Email automation: raising ROI without annoying anyone

Email is the highest-return channel we manage and the one clients most often break by automating too much of it. The failure mode is always the same: a stack of sequences built in month one, nobody maintaining them by month six, and a list that has quietly stopped opening anything.

Here is the smaller version that works.

Build four sequences, then stop

Welcome. Triggered when someone joins the list. Three messages over ten days. The first arrives within minutes and delivers whatever was promised. The second, three days later, answers the question that person is most likely to have next. The third introduces what you sell, once, plainly. Across our clients this sequence typically produces more revenue than every broadcast campaign combined, because attention is highest at the moment of subscribing.

Abandoned enquiry or abandoned cart. Triggered when someone starts and does not finish. One message an hour later, one the next day. For services this is a quote request that was never returned; for retail, a cart. Do not send four. The second message recovers most of what the fourth would, and the fourth is what gets you marked as spam.

Post-purchase or post-project. Triggered on completion. Useful information first, how to use it, what to expect, who to contact, then a review request, then a relevant next offer several weeks later. This is where repeat business comes from, and it is the sequence most businesses skip entirely.

Re-engagement. Triggered by 120 days with no opens or clicks. Two messages, then remove the contact from the active list. This one feels like throwing money away and is the reason the other three keep landing in inboxes.

Deliverability is the ceiling on everything

The major mailbox providers tightened bulk sender requirements substantially, and the practical effect is that a technically sloppy sender no longer reaches the inbox at all.

The minimum: SPF, DKIM and DMARC configured for your sending domain, one-click unsubscribe honoured in the header, and a spam complaint rate kept below one in a thousand. Send from a subdomain you control rather than a free mailbox address. Warm a new sending domain gradually rather than mailing forty thousand people on day one.

Then the ongoing part: remove hard bounces immediately, remove people who have not engaged in four months, and never import a list you did not collect. A bought list does not merely underperform. It damages the reputation of the domain you send everything else from, including your invoices.

Timing that respects the reader

Transactional and triggered messages go immediately. Marketing sends go on a predictable day, so people know what to expect. No more than one marketing message a week for most B2B clients, no more than two for retail with an active catalogue.

We do not use send-time optimisation for small lists. There is not enough data for it to learn anything, and it makes reporting harder to read.

Frequency complaints are relevance complaints

When unsubscribes rise, the instinct is to send less. Usually the problem is that the wrong people are receiving the message.

Segment on behaviour rather than demographics: what they looked at, what they bought, how recently they engaged. Three segments handled well beats fifteen maintained badly. One client cut broadcast volume by a third, split the list into bought before and has not bought, and revenue per send rose while unsubscribes halved.

Measure what still works

Open rate has been unreliable since mail privacy protection started pre-loading images, and it gets worse every year. Treat it as a weak directional signal, not a metric.

Report clicks, conversions attributed to the send, revenue per recipient, list growth net of unsubscribes, and complaint rate. Revenue per recipient is the number telling you whether the list is an asset or a liability.

The consent question

Collect consent explicitly, record when and how you got it, and keep that record. This matters legally in the EU and UK, increasingly in India under the DPDP framework, and practically everywhere, because unconsented lists complain.

The practical version: an unticked box with plain language, a separate checkbox for marketing distinct from service messages, and a preferences page that lets people receive less rather than nothing. The preferences page is the cheapest unsubscribe reducer we know of and almost nobody builds one.

What we do not build

Lead scoring for lists under about ten thousand contacts: it produces a number nobody trusts. Nine-message nurture tracks, because attention does not survive them. Personalisation tokens beyond a first name, which reliably break in public and embarrass everyone. And birthday emails for B2B, which have never once produced a measurable result in any account we have run.

A realistic build order

Week one: authentication, list clean-up, unsubscribe handling. Week two: the welcome sequence. Week three: abandoned enquiry. Week four: measurement, then let it run for a month before adding anything. Post-purchase and re-engagement in month two.

That is roughly fifteen hours of work in total, and in most of our accounts it recovers its cost within the first quarter. The pre-send rules that keep each message out of trouble are in our twelve email rules.

When to get help

The authentication step is the one worth outsourcing if DNS is not something you touch often, because a DMARC record with a wrong policy can stop your ordinary business mail as well as your campaigns. After that, most teams can run this themselves. We build the four sequences, hand over a documented account and leave, or keep running it as part of email marketing when there is nobody in-house to own it. If email is going in alongside search and paid, sequence the whole thing using the full-funnel order.