Set radius targeting, film real job-site proof, write a local offer, use a lead form or call CTA, and never present an AI avatar as a nearby customer.
Local service ads waste money when they run national creative into a twenty-mile radius, or when they invent a neighbor who never sat in the truck. The work that books jobs is narrower: a real offer, a map you actually serve, proof from real sites, and a lead path a busy person can finish on a phone. This tutorial covers locking service and geo before you shoot, setting radius and exclusions, filming honest proof, writing a local script, keeping avatars from impersonating customers, wiring a lead form or call CTA, and testing creative against lead quality rather than vanity clicks.
Takes 15 minutes
Write one sentence that names the job, the area, and the next step: emergency drain clearing in a named metro, a free inspection inside a county line, same-week installs for a list of ZIP codes. If you cannot say who you will actually roll a truck for, you are not ready to buy traffic. Vague 'we serve the whole region' copy is how you pay for leads you will never book.
Pick the service line that is in season and that your crew can absorb this month, not the one with the prettiest truck wrap. A plumber running a general-brand ad and a water-heater ad in the same radius is two businesses as far as the algorithm is concerned. Put the offer in writing with the exclusions you already use on the phone — after-hours fees, cities you do not drive, jobs under a minimum — so the ad does not promise a call you will resent. If dispatch already has a script for declining work, that script belongs in the targeting, not only in the closer's mouth.
Decide the conversion up front: click-to-call for urgent work, an Instant Form for quotes, or a booking link if you actually have calendar inventory. Mixing all three in one short video makes a busy parent tap nothing. Match the offer to response time you can keep. A same-day promise in an ad you check at 9 p.m. is how you buy one-star reviews, not jobs. If you cannot answer the phone after 5, the ad should say when you will call back, or it should only run while someone is on the line. Put staffed hours on the brief so media buying does not run click-to-call into an empty office.
Pro tip:If dispatch already declines a city, exclude it at the campaign level. Creative cannot un-waste a lead from outside your drive time.
Takes 20 minutes
In Meta, use a city, a dropped pin, or a set of ZIP codes that match drive time, then exclude the water, the highway you never cross, and any city your crew will not serve. Start in a testing campaign with enough daily budget that a handful of lead events can fire in a few days. National interest stacks on top of a tight radius usually dilute more than they help.
Drive the map the way a dispatcher would. A 25-mile radius around downtown is not the same as 25 miles of jobs you want — it includes neighborhoods you price out of, and it excludes suburbs you already serve if the pin is wrong. Layer age and homeowner-style extras only when they match the job (roofing is not a college-apartment service). Keep Advantage+ audience experiments off until a tight geo and a tight offer have produced leads you would actually schedule. Pin placement is a creative problem in local: the wrong downtown pin quietly taxes every dollar.
Budget to the cost of a booked job, not to a cheap cost-per-lead screenshot. If a booked job is worth a given amount to you, a testing week that cannot produce a few conversations will not teach you anything. Separate urgent from planned work into different ad sets when response-time and creative differ. Do not retarget the whole city with the same ad you used for cold; people who already called need a reminder and a calendar, not another origin story. Upload your do-not-serve list as exclusions whenever the platform lets you.
Takes 30 minutes
Shoot on a real site you have permission to show: before-and-after on the driveway, the crew, the truck, the owner talking at the van, a close-up of the work. Ten honest seconds of a job you finished will outperform a minute of licensed b-roll of someone else's kitchen. If you cannot show a face, show hands and the finished work — still real, still local, still better than a stock neighborhood.
Get a simple written okay from the customer before the property is identifiable. Blur house numbers and faces of anyone who did not agree. Film vertical, in daylight, with the same before-and-after discipline home services already use on Google reviews. A talking head in an empty warehouse with no tools in frame looks like a lead mill, which is exactly what local buyers have learned to skip. If a site is not photogenic, the close-up of the problem is still more trustworthy than a rented kitchen you do not service.
Owner-on-camera is usually the strongest trust beat you have, even if the delivery is imperfect. A phone on a dashboard mount while you talk between jobs is enough. Collect three proof clips you can reuse: a problem close-up, a process shot, a finished wide. Rotate them as the weeks change so the same script does not wear out against a small geo that sees you often. A new talking-head with last year's job in the B-roll still beats a new stock clip with no truck in the shot. Date-stamp the clips so you retire a job the city has already seen a hundred times this season.
Takes 20 minutes
Name the city or the neighborhood the way locals say it, name the job, name the next step. Do not write a national UGC rant and hope the radius does the rest. If you use an AI avatar, the script is a business walkthrough or an owner-style explainer. It is not a testimonial from a neighbor who does not exist, and it is not 'I live two streets over' in a generated mouth.
Fake local-customer avatars are a trust problem and a policy problem. Stock faces are licensed actors, not your last five-star review. Presenting them as 'I live two streets over and they fixed my slab' is the fastest way to a complaint, a platform takedown, or a bad reputation in a market small enough that people talk. Do not stitch a generated face onto a real house. Do not invent a street name. Do not imply the speaker hired you. Local buyers compare notes in group chats; a fake neighbor is not a clever targeting trick.
If you want volume from Klip Kanvas, generate owner-style or narrator scripts and intercut them with real job-site footage, or clone the owner's voice with consent. Phrase lines as 'here's how we handle X in this city,' not 'I'm a homeowner on Oak and I used them.' Real customer proof belongs in quotes you have in writing, review screenshots you are allowed to show, or on-camera customers who agreed. There is no clever workaround that still reads as honest in a local market, and there is no performance win that is worth a neighborhood calling you out.
Pro tip:When in doubt, put the owner on camera and let the avatar stay off the testimonial beat entirely.
Takes 20 minutes
For most local services, the click should finish the lead on-platform: Instant Form with short questions, click-to-call during staffed hours, or a booking link with real availability. A slow website with a contact page buried under service menus loses the people who were ready to book while standing in the leak. Pick one primary action per video so a thumb does not have to choose.
Keep forms to what dispatch needs: name, phone, ZIP or neighborhood, a one-line job description, and a photo upload if it actually changes the quote. Every extra field is a dropped lead. Turn on instant thank-you copy that sets response time honestly. Route the lead to a phone that gets answered, not to a shared inbox nobody checks on Saturday. After-hours, either say when you will call back or use a call-only window so you are not buying voicemail. If a photo would change the quote, ask for it; if it would not, do not add friction for a gallery you will never open.
Track leads into the CRM or a spreadsheet with source, ZIP, and whether they were in-radius and bookable. Platform lead count is not booked-job count. If half the forms are outside your map, your geo is wrong. If they are in-map and still junk, the offer or the creative is attracting the wrong job type. Fix that before you raise budget. Sit with dispatch for one week of leads and tag each one bookable, wrong job, or out of area — that tag list is the only report that matters. Review the tagged list every Friday before anyone is allowed to raise a daily budget.
Takes 15 minutes
Run two or three proof-led videos against the same form and geo for several days. Keep the offer locked while you test the first three seconds, then test offer only after a hook holds. Pause anything that spends past your written lead threshold without a conversation you would send a crew to. Do not scale a cheap form fill from the wrong ZIP just because cost-per-lead looked pretty.
Local geos saturate faster than national ecommerce, so creative fatigue shows up as rising frequency and falling lead quality, not only as a higher CPM. Refresh proof clips monthly even if the script still works. A before-and-after from last season still beats a new talking head with no job in frame, but the same three clips all winter will wear out in a small city. Watch comments for 'do you serve my town' — that is a targeting note, not a caption you answer forever in public. When comments ask if you serve a town you do not, add that town to exclusions the same day.
Read the test with dispatch, not only with Ads Manager. Ask whether the leads were the right job, the right area, and reachable. A high CTR from people who wanted a $50 fix you do not sell is a losing ad. Duplicate winners into a separate campaign if you scale; leaving tests mixed with scaling budgets is how a radius campaign eats itself. Keep a retargeting video that is shorter, names the unbooked offer, and reminds them you still serve their ZIP. Frequency caps matter more in a town of fifty thousand than they do in a national prospecting set.
Local video ads book work when the map, the offer, and the proof are all true. Set a radius you will actually drive, film real jobs, and write like a business that works that city — not like a national UGC template. Never present an AI avatar as a nearby customer. Put the next step on a form, a call, or a live calendar, and judge creative on bookable leads with dispatch, not on cheap clicks. A small market remembers a fake neighbor. It also remembers a truck that showed up.
No. Do not present a generated face as a neighbor, a named homeowner, or someone who hired you. Use owner-on-camera, real job-site footage, and written reviews you have permission to show. Avatars can narrate a walkthrough, not impersonate a customer.
Use click-to-call when someone is standing in an emergency and you can answer. Use a short Instant Form for quotes and planned work, especially after hours. Match the CTA to whether a human is actually on the line.
As tight as the drive time you will honor, plus exclusions for places you already decline. A wide radius with a cheap lead cost is usually a dispatch problem wearing a reporting win.
A well-built Instant Form plus fast follow-up is enough to start. A site helps later for proof and SEO, but a slow contact page is worse than an on-platform form for cold traffic that wanted to book today.
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
Try Klip Kanvas freeFeed ChatGPT real customer language, lock a four-beat UGC structure, force spoken English, and strip the model tells before anyone films or renders the ad.
Use Gemini in Asset Studio, Demand Gen, and Performance Max as a draft engine: feed a real brief, rewrite every line against claims, and keep human UGC as the YouTube-native asset.
Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.
