Skip to content

Key Things to Know Before Starting

Please use this document as a reference for terms that are commonly used when working on the platform

Key Objects

  • Data Source: Set of data, in the form of a table, generated outside the platform and exposed to the platform through Data Vault.

  • Source Column: A column within a data source table

  • Data Element: A column or aggregated set of columns that has been registered into the platform using Data Vault.

  • Feature: A calculated object whose calculation formula has been registered into the platform. Feature calculations can only take Data Elements and other Features as inputs.

  • Model: An algorithm registered into the platform that can be used to predict the value of a Data Element or a Featiure (i.e. dependent).

  • Product Type: A group of business products that have similar structure and could be registered into the platform using a common set of parameters or features. Examples of Product Types are Installment Loan, Credit Card and Insurance Policy.

  • Product: A business product that is registered into the platform by assigning a set of values or range of values to the parameters of the Product Type that it belongs to. For instance a 36 month installment loan with loan amounts ranging from $1,000 to $20,000 and interest rates ranging from 0.05% to 0.29%.

  • Framework: A custom set of interrelated calculations that can be registered as whole package into the platform and used to make decisions on a product. A framework requires a set of inputs to perform its calculations and produces a set of outputs. 

  • Framework Input: Frameworks can take as inputs Data Elements, Features, Models and Product Features.

  • Framework Output: Framework outputs are specific results of the Framework calculations that are registered into the platform (e.g, Expected Monthly Default Rates).

  • Framework Metric: Framework metrics are Features created using Framework outputs and potentially other objects as inputs. These metrics can be used to create Policy simulation reports (e.g., Expected Annual Losses)

  • Valuation Function: A valuation function is a Framework metric that the User specifically intends to use as an objective function in an optimization strategy (e.g., NPV)

  • Strategy: A strategy is a set of decision rules that are written using Data Elements, Features, Models, Framework Metrics and Valuation Functions. 

  • Rules-Based Strategy: A strategy whose rules always result in a Pass or Fail outcome

  • Action Strategy: A strategy whose rules always result in an action (e.g., “set amount to $xxx”, “request income”, etc.)

  • Optimization Strategy: A strategy that consists in optimizing a valuation function

  • Policy: A set of strategies to be used to make a decision on a given product.

Key Tasks

  • Simulation: After defining a new object, a User can simulate it using filtered platform’s data. Simulations can use advanced filtering to test the object on a very specific sample of data. Simulation sample sizes are also unlimited and can be as large as needed.

  • Clone: Cloning an object allows Users to create a new version of an object. The new version will be in draft mode and can be edited, simulated and submitted for approval. Once approved the new version becomes by default the version that will be used to create any new object,

  • Copy: Copying an object allows Users to create a new object using information from the copied object. For instance, a User can copy an object, edit its definition and metadata and save it as a new object.

  • Approve: Approving an object allows Users to use it as an input to other objects. An Object can be used as an input to define a new object even in the draft state.

  • Activate: Currently Only approved policies can get activated. There also can be only one approved policy per product. Activating a policy for a product will result in all entities related to that product (prospects, applications, loans and customers) to be decisioned using that policy.

  • Execute: Currently only approved and activated policies can get executed. Executing a policy means effectively running it on a new set of data (typically production data) and getting a decision as an outcome. 

  • Exporting Artifact: An Artifact is a bundle of documents, including an executable script, that gets generated once an object is registered in the Platform. Artifacts can be exported for an object under any status (draft, pending approval, approved or rejected)

  • Exporting Documentation: A document can be exported once the object is registered on the Platform. This document can be exported either in docx or Html format. The documentation contains the simulation’s reports and charts along with the metadata of the simulated item. At present this functionality is available for data elements, features, models and policies.

  • Sharing Objects: An object can be shared with other users even in the draft status. User can be granted (Read and Write) access.

Key Concepts

  • Entity: An entity is an object to which a set of information is attached and on which a business decision can be made. For instance, an application is an entity to which a credit bureau file is attached and on which a decision to approve or decline is made. Currently, our platform supports four (4) types of entities: prospects, applications, accounts and customers.

  • Aggregation: An entity can have multiple information of the same type attached to it. For instance, a loan can have multiple monthly balance snapshots attached to it. Users can perform calculations to summarize multiple information of the same type into one new information (e.g., Maximum, Sum or any other calculation).

  • Definition (Registration): Any information (data, formula, model, policy…) used in the platform has to be defined (registered) in order to be used either to create new information or make a decision. A definition can consist of just a mapping to some raw data coming into the platform, a formula, a model algorithm or a set of rules.

  • Versions: A version of a definition is created every time a definition is either newly created or subsequently changed.

Info

A page can be printed to a PDF file by using option ctrl+p and can be saved on the local drive.