Email personalization: beyond "Hi {First Name}"

Email personalization is the practice of tailoring content, offers, imagery, and send time to each subscriber using the data you already hold: past purchases, browsing behavior, lifecycle stage, and engagement history. Done well it moves real money. Done as a {{first_name}} merge tag, it moves almost nothing.

The gap between what personalization promises and what most senders actually ship is wide, and that gap is what this article is about. McKinsey's Next in Personalization research found 71% of consumers now expect personalized experiences and 76% get frustrated when they're missing. Meanwhile the most common tactic in the wild, a first name in the subject line, has been pattern-matched by subscribers and shows little measurable lift on mature lists.

What email personalization actually means

Personalization is any change to the content, timing, or targeting of an email driven by data unique to the recipient. That can be a product recommendation block or a fully dynamic template where every subscriber sees different copy.

Vendor marketing counts any merge field as personalization; practitioners mean something narrower. A working definition needs three legs. You have to know who the person is, you have to know what they've done, and they have to receive something a different subscriber wouldn't. Drop any leg and you're doing segmentation with extra steps.

Segmentation is the upstream work that makes personalization tractable, and the two operate at different levels: segmentation decides which group gets which email, personalization decides what varies inside it. Good programs run both. For the data prep side, see email list segmentation.

Merge tags are table stakes

Merge tags felt like personalization because they used to be novel. In 2026 they're the floor. The classic subject-line first-name play shows a small lift on cold lists that fades as the list matures, and on established lists it frequently shows nothing at all. Subscribers learned the trick.

The aggregate numbers that circulate about name personalization hide their own distribution. Cold lists see a novelty bump; mature lists see noise. Averaging those two together produces a figure that describes neither.

Here's the framing I keep coming back to: merge tags personalize the envelope, behavioral data personalizes the letter. Both are worth doing; only one of them changes revenue. If you want patterns that go past the name field, the personalization subject line collection is organized by what the line actually references.

Behavior-based personalization

Behavior-based personalization uses what a subscriber did (browsed, bought, clicked, ignored) to decide what to send, when, and what goes inside it. It's the highest-return form of personalization because it runs on data the subscriber actively generated rather than data they passively disclosed.

You need less of it than most vendors imply. Signup date, purchase history, and email engagement cover the three most useful signals; on-site browsing is a strong addition and not a prerequisite.

Browsing behavior

A subscriber viewed a product, a category, or an article and didn't act. That's a signal with a short shelf life. A browse-abandon email sent within 24 hours of the session consistently outperforms your standard broadcast on both engagement and revenue per send, because the intent is still warm. Salesforce's State of Marketing reports that high-performing marketing teams are 1.6x more likely to use behavioral data across channels than underperformers.

Purchase history

Past purchases tell you product affinity, price sensitivity, category interest, and replenishment cycle. A skincare brand that knows a customer bought a 30-day serum 28 days ago has everything it needs for a replenishment reminder. A SaaS company that knows a customer sits on the team plan can suppress team-plan upsells and show the enterprise path instead. None of this is exotic; it's using data already sitting in the store.

Email engagement

Who opens, who clicks, and who has done neither in 90 days. Engagement data drives the most consequential personalization decision there is, which is whether to send at all. Continuing to mail dormant addresses damages deliverability, because inbox providers read persistent non-engagement as a spam signal. Watch your churn rate alongside campaign metrics; a list that engages beautifully while shrinking every month is a problem you'll feel two quarters from now.

Lifecycle-stage personalization

Lifecycle personalization adapts content to where the subscriber sits in the relationship: new subscriber, active customer, at-risk, churned, win-back candidate. It's coarser than behavioral targeting and it's usually the easier place to start, since the stages are obvious and the rules stay stable.

The rank order holds across every program I've looked at: VIP and new-subscriber streams produce the most revenue per recipient, and churned streams the least. So build those two first, and accept that the win-back series is mostly a list-hygiene exercise wearing a revenue costume. The welcome swipe file has the new-subscriber structure, and re-engagement subject lines cover the other end.

Dynamic content blocks

Dynamic content blocks let you build one template with conditional variants, so a single broadcast delivers different content to different subscribers. Instead of building five emails for five segments, you build one email with five blocks and let the platform pick per recipient. Of every technique here, this one has the best effort-to-impact ratio.

A retail example: one Black Friday send, one dynamic hero block. Women's apparel browsers see a women's hero, men's apparel browsers see a men's, subscribers with no browsing history see a best-seller, and VIPs see early access with a unique code. One template, one send time, four experiences.

The rules build on whatever data you have: tag membership, custom fields, purchase history, engagement score, even live inputs like local inventory. For the underlying concept, see the glossary on dynamic content and personalization.

One caveat worth stating plainly: dynamic content only works if the data is clean. Stale tags, gappy purchase history, or a frozen engagement score will surface the wrong variant to the wrong person, which is a worse outcome than the generic email you were trying to beat. Clean the data first, turn on the rules second.

Using AI for personalization at scale

AI extends personalization past what rule-based systems handle: copy variants per segment, predicted send times per subscriber, engagement scoring per message, product recommendations per recipient. The bottleneck used to be the marketer's time; now it's the marketer's data.

Three moves carry most of the return.

Subject-line optimization per segment comes first. Feed the model your top performers plus segment context and get variants tuned to each group's language. This beats universal subject-line testing because different segments respond to different hooks; our subject line guide covers the craft underneath.

Send-time optimization per subscriber comes second. Rather than picking 10 a.m. Tuesday for everyone, the model picks the hour each subscriber has historically been most likely to open. It needs real history to work, so it's close to useless on a new list.

Copy generation for variant testing comes third. Generating five versions of a hero paragraph and testing them costs minutes instead of an afternoon; how to use AI for email writing has the workflow.

AI personalization works as a drafting and targeting assistant. The teams that get flat results are almost always the ones auto-sending whatever the model produced, without anyone reading it.

The three ways personalization fails

Failures cluster into creepy, irrelevant, and broken. Each destroys trust faster than generic broadcasting would have, which is why a little personalization done badly costs more than none done well.

Creepy personalization uses data the subscriber didn't realize you had. Referencing their exact browsing session ("Still thinking about the 14-inch MacBook Pro you looked at on Tuesday at 9:47pm?") crosses the line; referencing the category ("Deals on laptops this week") doesn't. McKinsey's research locates that line at the moment a subscriber can no longer reconstruct why they received the message. Apply the reconstruction test: if they can trace it back to something they knowingly shared, you're fine. If the data path involves location tracking, cross-device stitching, or third-party enrichment, back off; the lift isn't worth the trust.

Irrelevant personalization comes from stale or wrong data. Recommending the product they bought last month. Sending a we-miss-you email to someone who purchased yesterday because the tracker missed it. Naming a company they left six months ago. Each of these is worse than a generic broadcast, because it proves nobody is paying attention.

Broken merge tags are the most embarrassing category: Hi {{first_name}}, Your {{company}} account, Dear Customer Name. Every senior email operator has a story about one that got through. The fix isn't more careful writers; it's fallback values on every merge field and a seed-list test before every broadcast.

And an honest limitation on the whole discipline: personalization raises the cost of every campaign. Cleaner data, more QA, more template logic, more segments to maintain. On a 500-contact list that setup cost rarely pays back. On a 50,000-contact list it's a rounding error against the lift. Personalize when your list size and data hygiene make it economic, not because a deck said to.