Generally wait until the update has fully rolled out before you act, unless you find a concrete, unrelated break to fix. During a rollout, rankings fluctuate as Google works through the change, so what you see mid-rollout is noise, not a settled result. Reacting to that noise wastes effort, you may “fix” a drop that reverses on its own before the rollout ends, or chase a fluctuation that was never the real outcome. The pivot is wait out the rollout unless there is a concrete break. Resist the instinct to react fast to recover, because speed against an unfinished update is mostly speed against a moving, unreliable signal.

The reasoning is that an update in progress has not produced a final state to diagnose. Positions can swing day to day while Google applies and adjusts the change across the index, so a page might fall, recover, and resettle before the rollout completes. Any conclusion you draw mid-rollout is drawn from data that is still in motion, and any change you make is aimed at a target that has not stopped moving. Once the update finishes, you have a stable picture to compare against your pre-update baseline, and that is when diagnosis and action become worth doing.

The exception is real and worth keeping. If, while waiting, you discover a concrete problem unrelated to the update (a page that broke, a noindex tag left on by mistake, a technical regression), fix it, because that is not reacting to update noise, it is repairing a genuine fault that would cost you regardless. The distinction is between chasing fluctuations and fixing a clear, identifiable break.

For your next rollout, hold non-essential changes until it completes, keep recording what is happening so you have data to read once it settles, and act only on concrete, unrelated breaks in the meantime. One thing worth confirming: Google’s specific guidance on what to do during and after updates is revisited over time, so check its current recommendations before you finalize your response to a live rollout.