Find the answers to frequently asked questions about Discovery.
Q: What is Discovery?
A: The Ooyala Discovery recommendation engine uses industry standard big data and recommendations technologies, including Hadoop and MapReduce, and collaborative filtering, to generate related and trending content recommendations for consumers. Recommendations are highly configurable. You can use labels and filters to customize the recommendations provided. You can configure the player to provided recommendations when the consumer is done viewing a selected asset and/or when the Player is paused.
Q: How does Discovery work?
A: Discovery provides recommendations to viewers. The site administrator, and anyone added to the distribution list, receives a weekly Insight report showing what recommendations are displayed and which recommended assets are viewed.
Q: How is Discovery enabled and managed?
A: Contact your sales representative to activate Discovery features for your site.
When Discovery is enabled, the recommendation options can be configured for a specific Ooyala player using Backlot (The Discovery options are available form the Publish tab). If your site uses multiple Ooyala players, Discovery must be enabled separately for each. Discovery profile settings are named uniquely for the linked player.
Although an Ooyala player is required to create a Discovery profile, the player and the Backlot interface are not required to power recommendations on your site. Once Discovery is enabled via Backlot, Discovery settings can be managed from the Backlot Discovery interface and the Discovery API. If using only the Ooyala player Discovery interface to deliver recommendations to consumers, the deployment can be managed completely using Backlot. However, the Discovery APIs are a powerful option for managing site recommendations with options for configuring the player display. For more information see:
Q: My site uses a different URL for each video asset. Can Discovery support this?
A: Absolutely! Discovery can be configured to open recommended videos in the same player, the same page, or new page. A recommended video will open in a new page as long as permalink URL is available in the metadata for that asset.
Does Discovery work on mobile platforms?
Q: Does Discovery allow editor control over the recommendations?
A: Editors can influence Discovery recommendations in several ways:
- Use labels to include or exclude specific content based on the applied label.
- Select spots in the Discovery tray (1st recommended video, 2nd recommended video, etc.) to be filled by the freshest content (asset) that may contain a specific label. This allows editors to promote certain content in a targeted manner.
- Override Discovery recommendations by selecting assets with shared labels. For more information about these options see Labels.
Q: How can I confirm that Discovery recommendations are increasing engagement on my site?
A: The Discovery Insights Report is generated weekly. This report provides metrics on engagement categories that help determine Discovery’s impact of discovery recommendations on consumer engagement. For more information about the Insight report, see Discovery Insights Reports (Deprecated).
If you want to test metrics independent of the insight report use two Ooyala players, one with Discovery enabled and one without Discovery on your site for A/B testing. A/B testing is also useful when using Discovery APIs to deliver recommendations outside the Ooyala player on your site.
Q: What can I expect in terms of a lift in streams and engagement?
A: Metrics around engagement based on Discovery recommendations are dependent upon the volume of traffic for your site.
Q: I have a proprietary recommendations engine today. How can I try Discovery out and measure growth over the current engine?
A: Trial versions of Discovery are available. With a trial version of the recommendations engine you can review the metrics available from the Insights report, and perform A/B testing. For more information about starting a Discovery trial, contact your sales representative.
Q: How can I disable autoplay of the next video from the Discovery tray?
A: Autoplay cannot be disabled entirely. However you can use the Discovery API to set the default_countdown parameter to a high number. This number specified the number of seconds after which the next video will autoplay. The default is 10. The Backlot Discovery interface limits this setting to 99 seconds or less. For details about using the API to set this limit higher than 99 seconds using the default_countdown parameter, see Managing Discovery Profiles with the Discovery API.
Q: What data is available to track engagement from Discovery?
A: The Discovery Insights Report is generated weekly. This report provides metrics on engagement categories that help determine Discovery’s impact on consumer engagement, in relation to organic consumption. For more information about the Insight report, see Discovery Insights Reports (Deprecated).
Q: Why do I see the same asset recommended as in the previous batch?
A: Discovery is a stateless recommendation engine. Session specific consumer behavior data is not stored. In any given session the consumer’s historical usage data is leveraged, which may result in the same video asset being recommended again over subsequent recommended lists. The Discovery UX in the Ooyala player filters out any videos that may have been in the previous batch of recommendations. Ooyala recommends that publisher sites use a similar front-end/client side filter when using the Discovery APIs.
Q: What is the Discovery Feedback Loop?
A: Discovery recommendations are based on machine learning algorithms that track consumer behavior. The Feedback Loop is a critical piece of this infrastructure that registers impressions and clicks for recommended content and processes the data to refine the preferences for each consumer in real-time. For more information see Discovery Feedback Loop.
Q: How does Discovery relate different impression and click data passed through the Feedback Loop API for each consumer?
A: The device ID is a required identifier for recording consumer impressions and click data. Consolidation of data points is based on device IDs. In addition, the Ooyala player bucket info may also be sent via the Feedback Loop, enabling Discovery to gather additional data about the impression or click event. For more information see Discovery Feedback Loop.
Q: Can I influence the algorithms that Discovery uses?
A: Discovery uses machine learning algorithms that learn continually based on consumer behavior. In a cold start scenario, when no usage data is available for a consumer, content related to the current asset is recommended. When no current asset is available, Trending data is recommended.
Discovery provides several options to help you influence asset recommendations:
- Use labels to include or exclude specific content based on the labels. For more information see Labels.
- Use the Backlot Discovery interface to configure asset profiles. For more information see Creating Discovery Profiles Using the Backlot UI.
- Use the Discovery API to configure editorial parameters for filtering content. For more information see Query String Parameters.
Q: Editorial query string parameters and Discovery Profiles appear to do the same thing. Are they any different?
A: Discovery API requests return a base set of recommendations from the machine-learning algorithm. This base set can be modified using Discovery profiles (see Discovery Profiles), or by leveraging editorial query string parameters (see Discovery Editorial Query String Parameters). Both options can be used to customize the base set of recommendations, however Ooyala recommends using Discovery profiles rather than editorial query string parameters.
Q: I used filters when calling the Discovery APIs and do not get any assets returned. Why?
A call that mixes the filter_by and the discovery_profile_id could cause this issue. The filter_by parameter always take precedence over discovery_profile_id, effectively nullifying the profile. For more information see Discovery Design Considerations.