the ECC generation ends in 2027, with paid extended maintenance running only to 2030, which is pushing the entire installed base toward S/4HANA on a fixed clock. For a global brewer that clock does not buy a single weekend of downtime; it buys a program. A top-3 global brewer told investors in October 2025 that its estate still runs

43 different ERPs, non-harmonised process, non-harmonised data.

and that harmonising them underpins a raised cost-savings commitment of €400 million to €500 million a year. Forty-three systems do not migrate at once; they go country by country, each with its own cutover, freeze and hypercare tail. Stacked end to end, the commerce layer's effective freeze is measured in years, not weekends — and as of November 2025 most large brewers are early in that sequence, not through it. The window you feel is the union of every local freeze, not any single one of them.

What actually breaks while the estate is frozen?

Not the systems — they keep running. What breaks is throughput, and the damage compounds quietly rather than failing loudly.

What degrades	How the freeze causes it
Queue time explodes	Every non-trivial change routes through a migration change board, so features sit waiting instead of shipping — the queue, not the work, becomes the bottleneck.
Coordination tax	Each exception needs sign-off across program, finance and integration, turning a one-line change into a standing meeting that costs more than the change itself.
Context decay	Multi-year programs churn people; by the time the freeze lifts, the reasons behind half the backlog have evaporated, exactly as they do when context is handed off through decks instead of working code.
Backlog cliff	Everything deferred lands at once after go-live, colliding with the least stable moment of the whole program.

None of these appears as an outage, which is why they rarely reach a steering committee. They show up as a roadmap that quietly slipped two years and a trade team that stopped bothering to file requests it knew would be refused.

Who is really paying for the freeze?

The growth engine pays, so that the ledger can stay clean. That is the inversion worth naming. The ERP is a system of record; your order capture, loyalty and delivery layer is a system of differentiation. Freezing the thing that differentiates you in order to protect the thing that merely records you is backwards — and it is invisible on the program plan, because a frozen roadmap has no line item. The ideas did not dry up. The window to ship them closed. The cost surfaces later, as share that moved while you were heads-down in reconciliation, and it never gets attributed to the migration that caused it.

The way out is not to skip the migration — the 2027 deadline makes that a fantasy — but to stop letting the ledger's calendar govern the commerce estate. When the route-to-market suite talks to the ERP through an event bus rather than point-to-point interfaces, a migration changes a handful of adapters instead of every downstream system, and the SAP financial core can be rebuilt underneath a commerce layer that keeps its own release cadence. That is the practical case for putting the estate on one event bus.

With 2027 and 2030 fixed on the calendar, this wave of freezes is only starting; over the next few years nearly every large brewer will be mid-migration on something. The ones who treat the ERP as a swappable financial core — not the spine of everything attached to it — will keep shipping loyalty, delivery and pricing changes straight through their own cutovers. The rest will spend those years waiting for a window to reopen, and calling the wait a project.