Shopping for products

An important aspect of e-commerce is allowing users to find the products or services that they are looking for. The most common way to allow users to shop for products is to put up a website containing the entire catalog of available products. Entries in the catalog can provide detailed product information, reviews, customer comments and so on. The user can then browse items and pick the one(s) he needs.

Browsing products

Shopping carts

A well-known way to allow browsing products is the so called shopping cart model (US 5,745,681; US 5,715,314). Every item that a user selects is added to a virtual shopping cart. This allows him to browse and select multiple products. When the user chooses a check-out option, he can review his selection, and then pay all the items at once. However, shopping cart models appear to be less successful than most shops think. Recent studies have shown that over 60% of all shopping carts are never checked out, mostly because the interface for the user is too confusing or difficult to use.

Each merchant typically provides his own system for registration and for selecting and buying items. This can make it cumbersome for users to shop at multiple sites. Virtual malls can overcome this problem by allowing centralized registration and the use of one shopping cart system. This way, the user can shop at any booth in the virtual mall without having to provide his registration information at all of them. Alternatively, an intermediary can provide centralized registration and automatically translate the information to the format required at various merchant sites. The user can then shop at all those unrelated sites (WO 00/31657).

One-click shopping

A famous alternative for the shopping cart is Amazon's one-click model (US 5,960,411). After having registered (see user identification) a user can select any product and press one button to have it ship to his address. According to a statement by Amazon's CEO Jeff Bezos, a substantial portion of's revenue is obtained from the one click system. Bookstore Barnes and Noble provides a variation on the one-click model, which requires users to confirm their order after pressing the order button. This was deemed not to infringe the Amazon one-click patent.

Another alternative, especially useful for situations in which users browse content items like movies or music, is the Buy button (WO 01/57701). A special button is provided on a playback device. Watermarks are embedded in the content items, and the watermarks contain extra information about related products. When the user likes the content item, he presses the button and is provided with a screen that allows him to buy related products. Preferably, he is notified by means of a visual indicator that the content is watermarked and that he can use his buy button.

Visual content such as photos in a paper medium can also be watermarked, for instance with an URL for a website related to the photo. When the user holds the photo up to a webcam, the URL is extracted and the associated webpage is retrieved (US 5,841,978). This allows a newspaper to provide a link to its website, without having to print possibly long and complicated URLs.

Various ways exist to "bookmark" content. For instance, a portable radio can be equipped with a button that, when pressed, records the name of the song being played at that time. This information is present in the song by means of a watermark or through some other means. When the user connects his portable radio to his home computer system, the recorded names are fed to an e-commerce system. The system then shows a list of related products, such as CDs containing that song, an MP3 version of the song, and so on.

Product comparisons

With all relevant information on-line in various stores, it is possible to provide a service that allows comparisons between these products or services. The providers of comparing services earn money from referral fees received from the store that gets the sale (i.e., they operate as a sales representative).

Price-based comparisons

The earliest comparison services provide pure price comparisons: a robot retrieves the price for an item such as a book, or a CD from various stores and presents them side by side. Users can then pick the cheapest one from the list (US 6,085,176).

Price comparisons can also be used by an individual store to become and stay the lowest price offerer. To achieve this, the merchant simply regularly compares the prices of its own books against prices offered by its competitors and reduces its own price to be just below that (WO00/05666). The feature can also be made optional. The system presents a customer with the normal price and provides a button which, when pressed, compares the price against the price at other merchant sites. If any of those prices are lower, the system reduces its own price to beat the competition. To get the book at that price, the customer must immediately decide to buy. If he comes back later, he must press the button again. While this may seem inconvenient, it actually encourages sales, because users fear this may be a one-time offer. Further, by using the button, they get the feeling that this definitely is the lowest price, because they have checked at other sites.

A disadvantage of a pure price comparison is that other factors also determine the worth to a customer. If a site offers a low price but has very long delivery times, another store may be a better choice. So later developments led to sites which allow you to compare not only the price but also delivery terms and other relevant factors. To see this in action, visit for example

One example is a comparison service for travel arrangements, which not only shows price and options, but also the cancellation policy (US 5,926,798). Another example of a pure price based comparison system is, which allows users to pick the cheapest long distance or international phone operator in the Netherlands and to subscribe to it on-line. The operator gets a referral fee. This site also allows the selection of a GSM subscription based on expected or historical behavior of the user.

Feature-based comparisons

Another useful comparison service is a feature-based comparison. In this type of comparison, the user does not enter a desired product, but a desired combination of features. The system then determines the product or products that provide that combination. See for instance, where users can pick the desired features in a mobile phone (e.g., dual or triple band, light weight, replaceable covers, or programmable ringtones) and immediately buy one that matches the requirements. The Philips CE site allows visitors to pick a system such as a stereo system and compare it side by side with other systems, so the user can pick a system that he likes, and then compare other systems against it to see if there are any better choices.

Such features can also comprise the intended use of the product. For instance, the Philips DAP site allows a user to pick the "product that's right for you". The Lighting website lets a user take a questionnaire on his moods, and recommends the appropriate Bright Light product.

Recommending products

A user profile is a collection of information regarding the interests, lifestyle, likes, dislikes and so on of a particular user. This profile can be used to select or filter products, television programs, music and other content. For instance, a set-top box or television can be equipped with a filtering mechanism, which compares the content offered by a broadcaster against the user profile, and presents or records only the content items that "fit" the user profile to the user. This way, the user only sees the programs he likes. Similarly, a merchant site can use its profile for a user to select products that the user is interest in, and recommend those to the user.

The comparison can be a simple matching procedure, e.g. does a product in question match one of the genres in the user profile, or does it feature a favorite writer? The individual preferences can also be combined in a more advanced fashion by assigning relative weights, e.g. the genre of a movie is very important, but the actors in it are of secondary importance. The relative weights are then combined (for example, using weighted addition) to obtain an overall rating for a particular movie. That means that a movie in the "wrong" genre, featuring a favorite actor, will get a lower rating than a movie in the "right" genre featuring only unknown actors.

A site which offers recommendations can make money through referral fees received from sellers who received a sale due to a recommendation. When a recommender site gathers a number of regular participants, the user base may be regarded as a virtual community.

Determining recommendations from direct user input

It is important that the user profile accurately reflects the taste of the user. Since his taste may vary over time, or new facts regarding his taste may become available, it is important that a user profile can be updated.

The user can be given access to his user profile, so that he can enter new information regarding his tastes, or adjust his preferences. The user can for instance be given the opportunity to vote for or against particular programs. When he watches a particular movie or program for the first time, a pop-up window asks him if, or how much, he liked the movie. The rating the user gives is then entered into the user profile. The rating information can then be used to select other new movies or programs that are likely to be of interest to the user, based on their similarity to the rated item(s). This approach is used for books by Internet bookstore (US 6,064,980). The user can give ratings to any book in the collection, and the system then recommends books that have high average ratings and that are similar to the rated books.

Another application of this approach is a system which determines suitable upgrades and add-ons based on the hardware and software configuration of the user's computer system (WO 00/17789).

Determining recommendations from user behavior

Most users do not want to manually indicate which products they like or dislike. So, it is desirable to derive profiling information from an analysis of the user's behavior.

One important piece of information is the purchasing history of the user. It is likely, for example, that if the user has bought a lot of crime novels, he likes that genre, so the system should recommend newly published crime novels. The user's shopping cart can also be analyzed to see what he likes to buy (WO 00/17793). If a user frequently watches episodes of a TV series, then it is probable he likes that series, and so this series is given a higher rating in the user profile. If the user buys a lot of jazz CD's, the genre "Jazz" is ranked as more important in the user profile.

A known problem with adapting a user profile based on indirect information (that is, information the user did not provide explicitly), is that the inference may be wrong. For instance, if a user buys a book for someone else, the bookstore may infer that he likes that book's topic, even though this may in fact not to be the case. Recommendations based on the adapted user profile may then be quite off the mark. This is commonly solved by allowing the user to manually exclude such books from his user profile, for instance by providing a checkbox indicating the book is a gift, from which the system infers the book should not be used in adjusting the user profile.

Sharing recommendations

Merchants can also offer customers to post information and opinions about the quality of sellers, goods and services on their site. This allows other customers to find out more about the product, making the merchant site more attractive. This approach can also be used by an independent third party, such as, for instance, Opinions can be presented side-by-side with the seller's website (which not all sellers appreciate, leading to legal actions against the recommenders), or be browsed offline.

The recommender sites can monitor the opinions posted by users and provide this feedback to the merchants (for a fee). When the recommender is realized as an integrated portion of a web browser (, the recommender can monitor the user's buying habits and adjust its recommendations to the user's interests.

Recommendations can also be made available as annotations to a webpage. A web browser is extended with a mechanism that can retrieve and show "sticky notes" with annotations retrieved from the recommender site. When a user visits the webpage, the annotations for that page are retrieved and presented. The user can also add his own annotations (WO 00/62169).

Recommending additional products

A merchant site may also provide an auction system, or a classified ads system where individual users can sell their own items. For instance, a bookstore site may feature a second-hand section. In order to encourage the use of the second-hand section, the bookstore can provide, on a page for a specific book, links to that book in the second-hand section. The consumer can then either buy a new edition or a second-hand one (WO 00/58894).

When the consumer buys a number of products for a price which requires change to be given, the merchant can present a recommendation for a product which is priced equal to the amount of change. This price can be lower than the product's normal price (US 6,119,099). This saves the merchant from having to give back a number of coins. And since the recommendation is presented as a discount, the consumer is more likely to buy it.

This technique (called "upselling") is primarily useful with point-of-sale terminals, where consumers have to pay with cash money. However, it may also be attractive in Internet-based transactions, for instance with prepaid access schemes. This way, the consumer will have to buy a new prepay card faster.

Wish lists

Using a wish list, a user can keep track of items he is interested in, but not willing to buy at this time. The merchant site can make wish lists searchable (for instance, on user name), allowing friends to buy a product for the user's birthday or other occasion. When a friend selects an item on a user's wish list for buying, a purchase order is automatically generated with the billing address set to the friend's address, and the shipping address set to the user's address. Preferably, the ordered item is also giftwrapped and provided with a personalized message entered by the friend.

Recommendations based on a wish list

If all products on the wish list are too expensive, it is possible to generate a list of related items (using recommendation algorithms), and to pick one at a suitable price. For instance, the user may have a hardback edition of a book listed on his wish list. Since hardback editions are typically expensive, a friend wants to buy a paperback instead. So, he clicks on the entry in the wish list, and is presented with a detailed information page for the book. This page contains a link to the paperback edition. The friend can now order that edition instead.

The merchant site can provide the option of notifying people when a user's birthday is coming up. The URL to the user's wish list is then included in the notification. The moment at which to send the notification should be chosen so that the recipient has sufficient time to pick a present from the wish list, and the present can still be shipped in time to the user.

Wish list maintenance

When someone orders a product that is on someone else's wish list, the item can be deleted from the wish list. However, this reveals to the wish list owner that someone has ordered this product. The system should therefore distinguish between the owner of the wish list and other people, and present the owner with the full list at all times, even when items on it have been bought by others.

The wish list can be maintained on the server, but also on the client. In the latter case, the user browses an electronic catalog, typically provided on a CD-ROM, and optionally updated with information that is downloaded from an Internet server. When the user sees a product he likes, he adds it to his wish list. When he has completed browsing the catalog, he submits the wish list to the e-commerce system, and the system processes the list and generates an order. The products are then sent to the user. This is known from electronic ordering systems for supermarkets, such as the Dutch Albert Heijn's "Thuisservice".

In cases where the user periodically needs to buy more or less the same set of items, a reusable wish list is very handy. This can be realized in a variety of ways. First, the user can create a general wish list (a "master"), and re-use that every time he goes shopping. He can then add or delete items as necessary without having to re-enter every item every time. Second, the system can remember the wish lists that the user used before, and allow him to pick one. The wish list can then be modified as needed.