Personalisation is not “Hello, George”. It is the customer not having to hunt for the thing they would have bought anyway.
And it carries a prerequisite people skip: if your e-shop is not already comprehensible, personalisation saves nothing. Structure first, then adaptation.
Want us to look at what can be personalised in yours?
Tell us what you sell and on which platform and we will tell you what pays off first.
The forms that pay off, in order
| Form | Difficulty | Impact |
|---|---|---|
| “Related products” on the product page | Easy | Medium |
| “Bought together” in the basket | Easy | High |
| Abandoned basket email | Easy | Very high |
| Recommendations from purchase history | Medium | High |
| A dynamic home page | Hard | Medium |
| Personalised pricing | Hard | Risky |
The third has the highest return per hour of work and comes ready-made on almost every platform. If you do not have it switched on, you do not need anything else right now.
The last one is not recommended. Different prices for the same catalogue per user creates legal and commercial problems out of all proportion to the benefit — and when noticed, it costs you trust.
“Bought together” only works if your categories are coherent: see category structure.
What data you actually need
Less than you think. Three levels:
| Level | What it needs | What it enables |
|---|---|---|
| No sign-in | What they viewed this visit | Related products, recently viewed |
| With an email | Order history | Recommendations, repeat emails |
| With an account | Preferences, lists | Full personalisation |
The first level gives 60% of the benefit with no personal data at all. Start there: behaviour within the same visit is the strongest signal you have.
And a minimum volume: below 500 orders in total, automated recommendations produce random results. Until then, manual rules do better — “anyone buying X, show Y”.
The boundary: consent and intrusion
Two separate issues, both critical.
The legal one: tracking behaviour for personalisation requires consent through a cookie banner with a genuine option to refuse. Personalised emails require sign-up; an order by itself is not a newsletter subscription.
The commercial one: there is a point where personalisation becomes creepy.
- Intrusive: “You looked at this 3 minutes ago” in an advert on another site.
- Not intrusive: “Recently viewed” inside your own e-shop.
- Intrusive: referring to something the customer never explicitly told you.
- Not intrusive: suggesting a complementary product in the basket.
The rule: personalise on what the customer did in your shop, not on what you infer about them.
Not sure where the line sits in your case?
Send us what you collect today and we will tell you what is permitted and what will grate.
How to start, in four steps
- Switch on what already exists in your platform — related products, recently viewed, abandoned basket.
- Clean up categories and tags. Recommendations are only as good as your categorisation.
- Split your email list into three — new, repeat, dormant — and send each a different message.
- Measure for a month before adding anything else.
The third step is the cheapest real personalisation there is. Three segments with distinct messages outperform any algorithm on a small catalogue — the structure is in email marketing.
How to produce the copy variations quickly: AI in digital marketing.
What to measure
| Metric | What it shows |
|---|---|
| Clicks on recommendations | Whether the suggestions are relevant |
| Average order value | Whether the basket grows |
| Abandoned basket recovery | The most immediate revenue |
| Repeat purchases | Whether a relationship is forming |
Always compare against a control group. Without one you will credit personalisation with a rise that came from the season or an offer — the same measurement logic as the loyalty programme.
Frequently asked questions
Do I need a separate tool?
Usually not at the start. The major e-shop platforms have built-in recommendations and segmentation.
What does it cost?
The basic functions are included. Specialist tools start around €30–50 a month and make sense with a large catalogue.
Does it work with few products?
Below 50 products, manual rules beat any algorithm.
Does it affect speed?
Yes, if you load many scripts. Measure before and after — a slow page cancels the gain.
What if the customer refuses cookies?
Only within-visit personalisation works. It is smaller, but it exists.
The first step
Go into your own e-shop as a customer, put a product in the basket and leave.
If no email arrives within 24 hours, you have found the highest-return personalisation you do not have running — and it is one setting.
Next step
Book a free 30-minute call with the team at The Dev Alley. We will look at what your platform supports and what is worth switching on first.