Security Alarm Leads: Which Sources Book Installs, and AI Follow-Up

Referrals, past customers and trade partners such as builders and estate agents usually turn into booked surveys and signed installs at the highest rate, your own Google Business Profile comes next, and bought lists or shared leads come last, though only your own counts can confirm that. AI helps by asking four qualifying questions within minutes and logging the source.

With security alarm leads, cost per lead flatters the weakest sources. A $28 shared lead that books one survey in eight costs $224 in cash for every survey, before your own time, while a referral costs a thank-you card. Rank sources by cost per booked survey, then by what those surveys turn into, and hand AI the repetitive part: asking the same questions in the same order, every time.

Follow me on Instagram@sagnikteaches

Count booked surveys, then work back to the source

Lead sellers and ad dashboards report leads, because leads are what they deliver. The number that fills a fitter's diary is the booked survey: a date, at a property, with someone who has said they want an alarm. The survey is the site visit where you count the openings, look at any old panel or wiring, agree what monitoring they want and leave with enough to write the quote. Everything before it is selling effort. Everything after it is a separate question about quoting and closing.

Connect on LinkedInSagnik Bhattacharya

Six columns do the work: date in, source, property type, existing system, monitoring wish and stage. These five rows come from an invented week.

Subscribe on YouTube@sagnikteaches

Label the source when the enquiry arrives, never afterwards. On the phone, ask who recommended you. On the web form, make "How did you hear about us?" a required dropdown that lists your real sources and nothing else. Treat a call from your Google Business Profile as its own source: the profile's performance report counts taps on the call button and clicks on the website link, but it cannot say which of them became surveys, so that log has to be yours. Stages stay simple: enquiry, qualified, survey booked, survey done, quote sent, installed, each with a date.

Free-text answers need tidying, and this is a good first job for a chat assistant. Paste the "how did you hear" answers with no names attached and ask it to map each one to your source list, flagging any it cannot place. Illustrative results:

Two rules stop the sheet turning into a mess. One person owns it. And a duplicate, such as someone who arrives from a shared lead and then phones you directly, is credited to the first contact that reached you, unless they name a referrer, in which case the referrer wins.

One month, seven sources, and what each booked survey cost

Here is a month for an invented installer with two fitters. Every enquiry costs about 12 minutes of someone's time to answer, qualify and book, which at $40 an hour is $8, so that goes in the cost column next to cash spend. Every figure in the table is illustrative; replace them with your own as soon as you have a month of data.

The sum behind each last cell is cash spend plus enquiries times $8, divided by booked surveys. For the shared leads that is $840 plus $240 of handling, which is $1,080, divided by four surveys: $270. Across the month the installer handled 173 enquiries, spent $2,860 and booked 47 surveys, a blended $90.30 per survey. The blend hides a spread from $21.71 to $596, and that spread is the finding.

Three things stand out. First, the three cheapest sources, referrals, partners and the profile, produced 27 of the 47 surveys for about $608 in cash and handling time, and none of them can be bought: you can only ask for referrals, earn partner trust and improve the profile. Second, the social ads and shared leads took 84 enquiries, which is 16.8 hours of calls at 12 minutes each, to book ten surveys. Third, one survey from seven replies to a bought list is too small a sample to condemn the list, but the seller's per-contact price never mentioned the $240 of printing and postage or the time spent answering replies.

What to do with a table like this: move effort before money. Add a line to every install handover asking who they know, and use the referral side of alarm installation marketing with contracts and referrals to set up a thank-you routine for customers and partners. For other trades the same exercise is worked in HVAC lead sources that book replacements, and solar lead generation by cost per survey shows the same survey-first arithmetic.

From booked survey to installed job

A booked survey is not an install, and sources differ in how many of their surveys end in a signed quote. For the sources where you have enough surveys to say anything, add three more columns: installs, close rate and the cost of each install, counting the $85 a fitter's visit costs. Here is the same invented month followed through to signed jobs:

Each cost is the source's cash spend plus handling plus $85 for every survey attended, divided by installs: for referrals, ($188 plus 8 times $85) over 6. By cost per booked survey the profile was cheapest and referrals close behind. Followed through to installs, referrals pull ahead because three surveys in four sign. The shared leads go the other way. A cost per survey that looked merely high becomes a loss on the first install once only one survey in four closes. Four surveys and one install is a thin sample, so read the pattern and not the decimals, and remember that this ignores monitoring income. It is also why the ceiling sum later should use the close rate you see for that kind of lead, not your overall average.

Reading the table without fooling yourself

• Wait for twenty enquiries per source. Four surveys from thirty shared leads and one from seven list replies are both small enough to flip next month. Judge a source on at least twenty enquiries or two months, whichever is longer.
• Group by the month the enquiry arrived. A survey booked in week one may not become an install until week five. Counting installs in the month they were signed makes this month's advertising look worse and last month's look better than it was.
• Count no-shows by source. A booked survey that nobody is home for costs a fitter's hour. Send a confirmation text the day before with the fitter's first name, the arrival window and what to have ready (clear access to the loft hatch, any old alarm paperwork), and record which sources no-show most.
• Compare value as well as volume. A property manager referring six flats is not the same as six single-house enquiries. Add a column for install value once quotes are accepted, then look at cost per dollar of gross profit, not just cost per survey.
• Treat the first month as a draft. Source labels will be wrong, some surveys will be missing and a few enquiries will be counted twice. The second month is when the pattern becomes believable.

Four questions that qualify an alarm enquiry

A survey booked on thin information wastes a fitter's trip: the wrong person is sent, the landlord's permission is missing, or it turns out the customer wanted a camera. Four questions fix most of that, and each one changes what you do next.

You do not need all four before booking. Two answers plus a willingness to have a surveyor visit is enough, and the other two can be collected on the doorstep. Where AI helps is in applying a simple rating to what came in, so the enquiries that need a call today are never buried under the ones that can wait. Four invented enquiries show the idea:

A first reply that asks for little and offers two slots

Speed matters because the enquirer has probably asked several installers, and the first one who makes booking easy often wins the survey. A reply that asks nothing, or asks eight things, loses to one that asks two things and offers two slots. This prompt makes the assistant do exactly that. Save it in a Project in ChatGPT or Claude, or a Gemini Gem (becoming a skill from November 2026), so the rules stay put between enquiries.

You write first replies for a small alarm installation firm. Write as the owner, in plain, friendly English, under 90 words, no emojis.

Rules:
- Never quote a price, a monitoring fee or a response time for the alarm itself.
- Never promise police response; monitoring options are explained at the survey.
- Ask only for the items marked UNKNOWN below.
- Offer exactly the two survey slots listed.
- Never ask for alarm codes, door codes, keyholder details or photos of a keypad.
- If the enquirer mentions a break-in, open with one calm sentence of sympathy.

Enquiry text (phone number and surname removed): [paste]
Property type: [house / flat / commercial / UNKNOWN]
Existing system: [none / old bells-only / wired panel / DIY app kit / UNKNOWN]
Monitoring wish: [none / app alerts / keyholder calls / professional monitoring / UNKNOWN]
Survey slots: [Tuesday 10-12] [Thursday 2-4]
Owner first name: [your name]

For an enquiry about a flat above a shop, where the property type was given but the other two items were unknown, the assistant returned this (illustrative):

Hi [first name], thanks for getting in touch about an alarm for your flat. To book the survey properly I only need two answers: is there an alarm in place now, even an old one that no longer works, and would you like alerts on your phone, a monitoring centre that rings a keyholder, or just a siren? I can visit on Tuesday between 10 and 12 or Thursday between 2 and 4. Which suits you better? [your name]

It is close, and there are two edits worth making every time. It lists three monitoring options in one breath, but your cheapest panel may not have an app, so strike anything you do not fit. And "a monitoring centre that rings a keyholder" describes how some services work, not necessarily how yours does, so use your own wording. Read the first twenty drafts before anything is sent; the mistakes repeat, and each one becomes a new rule in the prompt.

Follow-up for enquiries that go quiet

Most enquiries that never become surveys did not say no. They got busy, or someone else called first. A short, fixed sequence catches the middle group without nagging anyone, and a chat assistant can write each step in your voice once you give it two or three examples of how you sound.

Stop the moment they reply, say no or reach day fourteen, and never send more than four messages in total. Keep texts near 160 characters, because many providers split longer ones into several billed segments. Two samples written to that limit, again illustrative:

Day three: Hi [first name], it's [your name] from [company]. I've a survey slot free Friday morning if you'd still like the alarm looked at. If plans have changed, just say and I'll close this off.

Day fourteen: Hi [first name], I'll stop messaging after this one. If you want an alarm survey later, reply here and I'll find you a slot. [your name]

The difference between a weak nudge and a good one is specificity. Before: "Hi, just following up on your alarm enquiry. Are you still interested? Let us know!" After: "Hi [first name], it's [your name]. You asked about an alarm for the bungalow. A survey takes about an hour and I check every door, window and any old wiring, so the quote that follows is accurate. Friday 10 to 12 is free if that helps." The second version tells them what happens, how long it takes and what to say next. Paste the weak version into the assistant with the instruction "rewrite this to say what the survey involves and offer one slot" and you get the second version in seconds.

Once the wording is settled, a sheet and a simple automation can send the day-three and day-fourteen texts without you. Lead tracking from enquiry form to pipeline sheet covers that build. Run it by hand for the first month so you know what the replies look like, then add the quote stage using an alarm installer quote template by zones and monitoring.

Testing a lead seller before you give it a budget

A lead seller's pitch is a list: so many contacts, so many cents each. Your test is different: how many of them book a survey, and what is one worth to you? Ask five questions first. Is each lead sold once or to several installers? How was the enquiry created (a form the person filled in, a call, or a scraped list)? What evidence of consent does the seller hold for contacting that person by phone, text or letter? What happens to duplicates, wrong numbers and people who never asked? Can you buy 25 leads instead of 250? If the seller will not answer the first four, the fifth does not matter.

Then do the ceiling sum, with your own figures in place of these illustrative ones. Suppose the average install earns $950 in gross profit and four in ten surveys become installs (use the close rate you see for that kind of lead if you have one; the table above ran from 25% to 75%). A booked survey is then worth 0.40 times $950, which is $380 of expected gross profit. Subtract $85 for the fitter's time and travel to attend it, and $295 is the most you can spend to book that survey and still break even on the first install. Multiply by the seller's booking rate to get a ceiling per lead, take off $8 for your handling time, and round down:

A seller charging $28 breaks even at one in eight and loses money below that. The first batch tells you which side you are on. This sum ignores monitoring income, which can justify a higher ceiling if most of your installs take a monitoring contract, so enter your actual take-up rate rather than a hopeful one, and ignore anything the seller says about average results.

What stays out of the chatbot in a security business

An alarm company holds information that burglars would pay for. Keep alarm and door codes, keypad photos, keyholder names and numbers, floor plans with detector positions, and anything about when a property is empty out of every chat assistant. The assistant only needs a first name, the property type and the three answers.

A realistic slip: a customer replies to your text with "the old code was 4471, can you reset it when you come?" and the owner pastes the whole thread into a personal ChatGPT account to draft the answer. The code now sits in a third party's logs. The fix has three parts. Redact before pasting. Add "please don't send codes by message" to the enquiry form and the confirmation text. And check which plan you are on: on personal plans, use of your chats for training can be switched off in the privacy settings, while ChatGPT Business, Claude Team and Gemini in Workspace say they do not train on business content by default. Setting labels change, so look at yours rather than trusting this paragraph. The AI security checklist for email and files goes through the other settings worth five minutes.

A fifteen-minute Friday review that moves the budget

Update the stages, let the sheet total each source, then ask the assistant for commentary rather than arithmetic. Chat assistants slip on sums, so the totals come from sheet formulas and the assistant only reads them back.

Here are this month's totals per source, no names: [paste source, enquiries, booked surveys, cash spend, cost per booked survey].
Last month's cost per booked survey per source: [paste].
My ceiling per booked survey is $295.
Tell me: which sources are above the ceiling, which moved by more than 25% against last month, and which have fewer than 20 enquiries so far and are too small to judge. Use my figures as given; do not recalculate them.

A useful answer (illustrative) names the shared-lead and bought-list rows as above the ceiling or close to it, notes paid search up from $84 to $110, a rise of about 31%, and lists the bought list as under twenty enquiries, so it is a test and not a verdict. The rules that follow are yours: two months above the ceiling pauses a source, and a source at half the ceiling or less that you have fitter capacity for gets more budget. The assistant is reading numbers back to you. The decision stays with whoever owns the diary.

Questions installers ask about alarm enquiries

Should the first reply include a price?

Only if your price list supports a range you can stand behind, such as the starting price for a three-bedroom house with a basic panel. Otherwise explain that the survey is what makes the quote accurate, and say how long it takes. A number you cannot honour at the survey costs you more trust than a slower answer.

Should referrals get the same AI-drafted reply?

No. A referral arrives with a name attached, so the first contact should be a call or a personal text from you that mentions the person who sent them. Use AI to draft the confirmation message and the survey checklist afterwards, not to open a conversation that a friend has already started.

Do I need permission to text people who filled in a lead form?

Rules on calls, texts and emails to people who have not asked to hear from you differ widely. Make the form say how you will contact them, keep a record of what they agreed to, and ask a solicitor or the relevant regulator what applies to your sales calls and bought lists before you scale anything up.

What if the enquirer will not say what monitoring they want?

Treat 'not sure' as the normal answer. Book the survey anyway, and tell them you will explain the options on site: no monitoring, alerts to a phone, keyholder calls or a monitoring centre. Record 'unsure' in the sheet so you can see later whether those enquiries convert at a different rate.

Further reads

• Alarm Maintenance Contracts: Renewal Letters and Reminders With AI — Renewing customers are your cheapest source of referrals.
• Alarm Company Software: Monitoring, Service and Billing, and Where AI Helps — Check your software can store source, monitoring and billing together.
• Window Replacement Leads: Reply Fast and Book the Survey With AI — Another survey-led trade, with its own reply timings.
• Boiler Replacement Leads: Follow Up in Minutes Using AI — See how an installer follows up in minutes, not days.
• How to Get More Google Reviews With Automated Review Requests — Reviews are what make profile calls turn into surveys.
• How to Optimise Your Google Business Profile With AI — Make the profile the source that books the most surveys.
• AI Camera Software for CCTV Installers: Analytics You Can Sell Clients — Human detection, line crossing, counting and plate recognition compared for installers, with a 14-day false-alert test, privacy checks and margin sums.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Google Business Profile help pages on performance reporting, and the privacy and data-use pages of ChatGPT, Claude and Gemini. All figures in the tables are illustrative, not measured.