1. The ten stages every business runs
Every small service business runs the same ten stages, whether or not the owner has ever drawn them out.
| Stage | The question it answers | |
|---|---|---|
| 1 | Get found | Does anyone searching for this service find me? |
| 2 | Never miss | What happens when someone calls and I can't answer? |
| 3 | Find | Who could I be selling to that I don't know about? |
| 4 | Reach | How do I start the conversation, consistently? |
| 5 | Understand | What was actually said on that call three weeks ago? |
| 6 | File | Where does that end up, without me typing it? |
| 7 | Convert | How does a conversation become a signed number? |
| 8 | Deliver | How does the work get done and handed over? |
| 9 | Get paid | How does an invoice turn into money without chasing? |
| 10 | Grow | What brings a past customer back? |
The diagnostic is simple: score each stage, find the empty ones, rank them by what they cost. Most owners have built stage 8 and maybe stage 3 (the work, and the finding) and left both ends of the funnel wide open.
2. Finding customers you don't know about
The problem. The business had 26 prospects, all researched by hand over several evenings. At that rate the list would never get big enough to survive a normal rejection rate, and the owner was spending selling time on data entry.
First run. Nineteen searches across nine target industries within thirty miles.
What it does now. A program runs every Monday at 6am on Google's servers. It pulls back every business it finds with phone number, website, rating, review count and address, removes duplicates, scores each one, and files them.
Why it runs on a server instead of my laptop
A script on a laptop needs the laptop awake, unlocked and plugged in at 6am Monday. This runs whether the owner is asleep, on a roof, or on vacation. That distinction is the whole difference between automation and a chore with extra steps.
Why running it again does not create duplicates
Each business is keyed on its Google place ID, so a re-run updates a record instead of adding another. Without that, running weekly produces 288 records, then 576, then 864. With it the list stays clean forever and phone numbers stay current.
Why three fields are protected
Status, contact name and notes are never overwritten by a refresh. The moment an owner writes "spoke to Wade, call back after the 15th" and the system erases it next Monday, they stop trusting it. An untrusted system is worse than no system.
Why a bigger company scores lower
| Signal | Points | Reasoning |
|---|---|---|
| Has a phone number | 40 | You cannot sell to someone you can't reach. Everything else is secondary. |
| 5–150 reviews | 30 | Established enough to be real, small enough that the owner still answers the phone. |
| Over 150 reviews | 10 | Large operation. Procurement, gatekeepers, longer cycle. |
| Rating 4.0+ | 20 | Healthy business, likely has money to spend. |
| Has a website | 10 | Mild professionalism signal. |
A big successful company scores lower than a mid-sized one, which is the opposite of what most lead scoring does. A solo operator's scarcest resource is time, and a 400-review regional franchise will spend six months of it before saying no.
3. Missed calls
The problem. A missed call from a prospect isn't delayed, it's gone. Most people who call a contractor and reach voicemail call the next contractor.
Most owners already have a number on the truck, the sign and the Google profile. Telling them to replace it is how you lose the sale, and it's usually unnecessary: a phone number has two separate capabilities, voice and text, and different providers can handle each.
Four ways in, cheapest and least disruptive first:
- Turn on what they already pay for. RingCentral, Dialpad and Ooma often include missed-call auto-text and nobody switched it on. Check this first, every time. It costs them nothing and buys enormous credibility.
- Text-enable the existing number. Calls keep flowing to their carrier untouched; a texting platform adds SMS to the same number. Works on landlines, VoIP and toll-free.
- Conditional call forwarding. Most carriers can forward only unanswered calls to a capture line. Their number rings as normal; only the calls they were already losing get routed.
- A second number for marketing only. Old number untouched, and every call to the new one came from advertising, which finally measures the marketing.
Porting the number is the last resort, not the first. Never risk downtime on the line a business runs on when a cheaper option exists.
"You're keeping your number and your phone service. I'm not touching either. I'm adding something that catches the calls you're already losing."
Why the text matters more than the voicemail
A voicemail asks the customer to wait. A text arriving eight seconds later says you've been heard, I'll be with you, and it opens a channel they can reply to one-handed while standing on a job site. The gap between those two experiences is the gap between winning and losing the job.
Why a voicemail you read beats one you listen to
Listening requires finding ten quiet minutes. Reading takes four seconds between jobs. The owner triages in real time instead of at 9pm.
The constraint nobody warns you about. US carriers require every business sending automated texts to register through a program called A2P 10DLC, roughly $19.50 one-time plus a few dollars a month. Approval takes 10 to 15 days, it applies to every provider, and it registers a legal entity rather than a number, so a client with two companies needs two registrations and two waits. Voice works the day the number is bought; texting doesn't. I start that paperwork first.
4. When software you paid for doesn't work
The phone platform was supposed to push each finished call to my code. Everything was configured correctly: webhook enabled, URL verified character by character, a current event selected. Nothing was ever delivered, and their own log returned an error.
I ruled out trial limitations, then a blocked destination, then event misconfiguration, then their own suggestion that it was limited to their legacy event system. Their support confirmed the fault was theirs.
Then the important part: it didn't stop anything. Their other interface was healthy, so instead of waiting to be pushed, the system pulls. It checks for new calls every ten minutes, collects each one with its recording, speaker-separated transcript and summary, matches the caller against the prospect list, and files it.
The trade is latency: minutes instead of instant. For reading call notes between jobs, invisible.
It also produced a better result than the original plan. The polling logic is code the owner controls, not a setting in someone else's product, so if the phone platform is ever replaced, the destination doesn't change, only the source.
5. How I keep the data safe, and how I proved it
The addresses those calls arrive on are public URLs. Without protection, anyone who discovered them could post fake calls into the system. So every request is checked against a cryptographic signature only the phone provider can produce, and if the signing key is missing the endpoint refuses the request rather than accepting it. Failing closed, not open.
An unauthenticated public webhook that trusts whatever it receives is one of the more common ways small systems get abused.
I posted a forged call to all four endpoints from outside the system, with no valid signature. All four returned 403 Forbidden. Nothing was written to the database.
That distinction matters more than it looks. It's a demonstrated fact rather than a claim in a document: the difference between "the system validates requests" and "I attacked my own system and it held." The second sentence can only be said honestly if someone actually ran the test.
This is the standard the whole method is held to: every security control gets attacked once before it is described to a client as working.
6. What it costs to run
| Item | Cost |
|---|---|
| Cloud functions, database, scheduling | $0 at this volume, inside the free tier |
| Business search API | $0, 76 of 5,000 free monthly calls |
| Phone platform | ~$23/month annual, $33 monthly |
| Additional number | ~$5/month |
| Carrier registration for texting | $19.50 once, then $1.50–3/month |
The automation infrastructure is effectively free. The phone platform is the entire bill.
Worth saying plainly, because a prospect will assume the software is the expensive part. The comparison they'll make is against a $300-a-month all-in-one platform, and the honest answer is that the platform is easier and this is cheaper and theirs.
7. What I have not solved
Naming the gaps is part of the method. A consultant who claims everything is solved is selling something.
| Stage | Status | Detail |
|---|---|---|
| 1. Get found | Unsolved | Profile exists but doesn't surface in local search. |
| 4. Reach | Half built | Calling works. Sequences and follow-up automation don't. |
| 7. Convert | Not built | Quote, proposal, e-signature, deposit. |
| 9. Get paid | Not built | Invoicing and reminders. |
| 10. Grow | Not built | Review requests. |
The destination is vendor-neutral on purpose. Calls are written to a database the owner controls, not into the phone vendor's product. Whatever produces the call feeds the same place through a small adapter. That's what made the webhook failure survivable, and it's the argument that lets a client keep the phone system they already pay for.
8. What you get for $2,500
The AI Opportunity Snapshot is $2,500 and takes one day.
- Walk the ten stages with the owner. Score each one. Ask what happens today, not what's supposed to happen.
- Find the three that cost the most money. Usually 2, 9 and 10.
- Quantify one of them out loud. "How many calls a week do you miss? What's an average job worth? That's your number." The arithmetic is more persuasive than any slide.
- Deliver the map, the three gaps, and a quote for closing them.
Why this works. The owner already knows something is broken. They can't name it, and they've never seen it laid out. The map turns a vague feeling of drowning into three specific problems with prices attached.