How the change history and the 3-day cooldown work

Last updated July 19, 2026

An audit that forgets what you did yesterday will happily tell you to do it again. The change history is how the optimizer remembers: a timeline of what already happened on the account, plus a cooldown that keeps recommendations off anything you touched recently.

What the change history feed is

Every time an audit runs, uplads pulls the ad account's activity feed from Meta for the last 14 days and turns it into a plain timeline. It covers every budget, status, bid, targeting and creative edit on the account — no matter who made it:

edits you made by hand in Meta Ads Manager,
changes applied from uplads (recommendations, bulk edits, launches),
changes made by your automated rules.

Each entry reads as a single line: how long ago it happened, what changed (Budget 100 → 80 EUR, Paused, Activated), which campaign, ad set or ad it touched, and who or what did it.

Where to find it

Open the Optimize view for an ad account and run or load an audit. Below the recommendations you'll see a quiet Show change history (N) button. Click it to expand the timeline; the header reads last 14 days · manual edits and applied recommendations.

The first 10 entries are shown, with Show all to see the rest and Hide to collapse it again. The section only appears when the audit actually found changes, and it sits with the full recommendation list — so if you've filtered down to a single category, switch back to the all-recommendations view to see it.

What gets filtered out

The raw Meta feed is noisy, so uplads trims it before you ever see it:

Audience sync events are dropped. A shop syncing purchasers every few minutes would otherwise bury everything else.
Post-review duplicates are dropped — Meta logs the same status edit twice, once as your action and once as the version queued behind ad review.
Billing and permission events are dropped; they aren't account edits.
Near-duplicates are collapsed: the same result on the same entity within the same hour is kept once.
The list is capped, and if more changes existed than fit, the audit knows it saw a partial picture.

Budget amounts come out of Meta in minor units, so they're converted back to normal currency before display. Status changes are read from Meta's numeric run-status code rather than its localised text, so a German-language account still reads Paused and Activated.

The 3-day cooldown

Any campaign, ad set or ad with a budget, status or bid change in the last 3 days goes into cooldown. While it's in cooldown, the audit will not recommend touching it.

The reason is discipline, not caution for its own sake. A budget you just cut is still settling — Meta's learning phase hasn't finished re-optimising delivery, so the numbers you'd judge it on aren't real yet. And an ad you deliberately paused yesterday was a decision, not an oversight; the tool has no business overturning it the next morning. It also makes recommend → apply → recommend loops impossible.

Edits made by your automated rules count exactly the same as edits you made by hand. If a rule paused an ad set last night, that ad set is in cooldown this morning.

Which recommendations the cooldown suppresses

Cooldown only silences the categories where a fresh change matters:

Pause candidates — ads, ad sets and campaigns that would otherwise be flagged as underperformers.
Budget changes — scale-up and scale-down recommendations on ad sets and campaigns.
Reactivate winners — an ad you just paused won't be recommended back on.

Everything else still runs normally: creative fatigue, funnel leaks, tracking and signal-health issues, geo and audience findings, landing drop-off, placement and device waste, and trend shifts. Those don't depend on a recent budget move settling.

Suppression is silent — a cooled-down entity simply doesn't produce a recommendation this run. It becomes eligible again once the change is more than three days old.

Outcome tracking — "Since the last audit"

Applied recommendations get graded. When an audit runs, uplads loads the recommendations you applied in the previous audit for that account and re-measures the same entities against the fresh data. The result is the Since the last audit card near the top of the results.

Each line shows the recommendation, its before-and-after numbers, and a verdict:

Improved — the primary metric moved more than 15% in the right direction (ROAS up, or CPA down).
Held — the metric stayed within ±15%.
Declined — it moved more than 15% the wrong way.
Done — a pause where the entity is no longer running; the saving is realised.
No data yet — the entity had no spend in this window, so there's nothing to compare.

This only appears from the second audit onward, and only if you applied something in the previous one. There is no prior audit to grade on your first run, and nothing to measure if you dismissed everything. The before-numbers come from the moment the recommendation was made and the after-numbers from the current window, so read it as direction of travel rather than a controlled test.