Follow-Up That Runs Itself
Four things your job system already knows, or can be taught to know: a quote went out, a job finished, someone is running late, a client went quiet. Wire the follow-up to those four events and it runs without you.
By James Schramko · Updated September 2026
This is for businesses with staff in the field: cleaners, trades, installers, mobile services, anyone whose work happens at the client's address. It is written for owner-led operations, from two vans to a few dozen staff, where the follow-up still depends on somebody remembering. At some point the owner says the words "we need a CRM" in a meeting. What they usually mean is that the follow-up has been living in their head and their phone, the business has outgrown their memory, and quotes, reviews and past clients are slipping through the gap.
The instinct is to buy something big or build something custom. Both treat a trigger problem as a software problem. Follow-up fails at the moment nobody remembered to send the message, long before it fails at the tool. Your job system already records the moments that matter most. This playbook wires the messages to those moments.
Two Systems, Not One
You are describing two different systems, and they are different jobs.
One is the database: everyone who has ever bought from you or enquired, and it exists so you can go back to them. It is the most valuable thing the business owns, and it does not care what software it sits in as long as you can get it out. When I was General Manager of a Mercedes-Benz dealership, the dealer standards audit required the customer database printed in hard copy and locked in a fireproof safe on the premises. I would open the safe and show the auditor the printout, and six figures of dealer bonus rode on it being there. The manufacturer understood what the asset was.
The other is the jobs and scheduling system: who is at which property, when, and whether they are on time. Most service businesses try to make one tool do both, then spend six months and a lot of money discovering it does neither well.
Keep them separate in your head, even if one product ends up handling both. The database is what you protect. The jobs system is what you wire the follow-up to.
The Four Events
Four things happen in a job system that should each send a message on their own. Nobody has to remember to send anything. The event fires, the person moves onto the right track, the track sends.
A quote goes out. The job is completed. A cleaner or tradesperson is running late. A client has gone quiet.
Each event gets one short track, written once, from a named person at your business. That is the whole system for the follow-up that is leaking today. Your business will have other events worth a message, a deposit unpaid, a reschedule, a service due date. Add them one at a time once these four are running.
Track One: The Quote Went Out
Start here, because this is where the money is sitting. Count your unconverted quotes from the last ninety days before you read on.
Same day: confirm it arrived and ask one question. "Quote's with you. Anything you want changed before I pencil in a date?" A question gets a reply. A statement gets filed.
Day three or four: "Still want this done? I have two mornings open next week." Specific dates convert better than open offers.
Day ten: close the file and leave the door open. "I'll take this off my list for now. If the timing changes, reply to this and I'll pick it straight up." The closing message often gets the reply, because it is the first one that asks nothing of them.
Three messages over ten days for residential work, then it stops. Nobody gets chased for a month. Commercial and builder quotes run on the buyer's cycle, so set the clock to theirs, a month or a quarter, and keep the same three messages.
Track Two: The Job Was Done
Same day: "Done at your place today. If anything is not right, tell me and we fix it." For bond and end-of-lease work the promise covers the agent's inspection, so say so: if the agent finds anything, we come back. This message gives a complaint somewhere private to land before it becomes a one-star review.
Day three, or once the result has had time to show: ask for the review, with the link, one line.
For recurring clients, the track adds one more message: confirm the next visit the day before. That one message cuts the "nobody turned up" calls, because a good share of them are a client who forgot.
Track Three: Running Late
This is the one owners feel most, because it is the phone call they currently make themselves, from the car, between two other problems.
When the job before runs over, the system sends the next client a message before they notice. "Sam is running about thirty minutes behind and will be with you by ten thirty." Named person, new time, and a way to say the new time does not work. The client who is told before they notice feels looked after. The client who notices first feels ignored, and that is the same lateness.
This track depends on staff updating job status on their phones. If they do not, fix that first, because no message can go out off a status nobody sets. Most job systems already have an on-the-way or running-late message in the settings, often sitting there unused. Check before you buy anything.
Track Four: Gone Quiet
Every client base holds three groups that matter for follow-up: people who enquired and never booked, current clients, and past clients who drifted off and forgot you. The third group is where the cheapest revenue in the business sits, and it is the group that usually hears nothing.
Define quiet by the client's own cycle. A regular who has not booked in ninety days. A bond clean at the next lease cycle. A builder at the next project. For a client who left, the first message asks for advice before it asks for a booking: what would we need to do differently for you to use us again? Then the offer. For a service on a known cycle, a gutter clean or an annual treatment, just offer the date. The full win-back structure, including how to segment past clients by value, is in Essential Prospect Emails.
Past clients come back when there is still a line to you. I had a customer from 2010 reappear in 2026. Fifteen years. The line was still open.
Turn On Before You Build
Almost everything above already exists in software you can switch on this week for a few hundred dollars a month, and much of it is already inside the job system you run. Look there first.
Switching a track on is also the fastest audit of your job data. If jobs get marked complete before the work is checked, or quotes get duplicated, the messages will show it within a week, to your clients. So run each track on one crew or one suburb for two weeks before it goes to everyone.
Where the job system genuinely lacks a track, the smallest bridge is a light automation tool that watches the job system and sends the text or email, and somebody owns that bridge when either side changes. A custom build comes last, and only for the piece nothing else does.
A custom system has no one else fixing it. When it breaks, and it will, it breaks on a Tuesday morning with staff on site and nobody to call. Build only where you have looked and genuinely found nothing, and even then build the smallest piece.
Count Before You Decide
Before choosing a tool or a sending frequency, count four things: how many client records you hold, how many booked in the last ninety days, how many quotes are out and unconverted, and how many have lapsed. Add what an average job is worth, so you know what one recovered quote pays. For anything you already send, pull opens, clicks, unsubscribes and spam complaints, and above all replies and bookings.
Frequency is the second question. The first is who is on the list and why they are there. Send all three groups the same message at the same rate and one group says you contact them too much while another never hears the one thing that would bring them back. Your own reaction to emails you receive is a weak signal. You are not the person they are written for.
Who Owns It
Name the person who runs the tracks day to day, and make sure it is a name other than yours. You can own the result. You should not be the one who fires the messages or fixes the failed ones. If the owner of the business runs the follow-up system, it becomes one more thing the owner personally runs, which is the exact problem you were solving.
The thirty-day test is simple. Did any message go out only because you remembered? If yes, it is not wired yet. A message that stopped for a human decision, a disputed job or an unusually large quote, is the system working.
Keep The Record Where It Is
If your client records already live in one system, keep them there. The messages bolt on beside it. Do not migrate everything to get one message to send, and do not let a tool salesperson turn a follow-up question into a platform project.
If your records are split across spreadsheets, phones and an old package nobody trusts, the first job is one record in one place, and that is still a far smaller job than a platform. Wire the four events, count the results, and only then talk about what else you might need.
The Self-Diagnostic
How many quotes went out in the last ninety days, and how many did you follow up more than once. How many past clients have heard from you this year. When a cleaner runs late, who tells the client, and how do they find out. Which of the four events already sends a message without a human touching it.
If the answer to the last one is none, you have a jobs system and somebody's memory, and it is usually yours.
Quick Reference
- Two systems: the database is the asset you protect, the jobs system is what you wire the follow-up to.
- Four events, four short tracks: quote sent, job done, running late, gone quiet. The event fires, the track sends, nobody has to remember.
- Quote track first. Count the unconverted quotes before anything else. Three messages, ten days for residential, the buyer's cycle for commercial.
- Job-done track: same-day fix promise, review ask once the result shows, next-visit confirmation for regulars.
- Running late: named person, new time, a way to reply. Depends on staff setting job status. Usually already in your job system's settings.
- Gone quiet: define quiet by the client's cycle. Advice first for a client who left, the date first for a service on a known cycle.
- Turn on before you build, and pilot each track on one crew for two weeks. A custom system has nobody else fixing it.
- Count before you decide: records, active, unconverted quotes, lapsed, average job value, plus replies and bookings.
- The owner owns the result, somebody else runs the tracks. Thirty-day test: did any message go out only because you remembered?
Wiring the four events and deciding what gets switched on before anything gets built is the kind of pass I run with founders inside Mentor.