Meta Ads for DTC: What Can You Pay for an Extra Order?

Reviewed 4 November 2026. A shoe store sells a pair through a Reels ad, collects €90 and sees a healthy return on ad spend. Three weeks later the shoes come back. The ad still looks successful in the platform report. The next budget decision depends on what the store keeps after returns and variable costs.

What the October Andromeda update tells us

Meta describes Andromeda as a retrieval system. It selects a smaller set of eligible ads for subsequent ranking stages. On 28 October 2026, Meta described a new architecture that brings candidate selection and scoring together on a GPU. It still evaluates advertiser-defined targeting logic. Its published efficiency gain concerns computing work in retrieval. It gives you no forecast for your own ROAS. Meta Engineering on retrieval and the October technical report

The ad gives the system another signal. More importantly, it gives a prospective buyer information. A video showing the width of a shoe across the toes addresses a different objection from a discount banner. Those are two ideas you can test and explain.

Work out what an extra order is worth

Illustrative figures per additional order, excluding sales tax and fixed costs:

At €60 net revenue, the implied target ROAS is about 2.73 on that revenue basis. If your platform reports sales including tax and before refunds, its ratio cannot be compared directly. For customer acquisition, use the costs and revenue of new customers. Include repeat orders in the forecast only when you have evidence from actual cohorts. E-commerce unit economics

Plan the campaign and the creative together

For a single product, start with a sales campaign, a deliverable region and a verified purchase event. Advantage+ automation can manage audiences, placements and budget. Separate products or markets when margin, availability or a reserved test budget gives you a reason. A different campaign for every age group makes a low-volume account harder to evaluate.

Document the incumbent ad and the result that would make you replace it. A weekly budget of €220 at €22 per order would buy ten orders. Spread evenly across ten concepts, that would average one order per concept. Minor differences will tell you little. With automated delivery, a creative generating fewer sales may simply have received fewer impressions. Product feeds for shopping campaigns

For automatic placements, preview the crop, captions and text on the surfaces your account offers. Cheap clicks on one surface become informative when you can also examine carts, paid orders and returns.

From two event messages to one budget decision

Receipt. If the browser pixel and Conversions API both report a purchase, send the same event name and event ID for that order so Meta can deduplicate it. Reopen the confirmation page during testing. Verify that both routes respect the customer’s privacy choice. A server connection must not silently resume a transmission the customer declined. Meta’s deduplication documentation and first-party data in the store

Attribution. Record the attribution window, net revenue, share of new customers and whether the returns window is complete. A change in attribution can move reported cost per purchase while paid orders in the store stay flat.

Incremental sales. Meta offers modelled incremental attribution. A controlled lift test approaches the question by comparing randomly assigned treatment and control groups. Run one when your account is eligible and purchase volume supports it. A small effect with a wide uncertainty range calls for a cautious test of the next budget increase. Meta on incremental attribution

At the next budget review, put ad spend next to contribution from settled orders and your estimate of how many orders were additional. That gives you a way to distinguish customer acquisition from taking credit for purchases that were already going to happen.