Manual Sales Campaigns

What you are actually managing in a Manual Sales Campaign is the same conversion-optimization engine Advantage+ Shopping Campaigns use, purchase-event bidding against catalog or landing-page data, but wired to ad sets you build yourself: an optimization goal, a conversion location, an attribution setting such as 1-day click or 7-day click, a bid strategy like cost cap, bid cap, or minimum ROAS, and either detailed targeting, a custom audience, or a lookalike as the audience.

One structural change worth knowing if you learned this platform a few years ago: Meta removed the ability to exclude by detailed targeting, interest or behavior categories, at the ad set level in March 2025. The old move of excluding an interest from a prospecting ad set to keep it from overlapping with a lookalike ad set no longer works that way. Exclusions now have to be built from custom audiences instead, an uploaded customer list, a website or app custom audience, or an engagement custom audience. If an old playbook or agency document still references interest exclusions, that instruction is stale.

Placement and bid strategy are the other reason to run manual instead of an automated campaign: you can drop specific placements, excluding Audience Network for brand safety or Reels if the creative is not native to that format, and you can set a hard cost ceiling or a minimum ROAS floor per ad set rather than letting the campaign chase whatever looks cheapest in aggregate. Advantage+ audience expansion and detailed-targeting expansion sit as checkboxes inside an otherwise manual ad set; leaving expansion on lets Meta broaden delivery past what you selected, which quietly defeats a genuinely locked audience test if you forget to switch it off.

Conversion location is worth setting deliberately as well: website, app, website and app combined, offline, or Messenger and Instagram conversations each pull from a different signal source, and offline conversions in particular need a separate offline event set uploaded on its own schedule rather than firing in real time, which changes how quickly the ad set can actually react to what is converting.

The failure mode beyond over-segmentation is budget math: an ad set needs a real number of purchase events per week to leave the learning phase and produce a stable cost per result, and splitting one modest total budget across many narrow ad sets to test more usually means every single one is starved, none of them stabilizes, and the reported costs bounce around meaninglessly. Frequent edits compound the problem, since changing budget, audience, or creative resets the learning phase, so a manual structure that gets tweaked every few days never gets the chance to settle.

Reporting lives at the ad set level here, which is the whole point compared to an automated campaign: compare cost per purchase or ROAS across audiences directly, and use the placement and platform breakdown to catch one bad placement dragging an otherwise fine ad set's average down. When comparing ad sets that use different attribution windows, one on 7-day click, another migrated in on 1-day click, normalize which window you are reading before concluding one audience actually outperformed the other, since a longer click window alone will credit more conversions.