What the Andromeda update changed for local lead gen (and why your $25/day Instant Form campaigns feel broken)

If your Meta ads were generating leads at a reasonable cost six months ago and now the same budget produces half the volume at twice the price, the campaign probably didn't get worse. The auction did.

Meta's Andromeda update was a backend overhaul to how the ad delivery system ranks and prices inventory. It wasn't announced with a press release aimed at small advertisers. Most contractors running $25 or $30 a day never heard about it at all. They just noticed their numbers got harder to hit.

Here is what actually changed, who it hit hardest, and what it means if you are evaluating whether to run Meta ads again after a previous agency burned you.

What Andromeda actually changed

Before Andromeda, Meta's auction ranked ads primarily on a combination of bid, estimated action rate, and ad quality. The system was relatively transparent in that direction: better creative plus a reasonable bid plus a receptive audience got you efficient delivery.

Andromeda shifted the ranking model toward a much more complex, ML-heavy system that weights predicted downstream value, not just immediate click or form-fill probability. In plain terms, the algorithm now tries to predict which users are likely to become real customers for a business, not just which users are likely to tap a form. It's attempting to optimize past the lead and toward the job.

For national e-commerce advertisers with large conversion datasets and diverse audiences, this is a genuine improvement. For a local painter running ads to a 20-mile radius with 40 form fills in his account history, it's a problem. The model doesn't have enough local conversion signal to make confident predictions. So it either over-spends chasing high-confidence users, or it under-delivers because it can't find enough of them in a constrained geography.

The contractors hit hardest are running campaigns that look exactly like the ones most home service agencies set up by default: $25 to $30 per day, Instant Form lead objective, metro-area radius targeting, broad or interest-based audience. That setup worked reasonably well under the old auction logic. Under Andromeda, it is starving the algorithm of the feedback it needs to optimize.

Why Instant Forms specifically got squeezed

Instant Forms (Meta's native lead capture, where the user fills out a form without leaving the app) have always had a lead quality problem: friction is low, so form fills are easy to get, but the leads who fill them out are often not seriously interested. Andromeda made this worse by deprioritizing accounts where the conversion signal is shallow.

A form fill is a weak signal. A form fill followed by a call, a booked appointment, a tracked job, is a strong signal. The new system rewards advertisers who feed it strong downstream signals through the Meta pixel, conversion API, or CRM integration. Advertisers who only fire a "lead" event when someone submits a form are giving the algorithm the weakest possible data to work with.

Most agencies running standard home service campaigns are not setting up conversion API, pixel purchase events, or CRM-to-Meta feedback loops. They set up the campaign, point it at an Instant Form, and report on cost-per-lead. That number can look fine on a spreadsheet while the underlying delivery degrades. The leads get cheaper but worse, or the volume drops and the agency blames the market.

This is one of the more common patterns in lost deals: a contractor pays an agency for months, gets leads that don't answer or are geographically useless, and by the time they stop the engagement they've spent a thousand dollars or more with nothing to show. The campaign looked like it was running. The conversion signal it was building was garbage.

What the $25/day budget constraint actually does in a competitive metro

Budget is not just a spend cap. In Meta's auction, daily budget signals to the algorithm how many auctions you want to participate in and how aggressively to bid. At $25 per day in a competitive metro, the system is being asked to find home service leads in a dense market at a constrained spend rate.

Post-Andromeda, the system needs more signal faster to learn who to target. Low budgets in tight geographies extend the learning phase or prevent the campaign from ever exiting it. A campaign stuck in learning phase is not optimizing. It is still guessing.

The Safe Step campaign ASN ran, which produced 247 leads at $11 cost-per-lead on $2,800 total spend, was structured to give the algorithm what it needed: enough budget to move through learning, conversion events tied to real downstream intent, and creative that matched the trade. That combination is what produces an $11 CPL in a niche where most agencies are seeing $40 or more. The mechanism isn't magic. It's giving the algorithm the inputs it requires to make accurate predictions.

At $25/day, you are often not giving it those inputs. You are getting the algorithm's best guess with half the data it needs, in a market where your competitors may be feeding it more.

What this means if you are thinking about trying Meta ads again

If a previous agency ran your ads and the leads were poor or expensive, there is a reasonable chance the campaign structure was part of why. Not the only reason, but a real one. Generic setup, Instant Forms with no downstream tracking, budget too low to exit learning in a competitive area, no niche-matched proof that the approach worked for a business like yours.

The questions worth asking before signing anything are concrete ones. Does the agency track anything past the form fill? What conversion events are they passing back to Meta? What does their average CPL look like for a business in your trade and your market size? Do they have a case study from something close to your niche, or are they showing you a roofing result when you do epoxy flooring?

Generic proof actively backfires at this stage of evaluation. One sales call reviewed for ASN's research showed a prospect reject a case study from a different industry entirely, calling it irrelevant. He was right. The numbers from someone else's trade in someone else's market tell you almost nothing about what your campaign will do.

What a well-structured campaign looks like after Andromeda

ASN builds campaigns inside the client's own Meta Ads Manager, sets up conversion tracking that goes past the form fill, and uses professional ad creative with unlimited revisions until it fits the trade and the brand. The AI follow-up system, Remi, texts new leads within seconds and holds a real conversation, which means lead responses that feed back into the system carry more signal weight.

The offer is $400 a month, no setup fee, no contract. If it works, you keep going because it's working. If it doesn't, you leave without a fight. That structure exists because a well-built campaign should be able to prove itself before asking for a locked-in commitment. If an agency needs three months and $1,500 upfront before you can evaluate whether it's working, that's not confidence in their system. It's the opposite.

If you want to see how ASN structures campaigns for your specific trade before making any decision, reach out through the contact page at americanservicenetwork.com.