Garage Door Leads: Qualify and Answer Them Faster With AI

Reply within minutes with a message that asks what the door is doing and whether anyone or anything is trapped, then sort the lead into an urgency tier before you quote. Bought leads are worth it only when cost per booked job stays under a ceiling you set; I start at 30% of gross profit per booked job.

Speed is the part of this you control. It is also what a bought lead really sells you: a lead shared with three firms tends to go to whichever calls first. I will not quote a statistic on how quickly a lead goes cold, because it varies by source, hour and job type. What I can give you is a way to measure it on your own leads, a triage page your office can follow at 7am, and the arithmetic that decides whether a lead source is worth keeping. For a chatbot that collects the door type, fault and photos before anyone phones, see the garage door repair chatbot tutorial.

Follow me on Instagram@sagnikteaches

Where garage door leads come from, and who pays for what

Each source has a different cost, a different catch and a different thing you must record. Give every lead a source tag at the moment it arrives, because you cannot work out cost per booked job later from memory.

Connect on LinkedInSagnik Bhattacharya

On Local Services Ads specifically, Google's help page says valid leads are charged, that charged leads are reassessed over a 30-day review and may be credited automatically if they turn out to be poor quality, duplicate or invalid, and that you can use a feedback survey on each lead. It also says credits are not available everywhere, so check what your own account offers; Google's lead credits page is the place to look. If you want to challenge leads on that platform, how a plumber builds an evidence pack to dispute leads shows the routine.

Subscribe on YouTube@sagnikteaches

Two more lines on that page bear directly on speed and duplicates. A missed call during your business hours counts as a charged lead if the caller stays on the line for more than 20 seconds, so on that channel an unanswered phone is a lead you pay for and never speak to. And follow-up calls or messages from the same customer, on the same phone number or email, are not charged again while you keep interacting within 15 days; the duplicates that cost you money are the ones that arrive through a different source or a different number. Google has also begun moving Local Services Ads into a Performance Max campaign type with pay-per-lead goals, in phases from August 2026 into 2027. Manual bidding such as a maximum cost per lead is not supported there, and the old performance reports do not move across, so download yours and keep your own lead log rather than relying on Google's history.

Five questions to put to any lead seller before you pay

Bought leads are not one product, and the answers to these questions move the cost-per-booked-job sums further down more than any reply template does. Ask them in writing and keep the replies with your invoices.

1. Is the lead exclusive or shared, and with how many other firms? A shared lead is a race. It should be priced lower, and the reply-time bands below will matter more to you.
2. What counts as a billable lead? A call of any length, a form with no phone number, a customer asking about a different service, an address outside your area. Each answer tells you how many of the leads you pay for are real enquiries.
3. What gets credited, and how do I claim? Duplicates, wrong service, outside your area, existing customers, spam. Get the time limit and the evidence required.
4. What are the minimum spend and the notice period? A monthly commitment hurts in a quiet month.
5. Can I pause or cap it? A daily cap or a pause switch protects you on the days when the vans are full and the leads would go cold anyway.

Put the answers next to each source in your sheet. A shared lead at $38 with no duplicate credit and a 30-day notice period is a different purchase from an exclusive lead at $55 with credits for duplicates, and only one of them is cheap.

Instant reply: a fixed first message, then an AI-drafted second one

Treat the reply as two messages. The first goes within seconds and is a fixed, pre-approved text with no AI in it. Its job is to say you have the enquiry, ask the one question that matters, and give a safety line. The second goes within five to ten minutes and is personalised, which is where AI earns its keep.

First message (fixed): Hi [first name], thanks for contacting [company]. To get the right person to you: is anyone or anything trapped, or is the door stuck open? If anyone is in danger, call the emergency services first. Please do not try to force the door or touch the springs or cables; they are under tension. A technician will reply within [10] minutes.

Have your lead technician agree that wording, since it is a safety statement and yours to stand behind. It is fixed because nothing unsupervised should say anything new to a worried customer, and because a template is always available when the AI is slow or down.

The second message comes from the triage below. It uses the customer's own words, names the likely visit type without promising a cause, asks for the one or two missing details, and gives a time window. Whether a person or a script sends it, a person approves the first month's drafts. Your job software, phone app or form tool will normally have an automation that sends the first message; check which of your tools can do it before buying another.

Urgency triage on one page

Write the tiers down once and put the same page in front of the office and the AI. Tiers are about what the customer needs, not what you think is broken.

Two rules need settling before the table goes live. First, who is on call for Tier 1 outside working hours, and what the after-hours charge is, if you have one. You decide both, and the fixed first message should state the rule so nobody is surprised. Second, what Tier 2 means at 7pm. A first slot tomorrow morning, with the safety line repeated, is often a better answer than a rushed evening call-out for a door that is secure.

Now the AI step. It reads the enquiry, applies the table, cites the customer's exact words for the tier, and drafts the second message.

You triage garage door enquiries for a small repair company. Apply the tier table below.
For each enquiry output:
Tier (1-5) and the customer's exact words that justify it.
Job type, as the CUSTOMER describes it. Mark everything unconfirmed. Never state a cause as fact.
Door type, if given (sectional, one-piece, roller). Photos: yes/no.
Missing: the one or two details needed to book.
Duplicate check: compare with the recent leads I paste; say MATCH or NO MATCH and why.
Draft second message: friendly, plain, under 70 words, no price, no promised arrival time, no cause.
Tier table: [paste]
Recent leads: [paste last 14 days: date, name, phone, address, source]
Enquiry: [paste]

Try it on this enquiry: "Garage door won't go up this morning, heard a loud bang last night. Car is stuck inside, need to get to work. Can someone come today?" A good response is Tier 1, citing "car is stuck inside, need to get to work", job type "possible spring or cable fault (customer reports a loud bang; unconfirmed)", missing "address confirmation and whether the door is currently closed", and a second message that repeats the safety line and offers the first slot without promising a time. The fix to look for: if the draft says "your spring has snapped", delete it. The customer said there was a bang, not that a spring broke, and the technician decides what failed.

A quieter example: "Looking for a price to replace our double garage door with an insulated one. Photos attached. Not urgent, we're away until the 20th." The output should be Tier 4, citing "price to replace" and "not urgent", job type "replacement door quote (customer description)", missing "opening width and height, and whether the opener is staying", and a second message such as: "Hi [first name], thanks for the photos. To give you a useful price, could you tell us the opening width and height, and whether you want to keep your opener? We can offer a survey slot after the 20th, so please say which days suit." Check that the draft does not quote a price range. A good draft asks for dimensions and offers a survey, because a price from photos is a guess.

Cost per booked job: does a bought lead pay?

A booked job is an appointment on your diary with a date, time and address, confirmed by the customer. Cost per booked job is what you spent on a source, net of any credits, divided by those. Compare that against the gross profit a booked job brings you, and set a ceiling on the share you will spend. Mine starts at 30% of gross profit per booked job, because the rest has to pay for vans, office and insurance. Set yours from your own overheads.

Now the numbers, for an invented month at a firm whose average completed repair brings $240 of gross profit (a $400 ticket at 60%).

The Local Services row is 34 charged leads at $40, less $160 credited for four duplicate or invalid ones. Gross profit per booked job is completed jobs times $240, divided by booked jobs, which accounts for cancellations. At a 30% ceiling the Local Services source passes narrowly (28%) and the marketplace fails badly: it cost $190 to book a job worth $189 in gross profit, so the month ends $20 down on that source. The firm's total gross profit is $13,680 on $3,860 of lead spend, which looks healthy, and hides that one source added nothing.

Here is the test of whether speed can rescue it. Suppose faster replies lift the marketplace's bookings from 14 to 21 and completions from 11 to 15 (an assumption, not a promise). Cost per booked job falls to $127, but the ceiling is $51, so the source is still more than twice too dear. That is the honest finding for most weak sources: faster replies improve what you get from a lead, and rarely rescue one whose price is higher than the job pays. The levers left are to negotiate the price per lead, ask for exclusive leads, change the source, or stop. For comparing platforms on this measure, Angi leads for plumbers and HVAC with a worked cost per job uses the same arithmetic.

Duplicate leads: one customer, three sources

A customer with a broken door at 7am may call you, fill in a marketplace form and tap a Local Services listing within ten minutes. You then pay for three leads, phone three times and risk sending two technicians to one address. In the month above, 9 of the 70 marketplace leads matched an enquiry the firm already had, which is $342 of lead spend. The rule is simple: one customer, one record, one owner.

1. Normalise the phone number into its last ten digits so formats stop hiding matches. If your local numbers start with a 0 that disappears behind a country code, use the last nine digits instead (change the 10 in the formula below to 9).
2. Match within a window, say the last 14 days, on the phone number, then the email, then the address.
3. Merge, do not re-book. A second source adds a note to the first record. The second source's lead is still logged, so you can ask about credit.
4. Tell the customer once. If the second message has already gone, do not send another.

In a spreadsheet (Google Sheets or Excel) two formulas do most of it. In the sheet, column A holds the date and time received, column B the phone number as typed, and column C a cleaned version:

C2 (last ten digits of the phone number):
=RIGHT(SUBSTITUTE(SUBSTITUTE(SUBSTITUTE(SUBSTITUTE(B2," ",""),"-",""),"(",""),")",""),10)

D2 (possible duplicate within 14 days):
=IF(COUNTIFS(C:C,C2,A:A,">="&A2-14,A:A,"<"&A2)>0,"POSSIBLE DUPLICATE","")

Formulas cannot catch "J. Smith, 14 Oak Rd" against "John Smith, 14 Oak Road" when the phone numbers differ. That is where the AI step above earns its place: paste the last 14 days of leads with the new one and ask it to say MATCH or NO MATCH with the reason. Check any MATCH yourself before merging, because two neighbours with the same surname on the same street are not duplicates. If your source offers credits for duplicates, send the matched pair with the dates and times. Do not assume it will credit them; read the agreement first.

The lead log: nine columns that make every sum possible

Everything above depends on a log with one row per lead. A spreadsheet is enough. Fill it as things happen, not at the end of the week, and use lead IDs instead of names when you paste anything into an AI tool.

Cost per booked job, the reply-time bands and the duplicate cost are all summaries of this one sheet, so a fifteen-minute review each Friday is enough. Ask an AI to write the summary formulas or a pivot table if you like, then check one total by hand. Row L-0733 is the case that earns the sheet its keep: a second lead for a customer you were already helping, logged as a duplicate, merged into L-0731, and flagged so you remember to ask about credit.

Reply-time bands: measure your own speed effect

Once leads carry a source tag and a time received, add one column for the time of the first human contact, and sort each lead into a band. After 30 days you have your own numbers, which beat any statistic you could borrow. The table below splits the same 155 leads and 68 bookings as the month above, to show the layout, not to set a benchmark.

Read it with suspicion. Calls your office answers live are quick by definition and book at high rates, so speed and source are tangled together. Split the same table by source before you conclude anything, and look at marketplace leads on their own, where the first call wins. If the bands barely differ once you split by source, your reply speed is not your problem, and you should spend the effort on the price or quality of the lead instead.

What an automated reply must never promise

Three rules keep the instant reply from causing trouble. Do not state a price: a form guess is not a quote, and the customer will remember it. Do not state an arrival time unless a person has checked the van; "within [10] minutes" in the fixed message is a promise to reply, not to arrive. Do not name a cause as fact; "possible spring or cable fault" is a description, "your spring has snapped" is a diagnosis.

Texting rules vary, so ask your adviser what consent wording your form needs before you text anyone, and include an easy way to stop. If a customer asks whether they are talking to a person, say plainly whether the message is automated. For after-hours calls, an answering service can carry the same triage page; setting up an AI answering service that never misses an emergency covers that, and replying to trade enquiries in 60 seconds has more reply wording. The quote that follows triage is its own job; itemised spring and opener wording starts you off.

Three ways this goes wrong in a busy week

The loud-word trap. A customer writes "the door keeps banging when it closes". The AI, primed by earlier examples about loud bangs, tiers it as urgent and the office phones at 6am. The fix is to require the AI to quote the exact words behind each tier and to add three or four example enquiries with the correct tier to the prompt. If the quoted words do not describe a trapped vehicle or an unsecured property, the tier is wrong.

The two vans. A marketplace lead and a Local Services lead arrive 20 minutes apart for the same address. Different people in the office book each one. Two technicians arrive, and one visit, perhaps $140 of paid time, is wasted before the customer starts shopping around. The duplicate check has to run before anyone books, not after, and the lead owner has to be named on the record.

The promise in the template. Someone edits the fixed first message to say "we can be there within the hour". For three weeks it is wrong most evenings. Keep the fixed text under version control (one named owner, one saved copy) and review it monthly with the first week's replies.

Questions garage door firms ask about leads and AI replies

How fast should I reply to a garage door lead?

As fast as you can do reliably, which for most firms means an instant fixed acknowledgement and a human or AI-drafted follow-up inside ten minutes during working hours. Do not copy someone else's statistic. Split your own leads by reply time and compare booking rates, as the table further down shows, taking source into account.

Should the AI send replies on its own?

Let it send only a fixed, pre-approved first message. Have it draft the personalised second message for a person to approve, at least for the first month. Anything that mentions price, arrival time or the cause of a fault should be read by a person, because those are the sentences that cause disputes.

Can I get credit for duplicate leads?

It depends on the source, so read your own agreement. Google's help page for Local Services Ads says charged leads are reassessed and may be credited if they are poor quality, duplicate or invalid, with a 30-day review, but credits are not available in every category or place. Repeat contacts from the same number or email within 15 days are not charged again. For other sources, ask in writing before you assume.

What counts as a booked job?

Pick one definition and keep it: an appointment on the diary with a date, a time and an address, confirmed by the customer. A voicemail you left, or a quote you sent, is not a booking. Using the same definition for every source is what makes the cost-per-booked-job comparison fair.

Further reads

• Handyman Leads: Thumbtack, Angi and Bark Compared on Cost per Booked Job — Three marketplaces compared on cost per booked job.
• Pool Service Leads: Compare Paid Leads, Local Services Ads and Fast Replies — Paid leads, Local Services Ads and fast replies for another trade.
• Foundation Repair Leads: Screen Crack Photos With AI Before the Visit — A photo-and-questions screen for a higher-value enquiry.
• Insulation Leads: What They Cost and How AI Qualifies Them — Real price ranges for insulation leads by source, how to turn them into cost per booked survey, and an AI qualifying script that protects survey slots.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Google's Local Services help page on how leads and lead credits work, and Google Ads help on the move of Local Services Ads to Performance Max (both read October 2026). Lead sources and prices in the examples are illustrative; check your own invoices and agreements.