Showing each person the version of the ad that interests them most is the promise of dynamic personalisation. The limit lies in perception: when the message reveals too much of what the company knows, it stops feeling relevant and starts feeling like surveillance. It pays to decide how far to go.
The essentials in a definition
Dynamic ad personalisation is the automatic adaptation of messages, images or offers to the context, behaviour or profile of the viewer. It works best when useful and transparent. The GDPR and the ePrivacy Directive require a legal basis and, in many cases, consent, and EDPB guidelines help interpret the limits.
What can be personalised
Products viewed, categories of interest, approximate location, time of day and funnel stage are the most commonly used data. It is also possible to adapt the copy to the search intent or the device.
The closer to the purchase decision, the more useful personalisation is. At the top of the funnel, overly specific messages can put off people who are still discovering.
Where discomfort begins
Discomfort appears when the ad shows something the person did not know the company knew: a private enquiry, a piece of health data, a financial situation. Even if it is legal, it can damage trust in the brand.
A practical rule: if the ad were read aloud in a crowded room, would the person feel comfortable? If the answer is uncertain, simplify.
Three levels of personalisation
- Context: category, place and time
- Behaviour: products viewed and basket
- Customer: purchases and declared preferences
- Measure gain and reactions before moving up a level
A tiered model
Level 1: contextual personalisation (category, place, time). Level 2: by on-site behaviour (products viewed, basket). Level 3: by customer data (past purchases, declared preferences). Each level needs more care with consent and with how it is communicated.
Start with level 1 and move up only when you have quality data, a legal basis and tests showing gain without complaints.
The best personalisation is the kind nobody notices.
How to measure the effect
Compare conversion rate, return and negative reactions (comments, hidden ads, removal requests). Personalisation that improves conversion but increases complaints may cost more than it earns.
Test with control groups. If the generic version performs almost as well, the extra personalisation is not worth it.
Decision table
| Personalisation level | Example | Risk |
|---|---|---|
| Contextual | Ad adapted to page content | Low |
| Behavioural on your site | Recently viewed products | Medium; consent |
| With customer list | Complements to what was already bought | Medium to high; legal basis |
| Sensitive inferences | Health, orientation or financial situation | Very high; avoid |
Common mistakes to avoid
- Showing that you “know too much”. Messages that reveal private information cause discomfort and complaints.
- Ignoring consent. Personalising with data lacking a legal basis is legal and reputational risk.
- Personalising without clear benefit. If the ad is not more useful, personalisation only adds risk.
How to measure the result
Measure conversion rate and objection or complaint rate by personalisation level. If the gain is small and complaints rise, step back to a less invasive level.
Limits and exceptions
Dynamic personalisation is not always advantageous. In small audiences, gains are limited and production and control costs can be high. When consent is low, behaviour-based personalisation is restricted. Moreover, over-personalising can seem intrusive and drive customers away. Define clear limits (what data to use, which messages to avoid), test the effect against less personalised versions and be ready to step back if reactions are negative.
Illustrative example
Illustrative example: an online bookshop shows those who consented suggestions in the same genre as their last purchase, and those who did not, suggestions by the page’s category. Both versions respect the choice, and the second has a slightly lower click rate, but no risk. The company explains in its privacy policy how and why it personalises.
Frequently asked questions
Does personalising always increase conversion?
No. It depends on data quality and audience perception. Test with control groups.
Can I use past purchase data?
Yes, with a legal basis and transparency. Check with whoever handles data protection.
How do I avoid seeming intrusive?
By using only what is needed, avoiding sensitive topics and testing audience reaction.
Can I personalise ads without consent?
Contextual personalisation, without personal data, is generally possible; using personal data requires a legal basis and often consent.
How do I explain personalisation to customers?
In plain language in the privacy policy and cookie notice, stating what data you use and how they can object.
Dynamic personalisation is only appropriate when useful, transparent and lawful; the contextual kind is lowest risk, and inferences about sensitive data should be avoided.
Sources and recommended reading
- General Data Protection Regulation (GDPR)
- ePrivacy Directive (2002/58/EC)
- EDPB: guidelines and recommendations
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.