For a Quebec SMB, the marketing automation that actually pays comes down to four mechanisms: an immediate reply to a new lead, follow-up on quotes that went unanswered, a short email sequence for people who are not ready to buy, and a review request once a job is finished. Budget $0 to $150 a month in tools and two to three days of setup. Everything else — lead scoring, twelve-branch decision trees, multichannel journeys — was designed for companies receiving thousands of enquiries a month. When you receive forty, it costs more in upkeep than it returns.
This describes what works in companies of five to fifty employees, with one person handling marketing part time — often the owner. No $800-a-month platform, no six-month project. And one clarification that saves a lot of disappointment: automation does not create demand, it stops you losing it. It sits on top of acquisition that already works; it does not replace it.
Marketing automation is not a piece of software
It is a rule of the form when this happens, do that. When a form is submitted, send an email and create a task. When a quote goes four days without an answer, alert the salesperson. When a project is marked finished, ask for a review. Nothing more. The software is only where the rule is written down.
That distinction matters, because most automation projects fail by starting at the wrong end. You pick a platform, sit through the demo, discover sixty features, and try to use them all to justify the subscription. Six months later nobody knows what gets sent to whom any more. The right approach is the reverse: write the three or four rules on a sheet of paper, then look for the cheapest tool that can run them.
The four automations that pay, in order
One. The immediate reply to a new lead. The only automation whose effect is measurable in the first week. The most cited work on the subject shows that a lead contacted within five minutes qualifies far more often than one contacted after thirty — the gap is measured in multiples, not percentages. On the ground the mechanism is trivial: whoever calls back first is talking to somebody who is still looking. An automatic reply confirming receipt, giving a realistic timeframe and offering a callback window is enough to hold the position. The setup is detailed in a separate piece.
Two. Follow-up on unanswered quotes. The most profitable and the most neglected. In most service SMBs, a substantial share of quotes sent never receives a reply and is never chased — not by choice, but because nobody keeps the list. A rule that triggers a follow-up at four days, then a second at ten, recovers contracts already priced, already qualified, already visited. No acquisition costs less.
Three. The sequence for people who are not ready. Some of your leads have a project in eight months, not next week. Calling them is pointless, forgetting them equally so. Three or four spaced-out, genuinely useful emails with no sales push keep the name alive until the project becomes real again. It is the same mechanism as a newsletter, in short, triggered form.
Four. The review request after the job. Three days after the work finishes, an email with a direct link to your Google profile. Asked systematically, a review arrives; asked when you remember, it does not. And review volume remains one of the few levers that improves both your local visibility and your conversion rate with no ad budget — including when a bad one lands.
Automation wired to dirty data does not save time. It sends your mistakes out faster.
What it really costs
The four mechanisms above fit inside the free tier of several CRMs, or inside an entry plan at $20 to $60 per user per month. Add an email tool, often free under a few hundred contacts, and a connector between your forms and your CRM where the link is not native. An SMB receiving under a hundred leads a month stays below $150 monthly.
The real cost is elsewhere. It is the two or three days needed to write the messages, wire the forms, test every scenario with real addresses and fix what goes sideways. And the half-day per quarter it then takes to check nothing has broken — a modified form, a renamed field, an expired integration are each enough to silence a sequence without anybody noticing. An automation nobody inspects has always, eventually, stopped quietly.
What has to stay human
Four things cannot be automated without damage. Qualifying a complex project: two questions on the phone beat eight fields in a form, and cost fewer abandoned leads along the way. Replying to an unhappy customer: an automatic acknowledgement of a complaint makes the complaint worse. Price negotiation. And first contact when the average ticket is high — nobody commits forty thousand dollars of work to a sender who does not reply.
The line is easy to draw: automate what is repetitive and predictable, keep human whatever needs judgement. And when in doubt, remember that the most effective automation is often the one that creates a task for a person rather than an email for the customer.
It all rests on the CRM, which is why you choose that first
The four mechanisms need the same foundation: one single place where every lead exists, with its source, its stage and its last-contact date. Without that there is nothing to trigger — a rule cannot chase a quote the system does not know about. Which is why an automation project always starts with choosing the tool that will hold that list, not with choosing the sequences. The criteria that genuinely matter for an SMB here, data hosting included, are set out in our method for choosing a CRM as a Quebec SMB.
A warning on sequencing: do not automate anything before cleaning the list. Duplicates, wrong addresses, contacts nobody can trace — a sequence wired to that sends follow-ups to people who never asked for them, and you pay in unsubscribes and complaints. The tidying comes before the machinery.
How to tell whether it is working
Three numbers, taken before and after. Average time to first reply, which should fall from several hours to a few minutes. The share of quotes followed up, which should go from when we remember to all of them. And the lead-to-customer conversion rate, the only verdict that counts — if the first two move and the third does not, your problem is not your response time, it is your offer or your price. That is exactly the kind of judgement call the five numbers a business owner should track are there for.
What measures nothing: emails sent, active scenarios, open rate in isolation. A company can triple its sends and lose customers. If your leads arrive fast but go nowhere, the cause usually sits upstream of the automation — the usual reasons are few and fairly predictable.
The three mistakes that sink a project
First: building everything at once. Four scenarios launched the same day, and when something starts misfiring there is no telling which. One scenario at a time, two weeks of watching, then the next. Second: writing messages that smell of the machine. Dear FIRSTNAME, we acknowledge receipt of your request has never moved anybody closer to buying. Write the message the way you would say it on the phone, and sign it from a real person with a real number.
The third, more expensive since 2023: automating sends without valid consent. Quebec's Law 25 requires consent to be obtained for a specific purpose, and a quote request is not consent to receive a marketing sequence. The rule and how it applies to mailing lists are set out in our piece on consent and the newsletter. Worth settling before launch rather than after the first complaint.
If you want to know which of the four mechanisms is missing in your business and which would pay back fastest, book the diagnostic call: we look at your form, your response time and your list, and you leave with the order of work.
The sales funnel for a small business — which stage to automate first, and why.