A roofer opens a Meta ads proposal promising 30 to 40 percent cheaper leads in 90 days. I built the artefact for this teardown from a dozen real pitches: a composite four-page proposal for a roofer, HVAC company or plumber, complete with homeowner audiences, guaranteed volume and done-for-you creative. I chose it because some version of these six promises keeps landing in owners’ inboxes.
My earlier piece, Your Facebook Cost per Lead Is a Fraction, Not an Agency Promise, put the math in the headline. This is the second look an owner needs while holding the proposal: which promises can move that fraction, and which just make worse leads look cheaper? Cost per lead (CPL) depends on the auction price, click-through and the share of clicks that become leads. Nobody negotiates it directly.
Most of this pitch either cannot move CPL or lowers it by buying cheaper, worse leads. The lines worth paying for concern creative, the optimization event and booked jobs. Say you spend $5,000 a month and someone promises to cut CPL by 30 percent. That sounds like savings until the leads stop answering the phone.
Page one: the 90-day CPL promise
The composite opens with a 30 to 40 percent CPL cut in 90 days. Page two lists proprietary homeowner audiences, 80 guaranteed leads a month, done-for-you creative refreshed monthly and optimization for leads. Page three names a management fee of 15 percent of spend with a $1,500 minimum, plus a monthly reporting call. Page four offers logos and two short testimonials. I am taking it in that order, because each page asks for a little more trust than the last.
The headline fails its first test: it names a result but no mechanism. For context, Home and Home Improvement leads on the Leads objective average about $41.26, against $27.66 across industries. Verified high-ticket medians run much higher: near $116.75 for roofing and AC install and $72.97 for plumbing. Nobody sets those numbers in a meeting. A lower CPL requires a lower effective auction price, more clicks from the impressions you buy, or more leads from those clicks.
The proposal does not have to guarantee an exact result to be useful. It does have to say which input it intends to change and what it will do first. Without that, a percentage cut is a hope attached to an invoice. Verdict: do not buy the headline until you can see the work behind it.
Promise 1: “Proprietary homeowner audiences”
Page two starts with targeting. This is a familiar place to sound technical without committing to much. Under Advantage+ broad delivery, location, minimum age, language and exclusions are hard limits, while age ranges, gender, interests and lookalikes are suggestions Meta can go beyond. For a plumber in one metro, the service-area fence matters more than an interest stack labeled homeowner. Broad targeting can match the proprietary grid because the ad itself now does much of the selecting.
Verdict: mechanically weak. I used to sell neat audience builds. I was wrong about where the work lived. A list of interests can look painstaking on a slide, but the owner needs to know who can actually see the ad and whether the people responding are in the service area. Those are different questions.
Keep this line only if the vendor names the fence first: service-area zip codes, renter exclusion and age floor. If the pitch leads with interest piles such as home improvement, expect renters and other tradies in the lead list. Slicing a small budget into more audiences does not, by itself, lower CPM or raise click-through from the right people. Ask for the enforceable boundaries before admiring the proprietary grid.
Promise 2: “Guaranteed 80 leads a month”
Now the proposal promises a count. This is where CPL gets gamed, because a vendor can change how easy it is to become a lead without changing how likely that person is to book a job.
Instant Forms averaged $34.10 in a 138-campaign dataset, against $52 to $78 for website conversion campaigns. Pre-filled data removes typing. That reduced friction can lift conversion rates up to 50 percent versus external sites, with mobile users about 63 percent more likely to complete. Easier finish, more finishes, lower CPL. The dashboard looks healthy.
But lead-to-sale rates run 20 to 30 percent lower for instant forms. A casual browser can raise a hand in ten seconds. Verdict: the volume is possible; its value is unproven. You can hit the lead count and miss the job count.
Run the booked-job math before you sign. At Facebook close rates of 8 to 20 percent, modeled cost per booked job runs $780 to $1,463 for roofing at $117 CPL, $332 to $487 for plumbing at $73 CPL, and $252 to $420 for AC repair. The useful question is not whether the form produces names. It is what those names cost once someone has booked work. A vendor that guarantees leads without pricing the booked job asks you to pay for hands raised, not trucks rolled. Ask for cost per booked job in writing.
Promise 3: “Done-for-you creative refreshed monthly”
This is the first promise with a plausible route to cheaper useful leads. Specific creative can earn clicks from the right people and give Meta stronger predicted engagement. That is how it can improve click-through and lower effective CPM without touching the bid. Verdict: the mechanism is real; the proposed pace is the problem.
What the deck calls a monthly refresh, I call minimum maintenance. On Meta, a winner lasts 2 to 6 weeks at meaningful spend, with 4 to 6 new concepts a month needed under about $10k spend. One sign of fatigue is CPA running 25 percent above its first two weeks at similar spend for 3 days. A calendar entry saying new creative does not tell you whether anyone notices that change or responds to it.
For roofers and HVAC shops, one new video a month of a tech talking to camera can tire right when storm season or summer demand hits. Keep this line only if weekly testing and a kill rule come with it: which hook gets tested next, what click-through from homeowners in your zip codes counts as winning, and what gets paused when. I keep the vendors who can name the next three tests. I cut the ones selling a content calendar. Pay for iteration, not decoration.
Promise 4: “We optimize for leads”
The final promise on page two sounds reasonable until you ask what lead means to the system. The optimization event decides who Meta looks for next. Optimize for raw form submits and delivery learns to find more easy form-fillers. Send back a CRM stage such as qualified or booked and you give it a signal closer to the people you want.
Conversion Leads with Maximize qualified leads sends statuses back via Conversions API for CRM, but it requires Instant Forms, 200-plus leads a month and a target stage reached by 1 to 40 percent within 28 days, with about a month for full effect. Most $3k to $5k home-service budgets never clear that bar. Verdict: optimizing for leads is mechanically honest and commercially wrong for most shops. A vendor should explain the event it can actually use, not imply that a CRM connection solves the volume requirement.
The distinction shows up in a simple comparison: $3k on Campaign A at $15 CPL produces 200 leads; at a 4 percent close rate, that is 8 deals at $375 per deal. $3k on Campaign B at $45 CPL produces 67 leads; at a 20 percent close rate, that is 13 deals at $231 per deal. Ads Manager crowns A on CPL. Sales wins with B.
Keep this promise only if the vendor can show how qualified or booked stages get back to Meta, whether the account has enough events to use them, and how booked jobs appear in reporting either way. If the only answer is more leads, you already know which campaign the vendor will celebrate. Make the booked job the test, even when the platform cannot optimize to it yet.
Page three: a rising fee and a monthly call
The proposal now names its price: 15 percent of spend, subject to a $1,500 minimum, plus a reporting call each month. Freelancers run 15 to 25 percent of spend and agencies 10 to 20 percent, alongside setup fees of $500 to $7k and minimums of $3k to $5k. The structure is familiar. That does not make its incentive neutral.
At 15 percent, $5k in spend calculates to $750 and $20k to $3,000 for largely the same management work; the $1,500 minimum overrides the smaller calculated fee in this proposal. Above that floor, the fee climbs mechanically with spend. For HVAC, that collides with seasonality: CPLs drop 20 to 30 percent in shoulder seasons such as March to April and September to October as competition eases. In a dead month, the honest advice can be to spend less. A vendor paid more when spend rises has a reason not to volunteer that sentence.
The monthly call has a similar blind spot. A deck once a month can show blended CPL and screenshots. A daily action log can show which ad tired, which budget moved, which edit restarted learning and how many events each ad set banked that week. Under about $40 a day, one local ad set makes more sense than splitting $20 a day three ways and leaving each too thin to exit learning. If a small-budget proposal lists broad, local and retargeting ad sets, the log exposes that choice in week one. The deck in week four may just call it optimization.
Verdict: the fee rewards spend, while the call tells you too little about the work. Ask what you pay when spend falls and what you can inspect between calls. Pay for the log, not the slides.
Page four: logos and two testimonials
The proposal closes with logos and two short testimonials. They may tell you that other businesses hired the agency. They do not tell you which CPL input changed, what happened to lead quality or how many jobs were booked. That is not a criticism of those customers; it is a limit on what this page can prove.
Verdict: keep the page in its proper place. If pages one through three have explained the creative tests, optimization event, booked-job reporting and fee, the testimonials are a final reassurance. If they have not, logos cannot do that work for them. I would go back to the promised 90-day cut and ask for its mechanism before signing.
What all four pages leave out
Roofing, HVAC and plumbing do not live on CPL. A roofer needs cost per booked inspection and cost per closed roof. A plumber can break even on low-ticket repair via Meta while high-ticket installs carry the return; blended CPL hides that difference. Seasonality can do the same. Peak summer and winter can raise CPL alongside urgency and ticket size, so pausing may look cheap while costing the best jobs. Get the plan by season in writing before paying anyone to spend through it.
Speed-to-lead is the other omission. HVAC data puts contact within 5 minutes at 21 times more likely to convert than contact after 30 minutes. A higher-intent form with a review step cuts junk 30 to 40 percent for 15 to 20 percent higher CPL when paired with a 60-second text and a 5-minute call. The form may cost more per submission and less per worthwhile conversation. If the office calls back in four hours, no audience grid or creative test fixes that. The agency gets blamed for a phone problem.
Before accepting a cheaper-leads claim, ask who answers those leads, how fast they respond and where booked jobs get recorded. Otherwise the proposal ends at the form while the business starts at the phone.
Scorecard: which promises survive the fraction?
- Worth paying for: creative tested often enough to earn clicks from homeowners in the service area, and a workable path from CRM stages to qualified or booked-job measurement. These connect the work to the people and outcomes that matter.
- Possible but risky: guaranteed lead volume built on easier Instant Forms. Friction falls, so volume can rise and CPL can fall while cost per booked job climbs.
- Not a plan: a proprietary audience grid presented as a way around broad delivery, or a fixed CPL cut with no named input. The 15 percent fee is a commercial choice, not a mechanism for cheaper leads.
If the proposal cannot say which input moves first and how many weekly events each ad set can bank, I do not call it a media plan. For another example of the same pattern, see The “We’ll Halve Your Meta Cost per Lead” Agency Pitch, Taken Apart.
Three questions before you sign
Ask these in order. Do not let a polished case-study slide answer for the proposal in front of you.
- Which CPL input moves first? What creative or form change ships in the first 14 days, and how will you show whether effective CPM, click-through or form conversion moved?
- What happens after a form submission? Can you send qualified or booked stages back from our CRM, do we have enough events to use them, and will you report booked jobs even if the campaign must optimize for an earlier event?
- What do I pay and see when spend falls? How does your fee behave, and what actions can I inspect each day rather than hearing about them once a month?
A serious operator may not promise the cheapest CPL. They should be able to answer all three without changing the subject to lead volume.
Overall verdict: keep the mechanism, lose the pitch
I would keep the creative testing, the insistence on a meaningful optimization event and the demand to measure cost per booked job instead of stopping at cost per form. I would lose the audience mystique, the unsupported percentage cut and the idea that a bigger spend should automatically earn a bigger management fee.
That last point is where groas offers a better operating model for paid and organic search, rather than a claim that it runs this Meta proposal. Its flat $1,499 monthly fee does not rise with spend; search campaigns are worked around the clock, actions are logged with their reasoning, and a named human strategist owns direction. No cut of spend and no setup fee. If you are hiring someone for Meta ads, make them meet the same standard of visible work and booked-job accountability. Sign for what gets measured after the phone rings, not for a prettier lead count.

