In short: Personalising an online shop starts from signals you already have: pages viewed, purchases, where visitors come from and abandoned baskets. The changes that pay off most are recently viewed products, sensible related products, different messages for returning customers, behaviour-based emails and a site search that works, all with proper consent in place.
Two customers walk into the same shop and see the same window display. One is looking for running shoes. The other is buying a present. Personalising means showing each of them what they are most likely to be interested in. You don't need a huge system to get started.
The signals you already have
- What they looked at on this visit and on previous ones.
- What they bought and when, for registered customers.
- Where they came from: a campaign, a search, a link in a newsletter.
- Device and country, useful for language, currency and shipping.
- What they put in the basket without buying.
The changes that pay off most
Recently viewed products
A block on the home page or at the bottom of the product page that takes customers back to where they left off. It's simple and almost always useful.
Sensible related products
‘Customers who bought this also bought’ works when the links make sense. For a small catalogue it's better to choose them by hand. For a large catalogue you use rules or automated tools that you check from time to time.
Different messages for returning customers
A customer who has already bought from you doesn't need the welcome banner. They need to see what's new, or an easy way to reorder what they bought before.
Behaviour-based emails
A reminder about the abandoned basket. A message after the purchase with tips on using the product. An invitation to reorder when a consumable product is about to run out. These are the emails with the best ratio of effort to results. See email for an online shop.
Site search
A search that understands typos and synonyms and puts your best sellers at the top. People who use site search already know what they want. If they find nothing, they leave.
Privacy: what you need to do it properly
- Explain clearly which data you use and why.
- Ask for consent for the cookies and tools that require it.
- Don't use special categories of data (health, opinions, orientation) for personalisation.
- Offer a simple way to unsubscribe or to request deletion of data.
The more your data comes from what customers do in your shop, the more useful it is and the fewer problems it raises. See privacy for an online shop.
How to tell if it works
- Pick a single indicator for each change: clicks on recommended products, orders from email or repeat purchase rate.
- Compare with a group that doesn't see the change, or with an equivalent period.
- Keep what works and drop the rest.
What to leave for later
AI systems that predict what each customer will buy make sense when you have plenty of data and a large catalogue. Get results with simple rules first. For the cases where AI is genuinely needed, see the page on artificial intelligence for e-commerce.
Further reading: artificial intelligence for e-commerce · email for an online shop · privacy for an online shop
Frequently asked questions
What are the first personalisation steps to take?
A block of recently viewed products, carefully chosen related products, different messages for customers who have already bought, behaviour-based emails and a site search that understands typos and synonyms.
Does personalisation require the customer's consent?
It depends on the tools you use. Non-essential cookies and tracking require prior consent, and the privacy notice must explain which data you use. Special categories of data must not be used.
How can you tell if it works?
By choosing a single indicator for each change (clicks on recommendations, orders from email, repeat purchases) and comparing with a control group or an equivalent period.