The freeze window before certification.

Stabilise before you measure

Certification needs a fixed point

The certified count is a snapshot. It records what you have deployed and can evidence at a point near the end of the term, and that snapshot becomes your permanent entitlement. The problem is that a live estate never stands still. Servers are decommissioned, workloads migrate, features are toggled, and environments are rebuilt, often weekly. If the measurement is taken across a moving estate, the number drifts, the evidence stops matching the configuration, and the count you eventually sign is harder to defend because it never described a single settled state.

A freeze window fixes the point. By controlling change across the in scope estate for a defined period, you give the measurement something stable to describe. The snapshot then matches the evidence, the evidence matches the configuration, and the certified number reflects a real estate as it actually stood, not an average of several estates blurred together. This is the difference between a count that survives a later audit and one that unravels the moment someone asks what the environment looked like on the day.

What is a freeze window before ULA certification?

It is a managed period, usually six to eight weeks before the term ends, in which deployment changes to the licensed Oracle estate are paused or tightly controlled so the count can be captured and evidenced cleanly. It does not mean the business stops. It means changes that would move the count, such as decommissioning in scope servers, migrating databases, or enabling and disabling options, are held, scheduled around the measurement, or recorded with care so the snapshot stays coherent.

The freeze sits at a specific place in the sequence. Planned deployment growth, the genuine demand brought forward into the window, happens earlier in the term while there is time to deploy and evidence it. The freeze comes after that, once the estate is at the level you intend to certify, and its job is to hold that level steady while the count is measured, reconciled, and prepared. Growth first, then freeze, then certify.

What the freeze protects

A well run freeze protects three things at once, and each of them is a place where an uncontrolled estate loses value or invites risk.

• The count. Decommissioning an in scope server during the measurement removes processors you were entitled to certify. A freeze prevents legitimate entitlement from being lost to routine housekeeping at the worst possible moment.
• The evidence. Rebuilds, migrations, and configuration changes break the link between what the evidence shows and what is actually running. A freeze keeps the records and the reality aligned.
• The defence. A certified count tied to a single, documented snapshot is straightforward to defend in a later audit. A count assembled from a constantly changing estate is far harder to stand behind.

A freeze sequence in miniature, indicative

The table below shows how the freeze sits within the run up to certification for an anonymized example. The timings are indicative and exist to show the sequence, not to represent any real engagement, and the right length depends on the size and complexity of the estate.

Indicative only. The length of each phase depends entirely on the estate and the wording of the specific agreement. The principle is fixed even where the timings move: stabilise, freeze, measure, certify.

How to run a freeze without stopping the business

A freeze that blocks all change is unrealistic and unnecessary. The goal is control, not paralysis. Changes that do not move the in scope count proceed as normal. Changes that do are scheduled, deferred, or carefully recorded so the snapshot stays coherent.

• Define the scope of the freeze. Identify exactly which systems are in scope for the ULA, because only changes to those affect the count and need control.
• Set a change protocol. Route any change that would alter in scope processors, options, or environments through a single approval point during the window.
• Hold decommissioning. Defer retiring in scope servers until after the snapshot, so entitlement is not lost to routine clean up.
• Capture the snapshot. Measure the estate once it is stable, and record the configuration, dates, and tool output that evidence each figure.
• Lift the freeze after certification. Once the count is captured and reconciled, normal change resumes, now governed by the certified entitlement.

The next step

The freeze is the discipline that turns a maximized estate into a number you can sign and defend, and it works only when it follows planned growth and precedes the final measurement. Set the window, control change across the in scope estate, and capture a snapshot the evidence matches. To bring the right deployment into the window first, see planning deployment growth before exit, capture every named product during the freeze with maximizing across database options and packs, and ground the approach in our ULA deployment maximization guide.