Yusp Personalization for Marketplaces

The cold-start is the arch-enemy of recommendation systems. It can affect both users and items.

For users, the cold-start means when the system has limited or no information on the customer’s behavior and preferences.

The item cold-start represents the lack of user interactions with the data upon which item-to-item relations can be drawn (we still have the metadata, though, but that won’t often suffice for truly fine-tuned recommendations).

The item cold-start is an obvious domain for the aforementioned content-based approach as it makes the system less reliant on transactional and interactional data. However, creating meaningful personalized experiences for new users is a much trickier problem that cannot necessarily be solved by simply gathering more information on them. It is quite typical –especially in the case of eCommerce sites or online marketplaces with wide product portfolios –that customers visit a website with completely different goals over time.

First they come to buy a microwave, but the next time they’re looking for a mobile phone. In this scenario, the data gathered in their first session is not relevant to the second. An intriguing approach to tackling the user cold-start problem is session based or item-to-session recommendations. This roughly means that insteadof relying on the whole interactional history of customers, the system splits this data into separate sessions.

The model capturing the users’ interests then builds on session-specific clickstreams. It then analyses this information in real time and triesto predict what the customer is looking for based on data solely from that session.

For classified sites monetizing in a CPC model -advertisers can promote their advertisements in a CPC model increasing their revenues by:-Increasing the CTR of in-house ads by personalization -The logic behind internal ad personalization is similar to product recommendation only the list of recommendable items is limited to a smaller set of ads (promoted ads) and additional objectives likecoverage and frequency capping is also taken into account.

-For classified sites having ad revenue from external ads as well, and internal and external ads are competing for the same ad space, optimizing the selection between in-house and external ads bypredicting CTR and eCPM for each ad will increase the ad revenue of the site. Those ads are worth displaying in a specific context –user, location, time of the day, category, page type, etc. –which will result in the highest effective CPM (eCPM).

For CPC ads the eCPM can be estimated from the click price and the expected CTR on the ad. If Yusp is displaying personalized internal ads on the site, it can predict the expected CTR as well and from this, it can decide which ad type (internal CPC / CPM, external CPM) and finally which ad is the best choice for that specific user and context.

Sellers at internet marketplaces and classified sites often sell very similar products; this may cause certain difficulties in product recommendation. One of the challenges is the lack of metadata or incorrectly filled metadata. Since users do not necessarily fill out all the metadata related to the item, it is cumbersome to find relationships with other items. Also, same product can be uploaded by many different vendors which causes the same product to have multiple item IDs.

Micro-clustering algorithm is for finding the same items from different vendors. It uses metadata to estimate the same items. This way we can distinguish which item ID belongs to which specific product and calculate their events cumulatively.

Since the events will accumulate the predictions will get better by micro-clustering.