Email marketing campaigns

A campaign is a project with an end date

A campaign is one job: a launch, a renewal push, a webinar fill, a seasonal sale. It has a named audience, a single primary action, a send window, and a stop condition. A weekly newsletter is a program. A welcome series is a flow. Mixing those words in a planning doc is how a 'campaign' runs for eleven weeks with no owner.

Write the job in one sentence before anyone opens a template. 'Get 200 trial starts from customers who used feature X but never invited a teammate' is a campaign. 'Engage the list' is a wish.

Anatomy that actually ships

Audience: a saved segment, not 'everyone who ever downloaded a PDF.' Offer: the thing they get if they click, which might be a demo, a SKU, or a PDF, but it is one thing. Creative: subject, preheader, from-name, body, and the one button. Tracking: UTM on links, a conversion event in analytics, and a suppression so buyers do not get the same pitch twice.

The from-name is part of the creative. People open mail from a person they recognize more often than from 'Team Acme.' Reply-to should land in a mailbox someone reads. A campaign that cannot receive a reply is a broadcast with extra steps.

• Subject and preheader as a pair, not two slogans
• One primary link; extra links compete
• A plain-text part for clients that strip HTML

Goals you can score without a dashboard novel

Open rate is a weak goal after Apple Mail Privacy Protection, because many opens are prefetch. Click-to-delivered and click-to-open still move. The goal that pays the bills is the downstream event: trial start, order, meeting booked, renewal started. If you cannot wire that event, you are decorating a send log.

Unsubscribe rate and spam-complaint rate are guardrails, not KPIs to maximize. A campaign that 'wins' on clicks and spikes complaints lost. Mailbox providers keep that memory on the domain, which is email-aisle reputation, not a creative award.

Testing without theater

A useful test changes one thing: subject, from-name, or the first screen of the body. Two simultaneous changes teach you nothing. You need enough volume that a 10% relative lift is not noise. A 400-person list cannot support a four-variant multivariate test. Send the control, send one challenger, wait for the conversion window, then pick.

Holdouts matter more than most A/B tools admit. If 10% of the segment never gets the campaign and still converts at the same rate, the email was not the cause. That is boring to report and worth more than a 'winner' badge in Mailchimp or Klaviyo.