Est. reading time: 7 minutes

The hardest habit for experienced Meta buyers to break is the one that used to be the job. Audience micromanagement, the stacked interests, the lookalike ladders, the ten-ad-set matrix, now mostly fights the delivery system instead of steering it. Meta’s automation decides who sees what, and the advertiser’s leverage has moved to the inputs, meaning the conversion signal, the creative supply, and a handful of settings that constrain the machine where it needs constraining. This is the playbook for running that system full-funnel, from cold scroll to checkout, without surrendering ROAS discipline.

Know what you’re actually operating in 2026

Two structural changes reframe everything below. First, Advantage+ Shopping Campaigns were renamed Advantage+ sales campaigns in early 2025 and now cover leads and app installs alongside ecommerce, so older documentation referencing “ASC” describes the same machine under its old label. Second, as of early 2026 Ads Manager no longer asks you to choose between manual and Advantage+ at creation. There’s one unified flow with the AI features on by default and individually toggleable, which means Advantage+ isn’t a campaign type you opt into anymore. It’s the water, and your decisions are about which automations to keep on and what to feed them. Underneath it all runs Andromeda, the retrieval system deciding which ads even enter an auction, and it rewards exactly two things, signal quality and creative variety.

Map the funnel before the machine does

Automation consolidates execution, not strategy, so intent tiers still need defining with clarity. Cold means no brand interaction, warm means engagement or a site visit, hot means cart or checkout behavior, and each tier gets one job and one success metric, hook and CTR for cold, session quality for warm, purchase and ROAS for hot, so nothing gets judged on a metric it wasn’t built to move.

Then assign campaign roles. An Advantage+ sales campaign carries net-new revenue at scale, combining prospecting and retargeting in one structure with the delivery system routing between them. A second, smaller sales campaign handles warm and hot remarketing through catalog ads where you want message control the automated campaign won’t give you. Lead gen and app objectives run in their own flows so budgets don’t cannibalize. The count stays low on purpose, since fragmenting conversion volume across parallel campaigns competing for the same people is the old structure’s worst habit imported into the new system.

The measurement spine

Everything Advantage+ does well, it does because of the signal you feed it. Pixel plus Conversions API with event deduplication, value, currency, and content_ids passed consistently on every purchase, and Event Match Quality treated as a KPI, since hashed emails and phone numbers on your events are the cheapest CPA reduction available on the platform, the input-quality argument we made in Training Meta’s AI Faster: Smart Inputs, Better Results. One correction to older playbooks, including earlier versions of this one: the Aggregated Event Measurement step of ranking eight prioritized events is gone, retired when Meta made that aggregation automatic in mid-2025, so if your launch checklist still includes event prioritization, delete the step.

Standardize attribution and hold it still, 7-day click with the 1-day engaged and view columns understood for what they are, and reserve lift tests and holdouts for the question the dashboard can’t answer, which is incrementality. The full signal-loss operating environment, and why all of this became non-negotiable, is covered in adapting your Meta strategy for post-iOS signal loss.

Audiences: suggest, cap, and exclude, in that order

Cold prospecting now runs on breadth plus first-party guidance. With Advantage+ audience, your customer lists, pixel data, and catalog function as suggestions the system starts from and expands beyond, not fences it stays inside, so the job is seeding better signals rather than shrinking reach. Placements stay automated for the same reason. Every restriction you add is exploration you’ve paid to prevent.

The one control that outranks every other setting is the existing customer budget cap. Uncapped, the algorithm follows conversion probability straight to your easiest wins, which are the people who already buy from you, and the campaign quietly becomes an expensive retargeting program wearing a prospecting badge. Reported ROAS looks great, new customer acquisition stalls, and the dashboard congratulates you for it. Upload your customer file, define it as the existing-customer audience, and cap the spend, with 20 to 30 percent as the working starting range, adjusted against your new-versus-returning split over the following weeks. Where genuine acquisition focus matters, add the new-customer value boost so the system prefers first-time buyers without overpaying for them. Note that this cap has had a turbulent product history, removed and then reintroduced with improvements, so verify where it currently lives in your account’s interface rather than trusting a screenshot from last year.

Warm and hot audiences live mostly in the companion remarketing campaign. Custom audiences from 30-to-90-day engagers and site traffic, with exclusions keeping the prospecting side clean, and catalog ads retargeting product viewers, add-to-carts, and checkout starters on tight 3-to-14-day windows. Once volume supports it, move hot audiences to value optimization, and use the minimum ROAS bid control where available so the margin floor is enforced by the bid strategy instead of by your Thursday panic.

Creative is the targeting now

With audience levers surrendered, the ad itself selects the audience, which makes creative supply a structural input rather than a production chore. The floor is mobile-first assets that hook inside two seconds, brand early, show the product in motion, and carry captions by default in 9:16 and 1:1, and the accounts that win keep ten or more live creatives per campaign with genuinely distinct concepts, not one concept in five trims, refreshed on a rolling basis before fatigue sets in. Repurposed creator content earns its keep here, and the mechanics of moving it over without losing performance are in repurposing TikTok videos for Meta.

Match the message to the tier the system will route it to. Cold gets problem-solution demos, UGC, and category-framing angles that create demand. Warm gets proof, comparisons, and offer articulation that removes doubt. Hot gets urgency, risk reversal, and dynamic reminders carrying price, reviews, and inventory. As for Advantage+ creative enhancements, the auto-generated variations, treat them as a test, not a default, because the lift is real in some accounts and the off-brand variations are real in others, so enable them on a subset, watch what Meta actually serves, and decide on evidence.

Scale with rules the dashboard can’t talk you out of

The operating discipline is the same one that governs any learning system. Enough signal to learn, roughly 50 optimized conversions per week, no mid-flight edits while it’s learning, budget raises held to 10 to 20 percent with days between them, and performance judged on stabilized weeks rather than volatile days. Formal tests run through the A/B tool with decision rules declared before launch, sample size, window, and the success metric, whether that’s purchase ROAS, new-customer CAC, or blended MER, and then honored, because a test whose verdict gets relitigated by opinion was a very expensive coin flip.

And keep the scoreboard outside the platform. Blended MER against backend revenue and new-customer acquisition cost from your store’s own data are the numbers that can’t be flattered by attribution, and they’re what the existing-customer cap, the value boosts, and the whole full-funnel structure ultimately answer to. Advantage+ is an amplifier, and it amplifies whatever it’s given, elite signal and varied creative into compounding acquisition, or thin signal and three tired ads into expensive retargeting. Build the inputs, set the guardrails, and let the machine do the part it’s genuinely better at.