Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

Optimisation Event

The conversion event an ad set or campaign is told to optimise delivery towards, such as a purchase, lead or install.
Paid MediaconceptUpdated

Definition

On automated ad platforms the advertiser no longer chooses each placement or bid. It chooses an outcome, and the delivery system predicts which people in the audience are most likely to produce it. For conversion campaigns that outcome is a named event: a purchase, a completed lead form, an install, a registration, or a custom event the advertiser defines and sends from its site or app.

The Meta campaign objectives guide describes the layering: the objective names the outcome at campaign level, and beneath it each ad set carries an optimisation goal and, for conversion objectives, the specific event. TikTok ad groups and paid app campaigns work the same way, with different names for the same setting. An event that the pixel or SDK has never sent cannot be chosen.

In practice: choosing the event and its volume

The first rule is to choose the deepest event that fires often enough. Deep events, close to revenue, describe what the business wants; shallow ones, earlier in the journey, are more frequent but further from the goal. The app install campaigns guide puts the same rule in app terms: install optimisation finds people likely to install, including many who never open the app, while in-app event optimisation finds people likely to do what the app is for, at a higher cost per install and a lower cost per useful user.

The second rule is to write the choice down. The paid social campaign build sheet records the event each ad set optimises for beside its attribution setting, so a reviewer can see before launch whether the event is too rare for the budget.

Shallow versus deep events and the learning phase

Delivery learns from optimisation events, and Meta's documentation ties the exit from the learning phase to a count of those events within a short window after the last significant edit. The learning phase guide explains the consequence: a purchase-optimised ad set on a modest budget may stay in learning for the whole window or never leave it, with unstable delivery and a high cost per result throughout.

The usual compromise is to move one step up the funnel, to initiate checkout or add to basket, accept that the system will find people who do that rather than people who buy, and move back down to purchase once volume supports it. Changing the event is a significant edit that restarts learning, so the switch is made deliberately, not weekly. The optimisation event test sets out how to decide between two events properly: identical ad sets that differ only in the event, both judged on cost per purchase rather than on the cheaper event each was told to find.

Checking the event fires correctly

An optimisation event is only as good as its tracking. If the event fires twice on some pages, fires on page load rather than on the action, or misses server-side purchases, the system learns from a distorted signal and optimises towards it. Testing the event in the platform's event tools before launch, and comparing event counts with orders or leads in the business's own records, catches most of these faults.

Illustrative example — not a real result

A wool shop moves its event up the funnel and back

Wexcombe Wool Shop spent £70 a day on one Meta ad set optimised for purchases and recorded about 6 purchases a week. After 3 weeks the ad set was still in learning, with a cost per purchase of £58 against an average order of £41. The team switched the event to add to basket, which fired about 48 times a week; the ad set left learning within 8 days, and although purchases were not the event, they rose to 11 a week at £38 each. By October, with a seasonal range and a larger budget, purchases reached 40 a week and the team moved the event back to purchase, accepting a new learning period, after which cost per purchase settled at £31.

The Meta campaign structure hub connects the event choice to budget level and consolidation, since fewer ad sets each receive more events.

Learning Phase is the period in which delivery is still working out who completes the event. Bid Strategy sets how much the system may pay for each event it finds. Tracking Pixel and its server-side counterparts report the event, so their faults become the system's. Key Event is the analytics-side label for an important action, which may or may not be the event an ad set optimises for.

Common confusions

  • The objective and the optimisation event are the same. The objective is set on the campaign; the event is the precise action each ad set pursues beneath it.
  • A cheaper result on a shallow event is a win. Only if those actions turn into sales; judge both events on the business outcome.
  • Optimising for purchases is always right for a shop. Not if purchases are too rare for the ad set to leave learning.
  • The event can be changed freely. Changing it restarts learning, so it is a planned move.
  • The platform's event count is the truth. Duplicate or missing fires distort both reporting and delivery.

More terms to look up

Where this term is used

Sources

  1. About the learning phase — Meta Business Help Centre

Related