Glossary of terms

  • Updated

Core concepts

  • Actor – A dataset representing a business entity, for example, User, Account, or Document.
  • Alert – A message produced from a rule evaluation that interests a user. For example, an alert about the daily active user count dropping 20% below its 3-month average.
  • Analytic block (Block) – A building unit for composing arbitrarily complex analytical computations. Analytics offers a library of pre-defined block types.
  • Application – A logical grouping of objects for a particular business use case. Managing different use cases across teams or initiatives is an organizational convenience. Applications do not share any state among them.
  • Attribute – A column in an Analytics Dataset representing a qualitative or categorical property of the entity the dataset represents, for example, an age column in a user dataset or a device column in an event dataset.
  • Catalog – The repository of all objects organized into folders for an application. Users see parts of the catalog depending on their permissions. Each application has its own independent catalog structure.
  • Cohort – A group of actors sharing common properties or behaviors. For example, a cohort of enterprise users hosted two meetings within the first week of signup.
  • Dashboard – A collection of visualizations organized in a grid view. You typically use dashboards to monitor commonly used operational metrics.
  • Derived column – A calculated column of a dataset based on other columns of that dataset or related datasets. For example, a derived column of the User dataset called Engagement Cadence has daily, weekly, or monthly values for each user depending on their usage activity.
  • Warehouse-Native Analytics – Optimizely’s platform that connects directly to your cloud data warehouse (Snowflake, BigQuery, Redshift, Databricks), enabling direct querying and transformation with no data duplication.​
  • Connection – A connection acts as a bridge, enabling direct querying of a data warehouse from within the Analytics solution. It's essentially a secure pathway that lets Analytics to interact with your chosen data source, such as Snowflake, BigQuery, Amazon Redshift, or Databricks, to retrieve and analyze data.
  • Data source – The enterprise repository of data. Typically, it is a data warehouse such as Snowflake or BigQuery.
  • Exploration – Provides analytical insights, such as retention, conversion, churn, and engagement. It is built using an exploration template and includes a query definition and a resulting visualization.
  • Folder – Contains objects that you can save with a name. Folders logically organize objects such as datasets, explorations, dashboards, and so on. Folders are hierarchical, and a folder can contain sub-folders.
  • Funnel – Maps website visitors' flow to specific funnel steps that result in conversions or signups.
  • Group – A collection of users organized together for a business function. You can use groups to control users' access to objects and application privileges.
  • Measure – A column in an Analytics Dataset representing a quantitative measurement of business activity, such as the number of daily active users.
  • Notification – A notification is a message to a user about an alert triggered in the system. You receive notifications within the Analytics desktop browser or the Analytics Mobile application.
  • Organization – Represents the company that is a customer of Analytics.
  • Query – A query defines an analytical computation using SQL, NetScript, or templates. Executing a query produces results that you can visualize.
  • Rule – Defines one or more criteria for alert generation and the payload for generated alerts.
  • Sampling – A query performance optimization technique that uses a representative sample of data for computations instead of the entire data.
  • Segment – A category for breaking down analysis. For example, you can break down a retention analysis by segments of user cohorts or attributes.
  • Template – A pre-defined set of configurations for defining a particular object, such as an Exploration. Templates offer an easy way to author commonly used types of objects. Analytics offers templates for Explorations, Cohorts, Derived Columns, and Metrics.
  • User – A user is a person belonging to an organization that has access to the Analytics application.

Data modeling and configuration

  • Dataset – A tabular view (rows and columns) of data used for analytics. Datasets are logical views of data in your data source representing a business entity or activity, for example, User, Session, Ticket, Account, or Event.
  • Source Dataset – A representation of a logical mapping to a table, view, or SQL query in the data warehouse.
  • Derived Dataset – A dataset created by combining existing datasets (source or derived) to create a compound dataset.
  • Union Dataset – A logical combination of multiple warehouse tables exposed to users as a single logical dataset. A typical use case for union datasets is to create a single logical event dataset from a set of event tables in the warehouse, one for each event type.

  • User (Actor) Dataset – A dimension table representing users, with a relationship to the event dataset. Essential for event analysis.

  • Column Actor Dataset​ – A special dataset representing an actor without a dedicated table-defined purely by its presence in an event’s identifier column.

  • Event stream – A dataset representing a stream of events, such as user interactions in a mobile app.
  • Event type (Event) – A representation of values in an event stream dataset column that describes the event type.

  • Event stream annotation – Defines event datasets by identifying the following:

    • A timestamp column (for example, event_ts) to order events.

    • An event type column (for example, event_type) for display.

Infrastructure and performance

  • Service account and Materialization schema – A Service Account is used for secure access to your data warehouse, while a Materialization Schema defines where and how your data is stored in the warehouse after being processed.​
  • Materialization – The process of creating and storing materialized tables within a data warehouse to improve query performance. These materialized tables contain pre-computed results of frequently used queries, allowing for faster access to data when running reports and analyses.
  • Schema – The organization or structure of a database.

Experimentation and Metrics

  • Warehouse-Native Experimentation Analytics – Lets you integrate experiment data (Feature Experimentation or Web Experimentation) with warehouse-held business metrics to build a unified Experiment Scorecard-facilitating comparison of variations directly in your warehouse.

  • Decision Dataset – A dataset that captures variation assignment and decision events from Optimizely experiments.

  • Metrics – The reusable analytical measures built in Analytics, including the following:

    • Numeric aggregation (for example, SUM, AVG).

    • Conversion (event counts/users).

    • Ratio metrics (for example, event/event).

    • Custom formulas combining the previous.

  • Guardrail metrics – A secondary metric used to monitor the potential negative impact of an A/B test or experiment on other aspects of a product or user experience. They act as a safety net, alerting teams to unintended consequences that might arise from focusing solely on a primary success metric. Essentially, they help ensure that while optimizing for one thing, you're not inadvertently harming others.

  • Baseline – The reference variant in scorecards against which other variations are compared.

  • Experiment Scorecard – A visual report in Analytics showing raw values, percentage impact, statistical significance (through Stats Engine), and confidence intervals.

Stats engine and advanced analysis

  • Optimizely Stats Engine – A modern statistical approach used in Optimizely's experimentation platform. It combines sequential testing and false discovery rate control to provide faster and more accurate results in A/B testing and other experiments.
  • CUPED (Controlled-experiment Using Pre-Existing Data) – A statistical method that reduces variance in A/B tests, enhancing sensitivity and making it easier to detect differences between groups. CUPED helps you run smarter experiments that reach conclusions faster and with greater confidence.
  • Outlier management​ – A feature designed to improve the reliability of experiment results by mitigating the impact of extreme values (outliers) on metric calculations. It works by capping extreme values at a defined threshold or smoothing them based on percentiles, preventing a few unusual data points from skewing the overall results. This is particularly useful for metrics like revenue, where a few large purchases can disproportionately affect the average.
  • Sample Ratio Mismatch (SRM) detection​ – A feature that automatically monitors the distribution of traffic across different variations in an A/B test to identify potential imbalances. This helps detect early signs of issues in the experiment's configuration or implementation that could lead to biased results.