E-commerce personalisation is software that shows each shopper a different version of an online store, based on what that shopper has viewed and bought before. Two people who open the same homepage at the same moment can see different banners, different products and a different order of categories.
Here is how it decides. Every time a shopper views a product, adds something to the cart or buys, the store records it against that shopper's profile. A returning shopper is recognised by her login, or by a cookie, which is a small file that the website stores in her browser. From these records, the software works out what she is likely to want next. When she opens a page, it chooses what goes in each slot: the top banner, the rows of products and the order of the products in each row. It does this in a fraction of a second, before the page appears.
Say a fashion brand's website has 2,000 products. Shopper A, in Bhubaneswar, viewed six cotton kurtas in size M last week, but bought none. Shopper B, in Mysuru, bought men's running shoes twice this year. When shopper A opens the homepage, the top banner shows the new kurta collection, and the first row shows cotton kurtas that are in stock in size M. When shopper B opens it, the banner shows a sale on sportswear, and the first row shows socks and running T-shirts. A first-time visitor, with no history, sees the store's best-sellers.
The software can change the search results too. If shopper A searches "white top", kurtas come before western tops. So personalisation is a part of e-commerce search and recommendations. It uses the same ranking idea as the recommendation systems that choose films and posts on other apps.
How does the brand know that it works? It tests it. For two weeks, half the shoppers see the personalised homepage, and the other half see the standard one. The brand then compares how many shoppers in each half bought something. Running tests like this is the job of conversion rate optimisation (CRO) software. If the personalised half does not buy more, the brand changes its rules.
Personalisation runs on the store's own records, called first-party data, which the brand collects with the shopper's consent. Used carelessly, it can make shoppers uncomfortable. A banner that shows a product someone looked at privately, such as a medicine, can embarrass them on a shared phone. So stores set limits, such as not personalising on health products.