Do not copy last season’s cost per lead into next season’s agency brief. A cheap form fill tells you nothing about the job that followed: whether it was in your service area, whether anyone completed the work, or what was left after discounts and direct costs.
I managed home-services accounts long enough to learn this the slow way. Say you spend $20k a month on Meta. One agency brings you $20 leads; another brings you $32 leads. Most owners pick the $20 agency without asking what offer produced them. I would pay $32 for leads that book one in four times over $20 leads that book one in ten. The first batch costs $128 per booking. The supposedly cheap batch costs $200. And neither figure tells you whether the work gets done.
A cheap lead can hide an expensive offer
A $20 lead from a $29 tune-up special and a $45 lead from an emergency no-heat call are not the same product. One may fill the calendar with low-margin maintenance visits that need upselling on site. The other reaches someone standing in a cold house who needs work this week. Brief an agency on CPL alone, and you give it every incentive to sell the first kind. Meta can find plenty of people who want a discount. They click and fill forms faster than someone considering a $12,000 system replacement.
That is why a blended target is dangerous. Home-services Meta medians run around $63 for HVAC service and plumbing and about $117 for HVAC install and roofing, against a blended Meta average near $45.50. Tell an agency to beat $30 across those offers, and the pressure is obvious: shift spend toward whichever trade and offer produce the cheapest fills. You may get your lower CPL and a worse season at the same time.
Distance and job mix decide what a booking is worth
Cheap leads also travel badly. I have seen shops with a tight three-county service area let Meta run statewide because CPL looked better. People forty miles beyond a sensible drive time still fill forms, and the agency reports those fills at the same price. Then a tech spends ninety minutes in the van for a $189 drain clear, or dispatch cancels because the roof is two towns past where the crew will go. Out-of-area fills are a common cheap-lead trap, alongside missing phone numbers or ZIPs and no exclusion for rejected inquiries. The lead counted. The job never had a chance.
The numbers change again when you price booked work. At Facebook close rates of 8 to 25 percent, cost per booked job runs roughly $252 to $420 for HVAC service, $332 to $487 for plumbing, and $780 to $1,463 for roofing. That spread is why a roofer and a drain cleaner should not share a CPL target. One signed roof can justify expensive leads. Ten discounted tune-ups that never turn into replacement work cannot justify the same acquisition cost. If the brief does not name the services, ZIPs and dates worth buying demand for, Meta will chase whatever filled cheapest last week.
Judge the offer before you judge the agency. Comparing a replacement offer, an emergency call and a free inspection on creative alone misses the point. What the deck calls a creative test, I call an economics test with different pictures attached. Each offer needs its own qualification, follow-up and close rate. Otherwise, you keep arguing about CPL while the job mix decides your season.
Put completed-job economics ahead of CPL
I stopped briefing on CPL years ago after watching the same pattern repeat. Short forms at $35 to $55 a lead and qualified forms at $80 to $100 a lead both landed around $150 to $250 per booked appointment, because the cheaper leads booked at a fraction of the rate. When one in four appointments signs work in the $15k to $300k range, that works out to $600 to $1k in ad spend per signed job. The higher-priced lead can cost less by the time it becomes a job. A dashboard that stops at form fills cannot show you that.

Even advertising cost per completed job flatters the work if you stop there. A $400 drain clear at 50 percent margin that closes one in five caps break-even CPL at $40 before you pay anyone to run the ads. Account for the discount, parts, labor and management fee, and the ceiling drops again. I used to tell clients to judge an agency on ROAS. I was wrong. Judge it on contribution after direct job costs, discounts and full acquisition cost, with the fee in the calculation where everyone can see it. Otherwise, a provider can win the CPL contest while you lose the season.
Write next season’s brief backward from the work
Start with the jobs you want crews doing in May and June, then tell the agency what demand to buy. Name the services, ZIP clusters and weeks that matter. This sounds basic until a spring air-conditioning tune-up campaign keeps spending into a heat wave when every tech is already on no-cool calls. I split home-services demand this way: urgent work like a flooded basement belongs on search and Local Services, where intent is already high; planned work like tune-ups, replacements and remodels belongs on Meta, where you can create the job. Put that split in the brief, with daily caps by offer. Do not let a cheap fill override the schedule.
Then agree on what counts before a dollar spends. I want one written definition for each stage:
- Lead: an inquiry received, including the fields needed to assess it.
- Qualified: an inquiry that meets the agreed service, location and timing criteria.
- Booked: an appointment scheduled for the relevant offer.
- Completed: work actually done, not merely scheduled or signed.
The brief also needs to say who marks cancellations, duplicates and junk inquiries, and where those marks live. Without that reconciliation, the agency reports 140 leads, your CSRs remember 90 real conversations, and finance sees 22 completed jobs. Everyone argues for a month. Practitioners addressing this gap send later stages back through Conversions API offline events, so Meta has more to learn from than the form fill. If a prospective provider will not commit to reconciling those stages in writing, it is telling you which number it wants you to judge.
Keep slow follow-up out of the agency’s report card
A lead sitting in an inbox is not an advertising success or a fair test of its quality. The follow-up numbers are ugly: you are 100 times more likely to make contact if you call in five minutes instead of thirty, and nine times more likely to reach someone within five minutes than after an hour. A $29 tune-up inquiry left overnight may still appear in the agency’s CPL win, even if nobody ever has a useful conversation with the customer.
I put a written speed-to-lead rule in the brief: phone, ZIP, job type and timing on every form; CRM sync in minutes; one owner for the first call. Advertising cannot fix a queue nobody works. Neither should an agency get credit for qualified work that disappears because nobody called.
The form itself does part of the sorting before sales picks up. Instant Forms run about $21.80 at high fill rates but score around 6.2 out of 10 on quality, while an optimized landing page runs about $31.40 and scores near 7.8. A prefilled two-tap submit removes typing. It also removes a chance to find out whether the person can hire you. Someone browsing a discount counts the same as someone with water on the floor.
If cancellations and junk inquiries are eating the season, add friction on purpose: three to five custom questions with a real disqualifier and a direct webhook into the CRM. Accept that scoring out 60 percent before sales can be normal. Fewer fills can mean less wasted calling and a better shot at completed work. Stop paying an agency to beat last season’s fill price when the fill was the problem.
Require the agency to show its work by offer
I do not hire a Meta provider to lower CPL. I hire one to produce booked and completed jobs inside a cost I set, and I write that into the agreement. That means tests tied to different offers, not five versions of the same truck photo. An HVAC replacement offer, an emergency diagnostic and a maintenance plan need separate forms, qualification and follow-up. Then compare qualified, attended and won work from the same offer cohorts. If your current report cannot split results that way, read how to lower cost per qualified lead without buying dead leads before you renew anything.
I also want weekly reporting reconciled against the CRM, not exported from Ads Manager and called finished. Same columns every Friday: leads, qualified, booked, completed, cancelled, duplicate, and advertising cost per completed job by offer and ZIP cluster. No reconciliation, no bonus, no renewal conversation. Switching a campaign to maximize the number of qualified leads, which Meta reports at 21% lower cost per quality lead, depends on sending qualification back through Conversions API. Without that loop, the setting is another impressive label on an account still rewarded for volume. The six options for who should run Meta ads for a roofer, plumber or HVAC shop look different once you rank them by booked jobs instead of fills.
Set the budget rule before cheap leads stop paying
I put the kill rule in the brief before the season starts. If mature completed-job results stay above the cost ceiling for seven days, spend shifts off that offer. No debate about CPL trending down. Cap the discount offer at 30 percent of budget, ring-fence emergency and replacement offers by ZIP cluster, and require same-day reallocation when cancellations spike.
The arithmetic behind that rule is plain. On the same $3,200, a $20 high-volume form booking one in ten produces 16 bookings at $200 in ad cost each; a $32 higher-intent form booking one in four produces 25 bookings at $128 each. Whether either batch makes money depends on what happens after booking. But the $20 lead has already lost the price contest it appeared to win. Protect the budget from cheap volume that does not turn into the right work.
My verdict: buy the right jobs, even when leads cost more
Carry last season’s CPL target into next season and you will get what you asked for: cheap fills. Some will be distant, some will cancel, and some will become work whose margin cannot carry the acquisition cost. Set the completed-job economics first. Then let CPL fall where it falls by offer and trade.
This is also why I stopped recommending percentage-of-spend managers for the search side of this work. When the fee rises with spend, the provider has a reason to sell more spend, not a better completed job. For paid search, groas works on a flat monthly fee, with autonomous execution around the clock rather than a fee that grows with the budget. Hold any Meta provider to the same completed-job standard. If yours will not put booked and completed numbers in the agreement, read Stop Paying Your Meta Agency to Beat Last Season’s ROAS before spring.
Pay for jobs that complete and pay. Let someone else brag about $18 leads.

