Automation · Practical Guide
Most automation advice assumes a marketing department. Here are the five sequences that actually earn their keep for one person, in the order worth building them, and the ones to skip entirely.
The promise of marketing automation is that you build it once and it works forever. The reality is that you build it once, it works for four months, then a product name changes and for eleven weeks you send everyone an email referring to something that no longer exists, and you find out when a customer replies to ask what you’re talking about.
I mention this at the top because the failure mode for solo operators isn’t building too little automation. It’s building too much of it and then not maintaining any of it. An automation is a piece of infrastructure that runs unsupervised and speaks to customers in your name. Every one you add is a thing that can silently break.
So this is a deliberately short list. Five sequences, ordered by return on effort, with an honest account of what each costs to maintain.
The Principle That Prevents Most Disasters Read before building anything
Automate a process that already works manually. Not one you think should work. One you have actually performed, by hand, enough times to know what people ask, where they get confused, and what makes them buy.
Automation amplifies whatever process you feed it. If your manual follow-up converts poorly, the automated version converts poorly at scale and you’ve spent a weekend building it. If a manual email gets replies you find useful, automating it multiplies something valuable.
The corollary: send the sequence to yourself first, on a real timeline, using a real test account that goes through the real signup. Not a preview. Preview mode hides broken merge tags, wrong links, delays that fire at 3am, and the discovery that step four references something step two never actually sent.
One: The Welcome Sequence Build first · Highest return, always
If you build exactly one automation, build this. Welcome emails consistently outperform everything else in a programme because you are writing to someone at the single moment of peak interest — they just chose to hear from you, deliberately, seconds ago.
The shape that works, over four or five emails and roughly ten days:
Immediately: deliver whatever they signed up for. If they were promised a guide, the guide is here, in this email, above the fold. Nothing else. Do not sell.
Day two: the origin story, briefly. Why this exists, what problem you were solving. People buy from businesses they have a mental model of, and this is where you install it.
Day four: your single most useful piece of content, given free with no ask. This email establishes that opening your emails is worth the time, which determines whether they open the next hundred.
Day seven: social proof or a case study. What changed for someone like them.
Day ten: the offer, stated plainly, with an easy no.
Two details that matter more than the copy. Send from a person’s name at a real, monitored address rather than noreply@ — replies to a welcome sequence are among the highest-quality leads you will ever receive, and a noreply address discards them. And ask a question in email one that invites a reply. Replies are a powerful engagement signal to mailbox providers, and a subscriber who has replied to you once behaves differently forever after.
Maintenance cost: low. Review quarterly. The main risk is links to content you’ve since moved or deleted.
Two: The Abandonment Sequence E-commerce only · Highest revenue
Someone put something in a basket and left. Three emails, and the timing is the whole thing.
One hour later: a genuinely helpful nudge. Assume a technical problem or an interruption rather than hesitation. “Did something go wrong at checkout?” outperforms “You left something behind” because it offers help rather than reminding them of a decision they consciously deferred.
Twenty-four hours: address the objection. Shipping cost, returns policy, sizing, security. Whatever your actual reason-for-abandonment data says, if you have it. Ask customers if you don’t.
Seventy-two hours: the last one. A discount here is optional and expensive — you will give money to people who were going to buy anyway. If you use one, make it modest and time-bound, and be aware you are training a segment of customers to abandon deliberately.
Stop the sequence the instant they purchase. This sounds obvious and is the most common automation bug in existence, because the exit condition often gets configured on a different trigger than the entry condition. Test it with a real purchase.
Maintenance cost: moderate. Product feeds, images and stock status all change. A recovery email showing an item that’s now out of stock is worse than no email.
Three: The Post-Purchase Sequence Chronically neglected · Compounds
Everybody automates acquisition and almost nobody automates what happens after the sale, which is backwards. Existing customers are the cheapest revenue available and the least contested.
Immediately: confirmation, expectations, what happens next and when.
On delivery or access: how to actually get value from the thing. The single largest driver of refunds and churn is people who bought something and never worked out how to use it.
Day seven to ten: a check-in that asks a real question and accepts a real reply. This is your best source of product feedback and testimonials, and it costs nothing.
Day thirty: the review request, or the logical next purchase, depending on what you sell.
Maintenance cost: low, and the return builds rather than decays. A good post-purchase sequence raises lifetime value permanently, which changes what you can afford to pay for acquisition — which is the actual unlock.
“Everybody automates the sale. Almost nobody automates what happens after it, which is where the compounding actually lives.”
Four: Re-engagement and Sunset Boring · Protects everything else
Not glamorous, and arguably the most important on this list, because it protects your deliverability.
Contacts who haven’t opened anything in six months are doing three things to you simultaneously: costing money on contact-based billing, depressing your engagement rates, and increasing the odds that mailbox providers start routing you to spam. With bulk sender rules now enforced at a 0.3% complaint threshold, an unengaged segment isn’t merely dead weight — it’s a liability.
The sequence: two emails to anyone silent for six months. The first asks directly whether they still want to hear from you, with a clear yes button. The second, a week later, states plainly that this is the last one unless they act.
Then actually remove them. This is the part people flinch at. Watching a list shrink feels like losing something. You are not losing anything — a subscriber who hasn’t opened in six months was never going to buy, and their presence is actively degrading delivery to the people who would.
Run it quarterly. The first time hurts. After that it’s routine, and your open rates and inbox placement will both be visibly better within two sends.
Maintenance cost: essentially nil. Set it and let it run.
Five: The Behavioural Trigger Build last · Highest skill required
One automation that fires on a meaningful action: viewed a pricing page three times without buying, downloaded a specific resource, used a feature that signals readiness, visited a particular category repeatedly.
Build exactly one of these. The temptation is to build fifteen, at which point you have an unmaintainable web of overlapping triggers, customers receiving four emails a day, and no ability to work out which one caused the unsubscribe.
Pick the single behaviour that most reliably precedes a purchase in your business — you can usually identify it by looking at what your last twenty customers did — and automate the follow-up you would have sent manually.
Maintenance cost: high. Behavioural triggers depend on tracking, and tracking breaks. Every site redesign, tag manager change and platform update is a chance for the trigger to stop firing, silently, with no error message. Set a calendar reminder to verify it fires, quarterly.
The Build Order
| Sequence | Emails | Build time | Maintenance | Why this position |
|---|---|---|---|---|
| Welcome | 4–5 | A weekend | Low | Peak interest; sets every downstream engagement rate |
| Abandonment | 3 | Half a day | Moderate | Most direct revenue, if you sell online |
| Post-purchase | 4 | A day | Low | Raises lifetime value, which changes acquisition maths |
| Re-engagement | 2 | An hour | Negligible | Protects deliverability for everything above |
| Behavioural trigger | 1–2 | A day, plus tracking | High | Real return, real fragility. One only |
What You Actually Need to Run This Less tooling than you’d think
The automation software market wants to sell a solo operator a platform. In practice the requirements are modest, and the tool matters far less than whether the sequences are any good.
An email platform with visual automation. Everything in this article is achievable on entry or mid tiers of the mainstream platforms. The specific thing to verify before subscribing is whether the tier you’re buying supports the trigger types you need — behaviour triggers and conditional branching are commonly gated above the entry plan, and discovering that after you’ve committed annually is a familiar frustration.
Working conversion and purchase tracking. Without it, abandonment and post-purchase sequences have nothing to fire on, and you cannot tell whether any of this earns its keep.
A single test account with a real, monitored email address. Not an alias you never check. You will use it constantly.
A written record of what exists. One page listing every live automation, what triggers it, what exits it, and when you last verified it. This sounds bureaucratic for one person and it’s the difference between a system you control and a collection of things you half-remember building.
What you don’t need: a customer data platform, a workflow orchestration tool, a separate personalisation engine, or anything described as a “growth stack.” These solve coordination problems that arise when multiple people and multiple channels must stay consistent. Solo, you are the coordination layer.
The First Ninety Days A realistic schedule
Building all five sequences at once is how people end up with five half-tested automations and no idea which is responsible for the unsubscribe spike. Spread it.
Weeks one to three: welcome sequence only. Write it, test it end to end on a real signup, launch it, and then leave it alone long enough to see a full cohort pass through. Watch per-email unsubscribes rather than just the aggregate.
Weeks four to six: the revenue sequence for your model. Abandonment if you sell online, post-purchase if you don’t. Same discipline: build, test with a real transaction, verify the exit condition, launch, observe.
Weeks seven to nine: re-engagement. Quick to build, and the first run is where you’ll remove a meaningful chunk of your list. Do it deliberately, and record your open rate before and after so you can see what it bought you.
Weeks ten to twelve: the remaining sequence, plus your first full maintenance pass. By now you have three or four things running and enough elapsed time for something to have quietly broken. Find out what.
The pacing feels slow when you’re keen. It’s the difference between infrastructure and debt.
Measuring Whether Any of It Works Three numbers
Automation reporting defaults to vanity metrics — emails sent, total opens — which tell you the machine is running rather than whether it’s useful.
Revenue per subscriber entering the sequence. Not per email, not per open. Total revenue attributable to the sequence divided by the number of people who entered it. This is the number that tells you whether the sequence is worth its maintenance, and it’s directly comparable across sequences.
Per-email unsubscribe rate. Aggregate sequence performance hides the specific email that’s costing you subscribers. One email in five driving most of the attrition is a fixable, specific problem, and you cannot see it in a summary.
Completion rate. What proportion of entrants reach the final email. A low completion rate with a low unsubscribe rate usually means people are exiting on a condition you set and forgot — which is either working correctly or silently truncating your sequence, and it’s worth knowing which.
What to Skip Genuinely, don’t build these
Lead scoring. Assigning point values to behaviours to rank prospects. Valuable when a sales team must triage hundreds of leads. Meaningless for one person who can read a list of forty names. You will spend a weekend building a scoring model and then act on intuition anyway.
Multi-branch decision trees. Every branch multiplies the states you must test and maintain. A three-branch, four-stage tree has more paths than you will ever verify. Linear sequences with clear exit conditions do most of the work with a fraction of the fragility.
Cross-channel orchestration. Email plus SMS plus push plus retargeting, coordinated. It’s a genuine capability sold to teams with people to run it. Solo, it’s a way to build four things that each break independently.
Birthday and anniversary emails. The data is usually wrong, the sentiment reads as automated, and the return is close to zero unless you’re a restaurant.
Anything requiring data you don’t reliably have. Personalisation is excellent when the data is right and actively damaging when it isn’t. “Hi {{first_name}},” rendered literally, arriving in someone’s inbox, does more harm than the personalisation was ever going to do good. If your data quality is uncertain, use a fallback value everywhere and check it works.
Maintenance: The Part Nobody Writes About
Everything above assumes you will maintain it, and the honest truth is that most solo operators don’t, because automations are invisible when working and only surface when a customer complains.
The minimum viable discipline, quarterly, taking about an hour:
Enrol yourself in every live sequence using a fresh email address and go through the actual signup path. Read every email as it arrives, on a phone.
Click every link. Not a sample. Every one. Links to moved pages are the single most common decay.
Check the exit conditions still work. Make a test purchase and confirm the abandonment sequence stops.
Look at per-email performance, not just sequence totals. One email in a five-email welcome sequence driving a spike in unsubscribes is a specific fixable problem hidden inside an acceptable aggregate.
Read the copy for things that have changed. Prices, product names, policies, people who no longer work with you, offers that expired.
Sequence structures and timings reflect widely-observed practice rather than any single published dataset; test against your own audience. Deliverability considerations reference the bulk sender requirements enforced by major mailbox providers. This article contains no affiliate links.