Delivery

Learning phase

Updated 20 August 2026

The learning phase is the period after a Meta ad set is created or significantly edited during which the delivery system is still testing audiences and placements, and Meta documents roughly 50 optimisation events within 7 days as the point an ad set leaves it.

Key facts about Learning phase

Also known as
Learning, Learning limited
Category
Delivery
Last updated
20 August 2026

In AdOps

How does Learning phase work inside AdOps?

AdOps does not read learning-phase status; it is not one of the 45 rule metrics. Two rule-builder controls limit how often a rule can disturb delivery: the per-task action frequency, with 11 cooldowns from 15 minutes to Once in a lifetime, and Hours since creation (`created_time`).

The AdOps metric picker open over a condition row: a panel with Meta Ads and Custom metrics tabs, a search box, and a scrolling list of metrics with Spend selected. Sample data
Forty-five Meta metrics and your own sheet columns in the same list, each carrying a reporting period and any of six operators. Read the details

The learning phase is the window after an ad set is created or significantly edited while Meta’s delivery system is still working out who to show the ads to. AdOps cannot see that status, because learning phase is not one of its 45 metrics, but two of its controls exist precisely to stop a rule from resetting it repeatedly.

Why do automated budget changes matter here?

Meta restarts learning after a significant edit to an ad set, and it documents roughly 50 optimisation events in 7 days as the threshold for leaving learning. A rule that raises a budget every 15 minutes can keep an ad set in testing indefinitely, which is a delivery problem no threshold on ROAS will reveal.

How does AdOps limit how often it changes a budget?

With action frequency, set per task rather than per rule. Increase budget, Decrease budget and Set budget each accept one of 11 cooldowns from 15 minutes to Once in a lifetime. The shipped Tiered Budget Scaling template caps every one of its three budget steps at once every 2 hours.

How do you keep a rule off a brand-new campaign?

Add a condition on Hours since creation, which reports the hours elapsed since the campaign was created and is point-in-time, so it takes no reporting period. Requiring Hours since creation above 48 alongside a spend floor of Rp 1,000,000 stops a rule from judging a campaign before it has data.

Glossary

The terms an advertiser usually reads in the same sitting.

Turn this metric into a rule that watches it.

Pick the metric, pick one of the 11 reporting periods, set the threshold, and choose what happens when it is crossed. Nothing runs on your campaigns until you set the rule live.

Contact us See pricing

14-day trial · no card

Inside the product

What the screens actually look like

Nine screens from the working dashboard — the rule builder, the metric picker, the dayparting grid and the log that records what happened. Scroll the strip.

  • The AdOps Performance Dashboard showing a Purchase ROAS card at 2.380x and an Aggregated ROAS card at 2.088x, both badged Profitable, a Monthly Budget card at Rp 155jt, and a Spend Breakdown ranking the top five ad accounts against a budget utilisation bar at 45.5 per cent.
    Performance Dashboard. Two ROAS figures to three decimals, a budget meter, and spend ranked by ad account — over Today, Last 7 days, Last 30 days or This month.
  • The AdOps rule list showing twelve automation rules, each with an on/off toggle, the ad accounts it manages, and when it last triggered — some minutes ago, others on a dated timestamp.
    Rule List. Every rule, what it manages and when it last fired. The toggle is the only thing between draft and live.
  • The AdOps condition builder showing a task with time-of-day conditions across seven day tags, a nested AND group holding a lifetime spend condition under 400,000, and a second task with four stacked metric conditions.
    Conditions. Metric, period, operator, value — joined with AND or OR, and nestable, so a rule can say something a dropdown cannot.
  • The AdOps metric picker open over a condition row: a panel with Meta Ads and Custom metrics tabs, a search box, and a scrolling list of metrics with Spend selected.
    Metric picker. Forty-five Meta metrics and your own sheet columns in the same list, each carrying a reporting period and any of six operators.
  • The AdOps dayparting timetable: a grid of hours against the seven days of the week, with the daytime hours filled navy for every day and the night hours and weekend evenings left empty.
    Dayparting grid. Or draw the hours instead. Anything outside the shape you fill in simply does not run.
  • An AdOps execution log detail: one campaign, two action panels badged Not Executed, each listing the action parameters and every condition evaluated with its actual value, its expected value and a Pass or Fail badge.
    Rule Log detail. Why a rule did nothing is recorded as carefully as why it did something — actual against expected, condition by condition.
  • The AdOps activity log listing budget increases and campaign renames, each row naming the affected campaign by id, the rule that caused it and how long ago it happened.
    Activity Log. One row per change AdOps made in your account, naming the campaign and the rule responsible.
  • The AdOps ad account list: eighteen Meta ad accounts with on/off toggles, account ids, Active or Inactive badges and this month’s spend in rupiah.
    Ad Accounts. Accounts discovered from Meta arrive switched off. Nothing is read, and nothing is changed, until you turn one on.
  • The AdOps custom metric editor mapping a Google Spreadsheet: a spreadsheet id, a sheet name, a column to match campaigns on and a column holding the values.
    Custom metric. Point AdOps at a sheet, name the matching column and the value column, and your own number joins the metric list.

Every figure is rebuilt from the product’s own interface and filled with invented data — no customer name, ad account or spend figure appears anywhere on this site. See how a run works