Personalisation at scale is segmenting the list by what people actually do, then changing the content blocks shown to each segment. First-name tags are table stakes; the lift comes from showing different offers, copy and imagery to different groups, in the same campaign.
What personalisation really means
Personalisation is not Hi {{first_name}}. It is showing different content to different people based on what you know about them: recent activity, purchase history, location, interests, engagement level. The aim is relevance, not flattery.
Segments that move numbers
- Engagement tiers: highly engaged, occasionally engaged, dormant.
- Behaviour: clicked X category in the last 30 days.
- Lifecycle: new subscriber, regular, lapsing.
- Value: spend or lead value bands.
Dynamic content in practice
One campaign, multiple content blocks, swapped based on segment. The hero image, the headline, the offer and the CTA can all vary while the campaign is sent in a single workflow. AI makes this fast: each variant is drafted from the same brief.
What to avoid
- Using data the recipient does not know you have. It feels surveillance-y.
- Over-fragmenting into 30 segments you cannot maintain.
- Forgetting to QA every variant. One broken merge breaks trust.
Frequently asked questions
How many segments should I use?
Start with three to six. That captures most of the lift without making the workflow unmanageable.
Do I need a CDP for this?
No. Most of the value is in engagement and recency, which any decent ESP can model. A CDP helps when you need cross-channel orchestration.
Does AI help with segmentation?
Yes. It can cluster engagement patterns and suggest segments you would not have written rules for.