Meta Purchase-Value Optimization: Revenue, Refunds and Margin

Purchase value and profit are different outcomes. Where eligible, a value-oriented delivery configuration can aim toward the reported purchase value. It does not know your full variable costs or automatically optimize accounting profit. Evaluate the data contract and commercial allowance before selecting the delivery hypothesis.

Verify VALUE in the actual account

The official Meta ad-set SDK includes VALUE. That name is not evidence of eligibility for your objective, dataset, bidding setup or account. Inspect the current available configuration and save its exact goal, currency and constraints for review. Do not infer access from a universal weekly purchase threshold.

Preserve the A/B counterexample

Illustrative per-purchase assumptions, not observed campaign results: A reports $150 revenue and contributes $60 before $50 media acquisition cost. B reports $220 revenue and contributes $55 before $60 media cost. The variable costs are therefore $90 for A and $165 for B, under this simplified definition.

The higher revenue ROAS in B coincides with a loss. Define variable costs explicitly: product, fulfillment, payment fees and other order-dependent expenses. Avoid subtracting a cost again if it is already included in contribution. The contribution-based cost ceiling makes the definition reviewable.

Add refunds before comparing a mature cohort

Hypothetical cohort extension: assume ten purchases in each option, one fully refunded A order and two fully refunded B orders. For simplicity, assume refunded orders have fully reversible variable costs except an unrecovered $10 per return. Media remains $500 for A and $600 for B.

A’s refund-adjusted revenue ROAS is 1,350 / 500 = 2.70; B’s is 1,760 / 600 ≈ 2.93. B still has the higher revenue ratio and worse contribution. Real returns may leave unrecoverable product or shipping costs; replace these assumptions with the ledger and a declared maturity cutoff.

A teaching Purchase-quality check

The Meta server-side custom-data model exposes value and currency. The examples below are a reconciliation worksheet, not runnable API payloads and not accepted events.

Also check discounts, tax and shipping treatment, occurrence time, cancellations and the distribution of values. A refund-adjusted finance report does not establish that an earlier platform Purchase event was automatically reversed. Verify any supported correction workflow separately. Use the event reconciliation guide for identity checks.

Compare mature margin under a spend limit

1. Predeclare the retained-order contribution metric, cost scope, refund maturity window, allocation method and cash allowance. Keep offer and underlying price changes separate from the goal change.
2. Where an appropriate controlled comparison is available, compare value-oriented and purchase-count treatments. Ordinary parallel campaigns can receive different traffic; describe that limitation.
3. Review paid unique orders, retained revenue, variable costs, contribution after media and cash exposure together. A few high-value orders can make a short report unstable.
4. Stop or retain an inconclusive result when the declared allowance is reached. Evaluate the next spending increase with marginal economics, rather than extending a favorable average ROAS indefinitely.

Prepare the comparison brief and proposed campaign as paused campaign and local ad-set drafts in NovaStorm. Verify required settings in the actual product and account before publication. This does not promise that NovaStorm supports or configures VALUE, connects the returns ledger, or proves a profitable lift.