Instagram ads for HVAC companies

Start with reach, and start by conceding it, because for this trade the app does not win on that axis and pretending otherwise is how a contractor ends up disappointed with a campaign that worked. Pew's November 2025 survey puts Instagram at 80% of US adults aged 18 to 29 and 62% of 30- to 49-year-olds, against Facebook's 68% and 80% — and then 40% against 74% in the 50 to 64 band, which is where a paid-off house and a fifteen-year-old furnace usually are. If the offer is a system replacement, the Facebook guide in this pair is the one to read. What this app has is the newer owner and the smaller job.

Now the thing that does not change, because it is the reason most contractors come to this page. A replacement sold on a fixed monthly payment is a credit advertisement, and the campaign carrying it declares financial products and services and loses location exclusion, ZIP, neighbourhood, sub-city, metro-area and small-geographic-area targeting, lookalike and saved audiences, custom age ranges, and any radius under 15 miles or 25 kilometres in the US and Canada. None of that is a Facebook rule. Since 31 March 2020 every campaign created through the Marketing API has had to declare a special ad category value, with NONE as the answer for everybody else, and the value sits on the campaign — the placements are chosen underneath it. Regulation Z travels the same way: it attaches to what the advertisement says, not to where it rendered. So the financed offer on Instagram is the identical cage bought with a smaller slice of the homeowners you wanted. It is the worse place to run that ad, not the way around it.

The third thing is the one that actually changes the playbook, and it is an absence. Instagram has no surface for capturing demand. On Facebook the 1.91:1 shape routes to the right-hand column and search results, which is where a six-in-the-morning search for a dead air conditioner lands, and Marketplace is where somebody prices a used window unit in the second week of a heat wave. Instagram has four image placements against Facebook's nine and none of them is a search result. Nobody opens this app to find a contractor while the house is at 88 degrees. Every reader here is mid-scroll, so every ad has to interrupt rather than catch — which is why there is no emergency dispatch offer in the table below, why nothing in the Browse batch is 1.91:1, and why all ten of those ads open on an object or a moment rather than on a service name.

Fourth, what the format gives back, and it selects your offers whether you meant it to or not. This app rewards showing the equipment and the work, and that rules out most of the money in a contracting business: a furnace in a basement is not an image, and neither is a load calculation. A mini-split head, a linear ceiling diffuser, a thermostat in its box, a return plenum sealed with mastic, a thermal pair of the same joist two hours apart — those are. The Browse batch's objective split lands at four Leads, four Traffic, one Sales and one Awareness, against the Facebook batch's five, three, one and one, and the Sales buy here is a $349 thermostat install that genuinely completes at a checkout. That the small-ticket end of the trade has a checkout at all, where the Facebook page barely does, is the clearest single read on what this app is for.

Fifth, the specs, and the received wisdom is backwards again. Meta's ads guide recommends about 40 characters for an Instagram Feed headline against Facebook Feed's 27 — the app everyone treats as the wordy one gives you thirteen more characters — with primary text about 125. The image is 4:5 at 1440 by 1800, identical to Facebook Feed, so one export serves both; but the aspect-ratio tolerance is 1% here against 3% there, and the minimum width is 500 against 600. A crop a few pixels out of true passes on Facebook and fails on Instagram from the same file. Measured against the Browse batch rather than assumed: 46 of the 50 headlines fit 40 and exactly one fits 27, all ten lead headlines fit 40 and none fits 27, none of the 50 primary texts fits 125 at 257 to 537 characters, and 8 of 10 descriptions fit 30 in a field the feed does not show. Across all six Instagram batches in this cluster, 263 of 300 headlines fit 40 and 14 fit 27, and not one of 300 primary texts fits 125.

Finally, a question worth asking before any of this matters: are you choosing Instagram at all? With Advantage+ placements on, which is the default and what almost every contractor runs, one ad fills all four Instagram surfaces and all nine Facebook ones from a single export and a single headline, and the app split is a delivery outcome rather than a decision. That is usually the right setting. It also means an Instagram strategy for this trade is mostly a creative decision — which offers you brief, and how you shoot them — rather than a media one, and that restricting placements to Instagram is a thing you should have to justify rather than a thing you do because you made a vertical video. It also means writing the 30-character description line even though this feed will not show it, because Marketplace will, and it is free.

Ads Manager offers the same six objectives here as anywhere — Awareness, Traffic, Engagement, Leads, App promotion and Sales — and this is the only page in this cluster where Leads is not the default. That is not a stylistic preference and it is worth understanding before you act on it, because the reason is the missing surface rather than anything about the audience.

On Facebook, a meaningful share of an HVAC ad's readers are already in the market: they searched, they were pricing a unit in Marketplace, their system failed this morning. Asking those people to book is asking at the right time. On Instagram nobody arrived with a problem. Every reader is mid-scroll and every ad interrupts, and most of what this app can sell is something the reader cannot name and has never priced — a ductless head for one room, decoupled moisture control, a return plenum rebuild, a concealed ducted retrofit. Asking for the appointment in the first ad is asking too early, and the click you actually want is the read. So Traffic takes the largest share here, and the Browse batch's four Traffic buys are all offers whose argument does not fit in an ad.

Leads is still a real and frequent answer — it is the other four in that batch — but it earns its place on a narrower test than on the Facebook page: the offer has to be priced, bookable and need no explanation. A $340 duct audit and a $129 pre-season start-up both pass. A concealed ducted retrofit does not. Sales is stronger here than anywhere else in this cluster, because the two offers in this trade that genuinely complete online — a flat-price thermostat install and a maintenance plan with a cancel button — are both small-ticket, which is to say both of them live on this app. And Awareness is more defensible here than on Facebook for the same reason Traffic is: with no way to capture demand, being the shop a household already knows the name of in March is a large fraction of what the money can buy.

This table is not the Facebook table with different pictures, and the difference is the most useful thing on this page. The Facebook guide lists twelve offers spanning an $89 diagnostic to a $14,000 replacement, and the replacement and its financing are the middle of it. Here the price ceiling is about $8,000 and most rows sit under $2,000, because the format selected them: every offer below is something you can photograph, and the ones that are not — the furnace swap, the load calculation, the emergency call-out — are missing on purpose rather than by oversight.

Each row pins the same four decisions as the Facebook page: who the offer is actually for, which angle family argues it best, which objective buys it, and a finished ad from the Browse page showing the whole thing assembled. Ten of the twelve have one, which is every ad in that batch used exactly once. Two do not, and an absence is more honest than borrowing a neighbouring ad.

One difference from the Facebook table deserves calling out rather than leaving to be noticed. That page recommends Identity and Emotion for no offer at all, because the batch behind it took a $249 zoning damper, briefed as a machine, and rendered "Stop tiptoeing down that hallway to check on her" into the image — second person, about the reader's own household, about somebody who needs looking after. This page recommends it for exactly one row, and the reason is an experiment rather than a change of heart. That brief kept the family and moved the subject from the occupant to the house: a 1920s plaster cornice, and equipment chosen so the cornice survives. What came back was "The wall you refused to ruin. You stripped it by hand." Still second person, still emotional, and not a personal attribute — because Meta's policy is a list of attributes and a renovation you chose to do is not on it. An identity ad is dangerous when it tells the reader what they are and workable when it tells them what they did, and that distinction is worth more than the row it produced.

Almost nothing in this section is different from the Facebook guide's, and that is the section's point. The special ad category is declared on the campaign, not on the placement. Instagram Feed, Stories, Reels and Explore sit inside a campaign that has already answered the question, so a financed offer running here loses exactly what it loses there — location exclusion, ZIP, neighbourhood, sub-city, sub-neighbourhood, metro-area, small-geographic-area and electoral-district targeting, lookalikes, saved audiences, custom age ranges, behaviour and demographic targeting, interest exclusion, and any radius under 15 miles or 25 kilometres in the US and Canada. The declaration is compulsory on every campaign either way, and has been since 31 March 2020, with NONE as the answer for the tune-up, the audit, the thermostat and the plan.

So the first Instagram-specific instruction is a negative one: choosing this app is not a targeting decision and cannot be used as one. If somebody has told you to move the financing ad to Instagram, they have confused a placement with a campaign field, and the cost of taking the advice is the same restriction bought against a smaller share of homeowners. Run financing in its own declared campaign on whichever placement actually performs, and keep every other offer in an ordinary campaign where a tight radius still works.

What genuinely changes is how much two of the levers matter, and both changes run in the same direction: the small-ticket offers this app selects for are less able to absorb a bad audience than the replacement is. A $349 thermostat install cannot pay for a forty-minute drive. A $340 duct audit cannot either. On the Facebook page a slightly wide radius costs you some wasted spend on a job worth thousands; here it can make the job itself unprofitable to attend. Think drive time rather than miles, remember that the ordinary floor is 0.63 miles or 1 kilometre, and set it tighter than feels comfortable.

Age is the second, and it inverts. On Facebook the useful bands are 30 to 64 and they are well populated — 80% and 74%. Here the 50 to 64 band is 40% and the 18 to 29 band, which is the app's strongest at 80%, is substantially renters who cannot commission the work whatever they think of the ad. The band that is both present and able to buy is 30 to 49, at 62%. That does not mean excluding everyone else, but it does mean age banding is worth more thought here than on the app where every band is a homeowner. And the usual caution applies: targeting a band is fine, telling the reader their age in the copy is not.

Two levers survive everything and both are under-used, and on this app the second one matters more than it does on the other. Custom Audiences keep inclusion, exclusion and expansion even under the financial products and services category, along with Advantage+ Audience and Detailed Targeting Expansion, which makes your own list the only precision instrument left in a financed campaign — check one with is_eligible_for_sac_campaigns before launch rather than after. And this app is where a large part of that list gets created, because the small jobs are the first thing a household ever buys from a contractor. That is the argument the funnel and budget sections are built on: Instagram's return is partly a list-building return, and a list is worth what it can be pointed at later.

One hard-error footnote, because it fails at launch rather than in delivery. A business admin has to accept Meta's non-discrimination policy in Business Settings before a housing, employment or financial campaign will deliver at all — the API returns error 2859024, "Certification Required". Its neighbour, 2909035, is what comes back for a custom age range, a saved audience, a lookalike or a radius under the floor. If a financing campaign will not start, read those two codes before you touch the creative.

The Facebook guide's answer to this is three-way and starts with the phone. This one starts by removing the phone, and that is most of the difference. Call ads are delivered on Instagram Feed and Instagram Stories, so the format exists here — but the reader does not. Nobody scrolling Instagram is standing in a 96-degree house waiting to be rescued, and building this app's plan around call ads is buying the format without the moment. Keep the call buy on the other page.

What replaces it, for most of the offers below, is a landing page — which follows from Traffic being the default rather than being a separate decision. Almost everything this app can sell is a thing the reader cannot name and has never priced: a ductless head for one room, moisture removed independently of cooling, a return plenum, linear slot diffusers in a plaster ceiling. Those need a page, and the page needs the price on it. An offer pitched at somebody restoring a house has to survive the question of what it costs, and a number answers that faster than a form does. Never the homepage — a reader who tapped an ad about a thermal image of their own ceiling and lands on a rotating hero of a smiling technician has been asked to start over.

Two offers want something better than a page, and this is the app where they live. A flat-price thermostat install and a maintenance plan with a cancel button both complete in a checkout with nobody leaving the depot, and where that is genuinely true the button should complete the purchase rather than open a form. Be honest about which side of the line you are on: if the money is collected at the visit, that is a lead wearing a Shop now button, and dressing it up costs you the trust the flat price was buying. For the priced, bookable visit — the audit, the pre-season start-up — use an Instant Form on the Higher intent type rather than More volume, because a pre-filled lead is a name and a number the person never looked at, and in a trade where the callback happens hours later that is most of them.

The fourth option is the one this app adds and the one to be most careful with. Lead ads can send people into a conversation on Messenger, Instagram or WhatsApp, and the instant forms and Messenger conversion location runs one campaign and lets delivery pick per person. For a two-van shop that thread is a real booking path — it is where a photograph of the wall, the panel or the old unit arrives without anybody asking for it — and it is also a real hazard, because a direct message is not a dispatch system. A household that messages your account at eleven at night has not booked anything and may believe it has. Use the thread only if a person actually answers it and the conversation ends with a date, and never for anything time-critical.

Start with the rule that is not Meta's, because it caught two of the ten ads on the Browse page and only one of them was a financing ad. Regulation Z, 12 CFR 1026.24(d), makes four things triggering terms in a consumer credit advertisement: the amount or percentage of any downpayment, the number of payments or period of repayment, the amount of any payment, and the amount of any finance charge. State one and the ad owes the reader the downpayment, the repayment terms across the full term of the loan including any balloon payment, and the annual percentage rate, using that phrase. Subsection (c) adds that a stated rate of finance charge must be expressed as an annual percentage rate using that term.

The financing ad on the Browse page was briefed as an experiment and returned the result cleanly. Its Facebook counterpart asked for the rate, the term and the total repayable to be printed on the advert, and produced a compliant and rather good ad: $128 a month, 9.9% APR fixed, a 120-month term, $15,360 total repayable, and the arithmetic checks. This brief is the same family and the same sub-angle with that clause deleted — a room fixed on a fixed monthly payment, full stop — and it came back as a poster reading $69 a month with "No down payment" in the copy and no rate, term or total anywhere in the image or the text. Two triggering terms, no disclosure, and the pipeline did nothing wrong: it wrote the category's house style, which is what it was asked for. The disclosure is a property of the brief rather than of the model, and the practical instruction is that whoever arranges your financing already has approved language, because this rule is why they have a compliance function.

And the credit line does not stay in the credit ad, which is the finding to carry into any trade that finances. The ductless ad on that page was briefed as a problem-and-friction piece about a window unit, with no financing in the brief at all, and one of its primary texts still ends with "0% financing available". That is a stated rate of finance charge not expressed as an APR, and it is enough on its own to make the campaign a financial products and services campaign with the fifteen-mile floor attached. Check every ad in a set for a credit line before launch, not just the one you think of as the financing ad.

The second thing this batch established is mechanical rather than stylistic, and it is the single most transferable rule on this page: choose a visual direction with no proof slot. The HVAC Facebook batch rendered an invented statistic into five of its ten images, one of them citing county inspection records as a source. This batch rendered zero. The difference is not restraint or a better prompt — it is that the Facebook run used stat heroes, offer stacks and comparison charts, which have somewhere to put a tally, and this one used a post-it, a packaging shot, a thermal pair, a notebook and a quiz, which do not. Where the direction offers a slot, the pipeline fills it whether or not you asked; choose a direction without one and the problem does not arise. Two ads here still carry invented figures in the copy, which is the residue rather than the pattern.

The third finding is the sibling of that one and it cost the batch its worst ad. One brief specified an offer, an audience and an objection, and did not specify who was talking. The direction was a phone Notes screenshot, which has to have an author, so the pipeline invented a homeowner and wrote a private diary entry in their voice, timestamped at 5:12 in the morning, with two of the five primary texts in the same invented first person. That is a fabricated customer testimonial in the most persuasive wrapper available. Name the speaker as well as the thing: any direction that implies an author — a note, a letter, a review, a message, a screenshot of a text thread — will invent one, and what it invents makes claims on your behalf.

Fourth, the shapes and the counts, and they are not the Facebook page's. Render 4:5 at 1440 by 1800, identical to Facebook Feed, so one export serves both — but the aspect-ratio tolerance is 1% here against 3% there, so a crop that passes on Facebook fails on Instagram from the same file. Do not buy 1.91:1: this app has no surface built to want it, which is also why the Browse batch contains none. Reserve 9:16 for Stories and Reels, where two of those ten sit. On copy, the recommendation is 40 characters of headline and about 125 of primary text, and this batch measured 46 of 50 headlines inside 40 with exactly one inside Facebook's 27 — every one of the ten lead headlines fits here and none of them fits there, which is the mirror image of the Facebook batch. None of the 50 primary texts fits 125, running 257 to 537 characters, and across all six Instagram batches in this cluster not one of 300 does. Longer is not forbidden, it is truncated, so the argument has to survive inside the first sentence.

Three stages here, and a fourth line that is deliberately not on this app at all. The shape looks like the Facebook page's and the weights do not, because what this app is buying is different: the cold stage is acquiring small first jobs rather than large ones, and the warm stage is larger in proportion because Traffic is the default and a Traffic campaign's whole output is a pool of people who read something.

Keep it this simple. A contractor's budget is small relative to the audience arithmetic Meta's delivery system wants, and every extra ad set divides the same evidence into thinner piles. Variation belongs at the ad level, inside one ad set, and the thing that varies is the offer rather than the audience.

The customer stage is the one to read carefully, because on this app it is not a retention line — it is the return. The small jobs sold in the cold stage are, for most of these households, the first money they have ever given a contractor, and the record that creates is what the maintenance plan, the second system and the replacement a decade from now all get sold against. That record is also the only precision instrument that survives into a declared financial campaign, where lookalikes and saved audiences are removed. So the honest way to describe an Instagram budget in this trade is that part of its return is a list, and a list is worth what you point it at.

Do the arithmetic before you copy a number from anywhere, including here. Meta's delivery system needs roughly 50 optimisation events per ad set per week to leave the learning phase; below that the ad set sits in learning limited, which is not a penalty so much as an admission that the system cannot optimise on what you are giving it. Take your own cost per lead, multiply by 50, and that is the weekly spend at which one ad set is delivering with full information.

Now the part that is specific to this app and is genuinely good news, because most of this page has been conceding things. The offers here are small, which means the events are cheap, which means 50 a week is a reachable number in a way it is not for a campaign selling five-figure replacements. The app with the worse reach has the better chance of getting an ad set out of the learning phase, because it is optimising toward a $340 audit rather than a $14,000 system. That is a real argument for putting the shoulder-season budget here rather than a consolation prize.

Which is the second point. The Facebook guide's central budget argument is that the received advice — watch the forecast, push hard when the weather turns — is harmful, because a hand edit during a cold snap is a significant edit and a significant edit restarts the learning phase on the one morning of the quarter when demand is real. Its resolution is to keep an always-on ad set alive through the shoulder season and change the ads rather than the ad set. This app is what that ad set should be running. Duct sealing in October, a start-up visit in March, a thermostat in any month, a ductless head for the room that was unbearable last August — none of those depend on a heat advisory, and all of them can carry an ad set through the quiet weeks so that it has already learned when the emergency campaign matters. The two apps' seasonality curves are complementary rather than parallel.

Finally, what to measure, and this is where most Instagram budgets in this trade get killed by the wrong report. Do not judge this app on cost per lead against the Facebook campaign: it is selling different offers at a tenth of the ticket to a different half of the market, and the comparison will make you turn off the thing that is working. Judge it on three numbers you have to get from your own office software rather than from Ads Manager — how many new customer records it created, what share of those bought a second job within twelve months, and what the maintenance-plan attachment rate is on jobs that started here. If the small job is only ever a small job, this app is a marginal buy. If a meaningful share of those households come back, it is the cheapest customer acquisition in the business and the report that shows it is not on this platform.

The first three are the ones that cost real money, and the first is the reason most contractors arrive at this page. The rest are the ways a good campaign quietly becomes a worse one, and several of them come straight out of the ten ads on the Browse page rather than out of anybody's opinion.