Keytalk Configurator

Keytalk List Page


  • Keytalk Overview

    • Keytalk Calls: This section shows the number of times users have used Keytalk in the last 7 days. The data is accumulated over time and can be used to identify Keytalk popularity trends and forecast user queries.

    • Calls by Category: This section shows the number of times a Keytalk from each category (e.g. ‘Climax & Ending’, ‘Emotional Response’, ‘Character’) was used in the last 7 days. This data can help you understand audiences’ interests and preferences.

    • In Use Keytalk: This section shows the number of Keytalk currently in use by your organization’s business.

    • Keytalk List: This section provides a list of all Keytalk. Clicking on a Keytalk name will take you to the configuration page for that Keytalk.

Keytalk unit detail page - Keytalk Information


  • Overview of Keytalk unit

    • Vertical: Refers to the item Category of the Keytalk you are viewing

    • Total Results: Number of items (in the same item category) mapped with the Keytalk unit

    • In use: Switch the Keytalk on/off from your service

      • Situations in which you should TURN OFF a Keytalk include: the Keytalk displaying inadequate results that cannot be resolved with given configuration options the Keytalk is out of season, the Keytalk can stir unpleasant controversy

      • Press ‘+ SAVE CHANGES’ on the top right section of the domain to properly deploy it onto your service

    • Call Volume: View monthly call volumes of the Keytalk unit. A time-linear graph on the popularity of the Keytalk (time of day/ day/ month) is available on the right.

  • Crowd’s Ranking: View actual items mapped with the Keytalk, ranked by the highest score to lowest. We call this data “Crowd’s Ranking” because our Keytalk data represents the Public

  • Attribute Settings: Quick view on the attributes that are forming the Keytalk

  • Modification History: View the changes made to the Keytalk

Keytalk Unit Detail Page - Configurator


  • You can view the detailed results of each Keytalk and change the attribute settings to match your business goals. Real-time simulated item results will appear on the right as you add and delete attributes and adjust attribute weights

  • Taste Attributes: Attributes containing NLP word data. Added attributes will recalculate what the Keytalk’s intended meaning is, and add higher-scoring words to the subject Keytalk’s data. More Taste attributes included in one Keytalk will result in richer results feed. For example, it is Valentine’s Day season and an OTT business is starting a project to expose more romantic movies to audiences in efforts to improve conversion rates and examine movie completion rates. At this time, the most popular Keytalk on the platform is “psychologically thrilling” the default taste attribute paired with the Keytalk is “psychological_thrill”. It has 560 movies attached in order. Because you don’t want to impede on the popular user experience of “psychologically thrilling”, you click on the taste attribute “romantic_subplot”, and attach it to the attribute “psychological_thrill”. The results on the right will simultaneously reorder itself, showing you recalculations of ranked movies that scored highest in both tastes, “psychological_thrill” AND “romantic_subplot”.

  • Meta Attributes: Attributes containing metadata. Added meta attributes will function as criteria for items to score bonus points and move up in the list of results. For example, an e-commerce platform has signed a partnership with the beauty brand, ABC. The business goal is to sell as many ABC lipsticks as possible. Beauty brand ABC chooses a handful of Keytalk that would be a good association with their brand image including a couple of “Most Popular” Keytalk. Meta attributes “brand:ABC” and “category:Lipstick” is applied to the selected Keytalk. The results on the right will simultaneously reorder itself moving up any items tagged “brand:ABC” and “category:Lipstick” a little more above the list. Remember that setting a Meta Attribute will help you moderate result slightly by giving items more points, but it cannot completely manipulate the default AI-generated results.

  • Adjust Weight: You can adjust the weight for added Taste and Meta Attributes which will result in a bigger impact on attributes . If when recalculating results. (Scenario continuing from the example on Meta Attributes) After adding the Meta Attributes “brand:ABC” and “category:Lipstick”, there is a problem. ABC has always been a consumer-favorite brand so items from ABC appear in the Top 10 list for the Keytalk it was added on to. However, none of the ABC items that made it to the Top 10 recalculations are lipsticks. ABC has always been a skin-focused beauty brand and they only started launching lipsticks 6 months ago. To give the unknown collection of lipsticks a little more boost, you may adjust the weight of “category:Lipstick” to 5.0 and “brand:ABC” to 3.0. This will give ABC Lipstick products an accumulative extra points of 8.0 and they will move up in the rankings accordingly.

  • Reanalysis and Deployment schedule is marked at the top of the page. The date listed is when the changes you make become deployed to your users.