$2,500 a month. Sixty days. One service business where the owner answers the phone. Under that constraint, I would make Meta leads harder to submit, not easier.

The fastest way to lower the cost of a lead worth having is to make the wrong person drop out before submitting. I ran Google Ads for home services clients for years, and the accounts that held up were not the ones with the cheapest form fills. They counted booked jobs. A $12 lead that never picks up costs follow-up time, office time, and a dead slot on the calendar. A $45 lead that books is cheap.

This is a filtering playbook, not a scaling playbook. The ceiling is $2,500 in each calendar month, with roughly $82 a day available when ads run. I will not spend the first week’s unused budget just to use it up. The plan is to buy fewer, better leads, see which ones become jobs, and feed those outcomes back to Meta when the account can support it. By week four, judge the campaign on booked jobs. If you cannot tell which leads booked, stop and fix tracking before adding spend.

1. Days 1–7: Spend $0 to define a booked job

Cost against the cap: $0 in media. Buys: a measurement setup you can trust. Keep the first seven days paused. That sounds slow until you see the contractor math: Meta home-service leads run $10 to $50 but close at 3 to 8%, so a $20 lead at a 5% close rate costs $400 per booked job, while a $60 higher-intent lead at 35% costs $171. I used to tell clients to chase the $20 lead. I was wrong. The cheap lead was the expensive one.

Install Pixel plus Conversions API and connect each lead to an outcome in the CRM. Define booked job before launch: the same stage has to mean the same thing when the owner marks it on Tuesday and when someone else checks it on Friday. Record the lead’s source, whether anyone reached them, whether they booked, and whether the estimate happened. If the CRM is not ready, use a sheet. A sheet taped to the wall beats a polished dashboard full of unreturned calls.

For context, Home and Home Improvement on Meta averages $41.26 per lead at a 5.22% conversion rate. Treat that as context, not a target. If you optimize only to raw fills, Meta can find people willing to fill forms. It cannot infer that the owner booked a job unless you record and return that outcome.

Cut the reporting clutter now. Cost per fill can help diagnose an ad, but cost per booked job decides whether the spend earns its place. Once the tracking path works, you can turn the campaign on without spending the rest of the test arguing over what a lead means.

2. Days 8–21: Spend about $1,150 on one filtering campaign

Cost against the cap: about $1,150, at roughly $82 a day for 14 days. Buys: one campaign, no more than two ad sets, and a form built to reject poor fits. The paused week stays unspent; it does not become permission to exceed the monthly ceiling.

At this budget, I pick a higher-intent instant form over a landing page. Its review screen asks people to confirm their details before submitting, and it runs on mobile feeds. A standard form makes volume easier. The extra screen and taps on the higher-intent form make submission less automatic. Expect to pay more per fill. That is the filter doing its job, provided more of those fills turn into bookings. Do not declare victory because the form looks tougher; check the calls.

Keep it short. Use one or two multiple-choice qualifiers and no more than three questions total, a setup Meta notes averaged 26% lower CPL. For most trades, I want to know:

  • What service do you need?
  • Are you the homeowner?
  • When do you want it done?

Contractors I trust report the same uncomfortable pattern: short forms run $35 to $55, qualified forms run $80 to $100, but booked cost holds $150 to $250 either way. If a question does not help disqualify a renter, a tire-kicker, or someone outside your service area, cut it. Every extra question needs a reason beyond making the form look serious.

Draw one tight service area around where the trucks actually go, not where you hope they might go. A roofer working a 30-mile sprawl can burn the day’s budget on estimates at the edge of the map. I draw the real service area, then cut 10 minutes off the edge for Meta. If a zip code has cost you two no-shows, exclude it. That is a better use of a hard location control than pretending every inquiry inside a large circle is equally useful.

Skip interest stacking. Advantage+ audience is now the default for Leads and Sales, and interests and lookalikes are suggestions Meta can ignore. Location, minimum age, language, and exclusions are the controls to concentrate on. Run broad inside the radius, set a 25-plus or 35-plus floor depending on the job, and let the form and creative do the filtering. Ten tiny audiences at $82 a day do not create ten useful tests. They split the money until none of them tells you much.

The rule for this phase is simple: two ad sets maximum. Keep enough spend in each place to learn whether the leads answer and book.

3. Days 22–35: Spend about $1,150 testing two creative angles

Cost against the cap: about $1,150 across these 14 days. Buys: a comparison of messages, not a collection of audiences. This phase crosses from month one into month two. Keep the same roughly $82 daily pace and the $2,500 ceiling in each month; do not treat the phase total as a new monthly allowance.

Run two angles with three variations each. The first is before-and-after: cracked shingle close-up versus finished roof, dirty coil versus clean airflow reading, or flooded under-sink versus new install. Shoot it on a phone on a real job. The second is price anchor plus urgency: a typical range for your market, what the visit includes, and the next open slot. The offer stays recognizable while the opening image or first line changes.

That is where the testing budget goes. I cut interest targeting, lookalikes, catalog tricks, and five extra audiences to pay for it. Creative gives a homeowner a reason to stop; an audience label in a dashboard does not. Two angles beat ten audiences when the budget can barely teach one campaign at a time.

Kill by spend, not gut feel. Give each ad about twice your target cost per lead before judging it. Say a qualified form lead runs $60. At $120 with zero leads, pause the ad. With one lead that went nowhere, let it reach $180 before deciding. Keep one winner per angle and replace the loser with a new hook on the same angle: same offer, different first three seconds. Do not change the audience, offer, form, and ad at once, then announce that you found the winner.

Watch the booking log alongside the ad results. A message can attract plenty of submissions while setting the wrong expectation for the call. If you have spent $600 with no booked jobs, inspect the offer and the follow-up before debating the font. The point of this phase is to learn which promise gets a conversation that can become work.

4. Days 36–50: Spend about $1,230 and return the CRM outcomes

Cost against the cap: about $1,230 over 15 days. Buys: a clearer booking signal and a decision about whether Meta can use it. By now, mark every lead as booked, no answer, junk, or wrong area. Send outcomes back daily through Conversions API for CRM where the setup allows it, and use the same records to calculate your true cost per booked job.

The tempting move is to switch straight to the Conversion Leads goal, which can optimize toward a downstream CRM stage such as qualified or booked rather than raw form fills. But that path has limits tied to Instant Forms, roughly 200 leads a month, and 1 to 40% reaching the target stage within 28 days. This $2,500 plan will not produce 200 leads a month. I would not call an unsupported switch an optimization strategy. By week four, optimize your decisions to booked jobs; use a downstream Meta goal only if the account qualifies. Otherwise, keep returning the outcomes, judge ads against bookings yourself, and fix any broken tracking before putting in more money.

The other half of this phase costs labor, not media: answer the phone. Contact within 5 minutes versus 30 minutes is about 100 times more likely to connect and 21 times more likely to qualify, and ServiceTitan 2025 data shows two-minute responders convert at 62% versus 28% at the 42-minute average. I have seen the failure pattern too many times: 30 of 36 leads never answered the phone, and a separate advertiser could not reach about 30% of lead-form numbers at all.

If you cannot call within five minutes, do not spend $82 that day. Set one phone to buzz, call first, qualify second, and log the outcome that day. A good form cannot rescue a lead that sits unanswered while everyone assumes someone else has it.

5. Days 51–60: Spend up to $820 to hold or stop

Cost against the cap: up to $820 for the final ten days. Buys: a decision, not another round of experiments. Pull three numbers:

  1. Cost per booked job: total Meta spend divided by jobs booked from Meta, not form fills.
  2. Contact rate: the share of leads you reached within five minutes.
  3. Show rate: the share of booked estimates where you got in the door.

Say you finish with $4,100 spent, 75 fills, 22 contacts, and 12 bookings. That is about $55 per fill and $342 per booked job. If a booked job is worth $800 in gross margin, hold the campaign and prepare to scale when the budget allows it. If it is worth $300, stop and fix the offer. The lower cost per fill would not change that decision.

Do not sneak a budget increase into this last stretch. At roughly $82 a day, the plan already comes close to the month-two cap. I cut audience expansion, duplicate campaigns, and lookalikes to keep enough money behind one campaign and its winning ads. If bookings are holding, keep the daily pace. If they are not, stop. Adding spend cannot repair a missing CRM outcome, a phone nobody answers, or an offer that fails on the call.

6. Before you commit: Cut businesses this plan cannot serve

Cost against the cap: $0. Buys: an honest reason not to waste the $2,500. Skip this plan if no one can call within five minutes during work hours. Skip it if you need jobs this week to make payroll; you will pull the plug before the form and booking data mean much. Skip it if your service area is under 50,000 people, because $82 a day can exhaust a small radius in 30 days and frequency will do the rest.

Skip it, too, if you sell a $99 drain clear with no upsell. If a qualified lead costs $60 and one in four books, you paid $240 to sell $99. I cut low-ticket lead gen from my client roster for that reason. One home services team I point to fixed lead quality by rebuilding tracking around booked jobs before scaling volume. The order applies here: tracking first, volume second.

This is not a challenge to make every service fit Meta. It is a constrained way to find out whether this service, with this follow-up, can buy booked work at a cost the business can carry.

7. Keep the work together; hand it off when the cap lifts

Cost against the current cap: no added management fee. Buys: a plan the owner can actually run. Every week, someone must compare spend with booked jobs, pause a losing ad at its kill threshold, upload booked and junk outcomes, and listen to two sales calls to hear why estimates did not close. Allow three hours a week for that work, plus the five-minute phone response. If nobody will do it, pause rather than pay for leads and blame the ads later.

When you can pay for management without taking it out of this media plan, hand off the whole chain. Do not hand off the ads while keeping follow-up and booking outcomes invisible. groas runs Meta Ads for a flat $1,499 a month, with audiences, budgets, and ads worked around the clock and new ads tested as winners emerge. One human owns the outcome, actions are logged with reasoning, and you keep the account. You set the booked-job goal and the budget. The execution does not wait for your Friday. That fee is not hidden inside this $2,500 media cap; adding it is a decision for when the constraint lifts.

When it does, do not add ten audiences. Keep the filter and raise the daily budget in 20% steps: from $82 to about $98 for three to four days, then about $118 if cost per booked job holds within 15%. If it rises more than 20% after a bump, roll back and wait a week. Once booked-job cost holds for three weeks, test a simple landing page against the higher-intent form for one service, expand the radius by one ring, and add a second angle built from real job photos.

Most agencies will not tell you this: the boring version wins. Fewer leads, harder forms, faster calls, and one number that matters. That is how $2,500 stops buying form fills and starts buying jobs.