Release notes for 15.13 (September 3, 2015)
We updated the style and consistency of our Dashboard. The new UI is easier to use, and more visually appealing. Example:
Catalog browser updates
We upgraded the catalog browser with more useful product information:
- Regional pricing
- Link to the product page on client site
- Recommendable status
- Last time the product was updated from a feed
- Release date
- Product image URL
Audit logging of site configuration changes
We implemented detail logging for site configuration changes. The activity log records names, times, and the specific change. This enables administrators to quickly identify the cause of changes in functionality or performance.
API updates for SEO-friendly integrations
The RecsForPlacement API responses now include the product URL and the click logging URL separately. This enables clients to render links to their domains for products we recommend while still capturing click events for RichRelevance products. This improves SEO: links to recommended products match the domain of the retailer website. This is important for Discover and Recommend server-side clients.
New region-aware Conditional Probability (CP) models and strategies: View-View, View-Purchase, Purchase-Purchase
The region-aware CP models and strategies bring additional accuracy to recommendations presented to shoppers in a specified region. This benefits clients that run region-based businesses where products and inventories are different.
Learn more: Recommendations Strategy Guide
Updated Mobile SDKs for iOS and Android to accelerate native app development
We updated our Mobile SDKs to simplify and package the RichRelevance personalization APIs into logical native interfaces for mobile developers. The SDKs support all RichRelevance applications (Recommend, Discover, and Engage). The SDKs also support our Preference Center, User Profile Service (UPS) and GetProductInfo APIs.
Using the SDKs reduces the learning curve of RichRelevance infrastructure. It also accelerates app development. With pre-developed objects (Strategies, Placements, Products, and Users), app developers can jump right into custom feature development and user interfaces.
Learn more: Mobile SDKs
Load shopper offline transactions to UPS
Individual shopper-level transaction details are now available in the User Profile Service (UPS). This enhances personalized strategies since the strategies now know what the shopper has purchased across online and offline channels.
All offline feed fields available in BYOS
The BYOS Hive table is now expanded from purchase transaction data to incorporate all of the fields provided in the offline feed file. This enables our BYOS clients to utilize all available data for building custom strategies.
- Merchandising rules could not be set when the start date was set for the current date, or with No End Date selected.
- Site analytics placement report was not dis-aggregating to display performance at the placement level.
Segment Export (15.13.02 update on 9/21/2015)
You can now export single or combination segments.
A text file is created and sent to your FTP based on your preferred schedule (for example: every 30 minutes or once daily). This file includes user IDs and other segment-related information. It allows your email service provider to process it, and send out a variety of follow-up emails; some examples include welcome, first order, cart abandonment, browse, and post-purchase emails. It can also be used for other purposes such as personalized text messages (SMS) or other use cases where it’s helpful to export segments.
Learn more: Segment Export user guide.
Link multiple user profiles
The User Profile Service (UPS) now merges data from multiple profiles. A User Link mapping file can be loaded through build-ftp. It provides mapping to merge profiles. A maximum of 10 separate profiles can be merged together to maintain a fast response.
Learn more: User Profile Service
Event Posting Service (Beta)
We released a new service to directly write events to update user information in UPS. This is currently beta to select clients. We’ll have more information on this update soon.
Science and MVT
Outlier detection now applied to all KOTH (Uber) revenue metricsWhen using metrics such as RPS, attributable revenue, or any other metric that is based on a sales amount, Uber can now do outlier detection. Using site config, it is possible to configure outliers to be handled:
- Normally (no filtering);
- Truncated (large order amounts are set to be no larger than the outlier threshold); or
The outlier threshold is the mean +3 standard deviations based on the previous 28 days of data.
Improvement to Deterministic Treatment Assignment in MVT
Prior to this release, if you set Deterministic Treatment Assignment in MVT, a technical issue was sometimes causing traffic allocation to deviate from the requested traffic. This would happen for small number of tests.
With this release, treatment assignment is redesigned to have an additional layer of random mapping. This avoids the traffic imbalance as well as possible biases in treatment assignment.