Engagement Custom Audiences: Retargeting Social Interactions

Social engagement can give a business a retargeting audience before it has substantial website traffic. Choose an interaction your account can actually use, distinguish audience unions from intersections, and test completed business outcomes. A save, comment or deep video view is a clue to investigate, not proof that someone will buy.

Inventory the sources you can save

Read the action labels and retention bounds in the current audience creation flow, then verify the saved definition. Permissions, source availability and campaign delivery settings can differ. The Meta CustomAudience SDK distinguishes engagement, website and video subtypes and exposes approximate size/status fields; it does not justify a universal menu of actions or a fixed maximum window for every source. Do not publish a campaign whose intended rule is only written in the brief.

OR expands a pool; AND requires both actions

Including video viewers plus profile visitors commonly describes a union: viewers OR visitors. It is not evidence that every included person did both. Only call the audience an intersection if your actual supported setup can represent and enforce both conditions. Do not try to reconstruct individual social identities outside an authorized mechanism.

Hypothetical set example: 800 unique video viewers, 500 unique profile visitors, and an assumed overlap of 200 yield 800 + 500 − 200 = 1,100 people in the union, versus 200 in the intersection. If overlap is unknown, the union lies between 800 and 1,300; the intersection between zero and 500. Raw engagement events cannot supply that overlap. An approximate Meta size range is not an exact identity ledger.

Interpret depth against runtime and context

In an illustrative comparison, 75% of a 12-second clip is nine seconds, whereas 75% of a 120-second demo is 90 seconds. These are different exposures. A short view can still contain the complete offer; a long view can be curiosity or criticism. Saves belong in one consistently named “saved-content” segment, without an invented high-intent ranking. DMs may be support requests or complaints, so distinguish purchase questions in authorized business records without uploading sensitive conversation content.

• Deduplicate within each source where the reporting contract supports it. Do not add video views, profile visits and saves and label the total unique people.
• If tiered delivery is required, check supported exclusions and assign an overlap priority. Otherwise use a combined pool and state that sequence and individual order of exposure are not guaranteed.
• Avoid dividing a small pool into many thresholds. Preserve a simple comparison with enough business outcomes to answer the question.

A completed social-to-sale test brief

Hypothetical plan for a reusable bottle: compare an eligible saved-content audience with an eligible broader engagement audience, using the same fit/size demonstration and landing page. Proposed message: “See the bottle dimensions and lid-care instructions before choosing your color.” Hypothesis: the saved-content group may respond better to practical detail. Check actual saved inclusion/exclusion and expansion settings; use an eligible randomized audience test when available, or label the result descriptive if overlap cannot be controlled.

Illustrative assumptions, not observed performance: each retained order contributes $18 after variable costs and expected refunds, and $3 is reserved for other costs/surplus; media CPA ceiling is $15. Approve a $150 allowance per arm and a common reporting age. Attributed outcomes of 12 and ten retained orders would produce CPAs of $12.50 and $15, respectively. They would not demonstrate a platform-wide advantage or statistically reliable result. Reconcile pending conversions and refunds before the decision; include any separate creative production cost in the full acquisition cost.

Record decisions without inventing an alert

Prepare the approved interaction brief, budget ceiling and creative in a NovaStorm campaign draft, then review the saved fields before publication. Keep the dated source/status and business-result worksheet with the campaign owner. This workflow does not assume automatic discovery of profitable engagement segments. Compare with website retargeting using the same commercial outcome, rather than declaring one source inherently superior.