Check out our new client-side integration support and deploy personalized recommendations faster
We knew we had to bring something new to the table, when participating as a Beta Startup at Web Summit, the largest technology conference in Europe. We have used this amazing opportunity to announce a release of our new client-side integration support. This improvement enables deployment of personalized recommendations on your site extremely easy and fast.
So far, we had only supported server-side integration, which meant that you could call our API solely from your backend server. The server-side integration has numerous advantages, such as complete control over sent requests (they are all sent from your own servers) or easy monitoring of the response times and availability.
We have however realized, that many portals prefer requesting recommendations directly from a web browser or mobile app to keep the backend part simple. This was not possible with our server-side SDKs where a proxy for requesting recommendations at customer side would be needed due to the used authentication mechanism. Implementation of such proxy requires some expertize and prolongs the integration phase by a few hours.
Now we offer an easy solution:
In order to achieve personalization, you need a unique identifier for each user. You can use Google Analytics for this purpose, as we show in this example.
As of now, we support Google Merchant Feed and Heureka Feed for eshops, and we are working on other formats such as Media RSS for media publishers. If you currently use an unsupported format, do not hesitate to contact us.
By default, each feed is updated every four hours, but it can be configured to synchronize more often (for example, a fast update is crucial in case of news portals). See the documentation for more info.
The accuracy of our recommendations improves when properties of items are stored into the catalog, and also when a sufficient amount of interaction data is gathered. Therefore, it is a common practice to first put into production sending interactions to Recombee, and after a while (for example a week, depending on your traffic) start asking for recommendations. You can, of course, use our SDKs to upload historical interactions to eliminate the data bootstrap period.
When you start recommending items, you can fetch properties of items from Recombee and display recommendations including product images without a need of querying your database for reading products data.
Here is one real-world example of the integration. View the website source code for inspiration: https://beerandbrewing.com/the-secrets-of-british-cask-conditioning/.
After you enable the recommendations, you can check the performance metrics in the web interface. Drop us an email so we can help you fine-tune the performance and increase your revenue.