Marketing automation: flows that react to what people actually do

Marketing automation is a set of rules that send messages when something happens, instead of when someone presses send. A subscriber joins a list and receives a welcome sequence; a customer abandons a cart and receives a reminder; someone stops opening and enters a re-engagement path.

The distinction from a scheduled campaign is that automation is per-person and continuous. A campaign has one send time for everyone; a flow has a send time per subscriber, determined by when they entered it. That is why a well-built flow keeps producing revenue long after the person who built it stopped thinking about it.

It is also why automation failures are hard to see. A broken campaign is obvious — it did not go out. A broken flow simply stops enrolling people, and the absence of email is not something anyone notices on a dashboard.

Every flow is built from three primitives, and almost every automation platform uses the same three under different names. Understanding them separately makes it much easier to reason about why a flow did or did not fire.

The common design error is to encode a condition as a trigger. 'Send when someone is a VIP customer' is not an event — there is no moment at which it happens. Expressed correctly it is 'when an order is placed, if lifetime value now exceeds a threshold, then send'. Getting this wrong produces flows that either never fire or fire endlessly.

A small number of flows account for most of the value, and they are the ones tied to a moment of demonstrated intent. Building them in order of intent rather than in order of ambition is the fastest route to automation that pays for itself.

Everything below is triggered by something the recipient did. That is what makes them effective, and also what makes them defensible: a message that follows an action is expected, and expected mail generates far fewer complaints than broadcast mail.

The most common cause is that a flow enrols each person only once, forever. That is correct for a welcome sequence and wrong for almost everything else: a customer who abandons a cart in March and again in July should receive the reminder twice. When re-enrolment is unavailable or left switched off, a flow appears healthy while silently serving a shrinking population.

The second cause is a dependency on open tracking. Any branch reading 'if opened' has been unreliable since Apple's Mail Privacy Protection began pre-fetching tracking pixels, because a large share of recorded opens are machine-generated. Flows built on that condition now route people down paths they never earned.

The third is missing exit conditions. A cart reminder that does not check for a completed purchase will email someone who already bought — which is worse than sending nothing, because it demonstrates the sender is not paying attention.

Automated messages carry the same consent obligations as campaigns, and the automation being triggered by the recipient's own behaviour does not by itself constitute permission to market to them. Under the GDPR and India's DPDP Act the lawful basis has to exist independently of the trigger.

The practical rule is that transactional and marketing content should not be blended into one message. An order confirmation containing product recommendations is a marketing email wearing a receipt's clothing, and it is treated as such by both regulators and spam filters.

Suppression must also be honoured globally rather than per flow. An unsubscribe has to remove someone from every marketing automation immediately, not just the one that most recently emailed them.