Blog What Meta's Andromeda update actually changed for local...

What Meta's Andromeda update actually changed for local home service lead gen campaigns

Most contractors who tried Meta ads and got burned didn't get burned because Meta stopped working. They got burned because the agency running their ads didn't know what changed, or didn't care enough to adjust. Meta's Andromeda update is a clean example of that gap.

Andromeda was Meta's infrastructure overhaul of its ad ranking and retrieval system, rolled out in stages through 2024 and into 2025. The short version: Meta moved from a single monolithic ranking model to a retrieval-and-ranking architecture, where candidate ads are first filtered through a retrieval layer before they reach the deeper ranking auction. That shift has real consequences for small, locally targeted campaigns, and most agencies running cookie-cutter home service ads either missed it or ignored it.

What the retrieval layer actually means for a local campaign

Before Andromeda, Meta's ad system evaluated a large pool of ads against audience signals and ranked them in a single pass. After Andromeda, ads have to survive a retrieval stage first, where the system filters for relevance before anything gets to the ranking auction. An ad that doesn't pass the retrieval filter never competes in the auction at all, regardless of the bid.

For a national brand with broad targeting, this is a minor calibration. For a local contractor running ads to a radius of 20 or 30 miles, the retrieval layer is a real obstacle. Small audiences with tight geographic constraints give Meta less signal to work with. If the ad creative and the campaign structure don't clearly signal relevance to that local context, the retrieval layer deprioritizes the campaign before it ever gets a fair shot at delivery.

The practical result: campaigns that look fine on the surface, with a live status and a modest spend ticking upward, can be getting throttled at the retrieval stage. Low impression volume relative to budget, poor reach within a defined radius, CPMs that seem high for the audience size, these are symptoms. Most contractors see the spend and assume the ads ran. They may have, just not where or how often they should have.

Why most agencies missed this, and why it hurt their clients

Andromeda wasn't announced with a contractor-facing press release. Meta published technical notes and the changes filtered through performance data before most practitioners documented them clearly. An agency managing dozens of accounts across different industries has a choice: track platform-level changes and adjust, or run the same playbook until a client cancels.

The second path is more common than most agencies would admit. A standardized setup, a templated creative, a broad interest stack layered over a geographic radius. That approach worked reasonably well before the retrieval architecture tightened. After Andromeda, campaigns that rely on generic creative and loose audience signals are more likely to stall in the retrieval stage, because they're not giving the system enough to read as locally relevant.

Contractors who paid $1,000 or more upfront, signed a three-month contract, and watched their campaign produce weak impressions and unqualified leads, many of them were dealing with exactly this. The agency's structure wasn't updated. The creative wasn't niche-specific. The campaign looked active but the retrieval layer was working against it from the start.

What a post-Andromeda local campaign actually needs

The retrieval stage responds to relevance signals, which means the ad itself has to do more work than it used to. A few things that matter now more than they did before:

Creative specificity is a ranking input. An ad that says "we do roofing" reads as generic. An ad that references the specific type of work, the service area, and a concrete outcome gives the system more signal to match against local intent. This is one reason the Safe Step campaign, 247 leads at $11 per lead on $2,800 in spend, produced the numbers it did. The creative was matched to the trade, the offer was specific, and the campaign gave Meta's system clear relevance signals to work with.

Instant Forms or landing pages need to match the ad. Post-Andromeda, Meta's system takes conversion signal feedback more seriously than it used to. If the ad drives clicks that don't convert, the retrieval layer learns from that. A form that's slow, generic, or mismatched to the ad's promise degrades future retrieval performance. The form has to convert at a reasonable rate for the campaign to stay healthy.

Follow-up speed feeds back into the system indirectly. A lead that books a job closes the loop on the conversion signal Meta is looking for. Slow follow-up means leads go cold, jobs don't close, and the campaign's downstream conversion data weakens. This is part of why instant AI follow-up through a tool like Remi, which responds to a new lead in seconds and books them onto a calendar, isn't just a sales tactic. It's also a campaign health input. The faster real conversions happen, the stronger the signal Meta's system receives that the campaign is delivering relevance.

What to look for when evaluating whether an agency has actually adapted

The tell isn't what an agency promises. It's how they describe their creative process and their campaign structure for local markets specifically.

If an agency can't explain how they handle retrieval-stage relevance for tight geographic targeting, they probably haven't thought about it. If their case studies are generic ROAS numbers with no trade-specific context, that's a sign their approach is standardized, not adapted. Niche-matched proof matters here. A result from rubber resurfacing tells an epoxy flooring contractor something real. A result from a vague "home services" campaign tells them nothing.

The other tell is what happens if it doesn't work. An agency that requires a three-month contract and a large upfront payment has no structural incentive to adapt quickly. An agency running no-contract, no-setup-fee engagements carries the risk themselves. If the campaign stalls in retrieval and leads don't show up, that agency loses a client immediately. That structure forces them to stay current, because they have no lock-in to fall back on.

What this means for your next step

If you're evaluating whether to try Meta ads again, ask the agency you're talking to how their local campaign structure accounts for the retrieval layer. Ask to see a case study from a trade that's close to yours, with actual CPL numbers. And look at what they're asking for before you've seen a single lead.

ASN runs done-for-you Meta campaigns for home service contractors with no setup fee and no contract. If the leads don't show, you walk. The terms are on the table before anything else is. If you want to see what a niche-matched campaign looks like for your trade, the contact page is the right place to start.

ASN manages Meta ads for home service contractors with no setup fee and no contract. If you want to see what this looks like for your trade before committing to anything, the contact page is the right next step.

See how it works for your business