# Glossary

Use this page as a quick definition reference for Birdie concepts and features.

For metric formulas and metric-specific definitions, use the [Metrics Glossary page](https://ask.birdie.ai/getting-started/metrics-glossary).

## A

* **Account (Audience)** - an optional entity used to segment feedback at the company/account level (e.g., "Acme Inc.").
* **Active customers** - the customer volume input used to enable and calculate the Contact Rate metric.
* **Agent** - a support person whose conversations are monitored in [Agent QA module](https://ask.birdie.ai/agent-quality-assurance/product-overview).
* **Agent Feedback** - a workflow to create and track coaching feedback for agents, with a manual status lifecycle.
* **AI Prompt Widget** - a dashboard widget that runs a prompt on the filtered dataset and returns a narrative answer (optionally comparing two contexts).
* **Alert** - a notification triggered by a condition, delivered via a configured channel (e.g., Slack/email).
* **Area** - a broad theme derived from feedback that aggregates trends, issues, and insights across the organization.
* **Area of Interest** - another name for an Area.
* **Audience** - the set of Accounts and Users used for segmentation and joins during analysis.

## B

* **Batch ID** - an optional ingestion field used to group uploaded records into a batch.
* **Breakdown** - a chart setting that splits a metric by a dimension (e.g., by segment, source, or reason).

## C

* **Calibration** - a review step where you validate and align an AI-created entity (commonly Opportunities and Criteria) with your expectations.
* **Channel** - where something is delivered or collected (e.g., a feedback source channel, or a notifications delivery channel).
* **Chart** - a visualization of a metric over time or by breakdown, used in Exploration pages and dashboards.
* **Collection** - a container that groups Areas and Opportunities, often mirroring products, squads, journeys, or themes.
* **Comment** - a collaborative thread on an Area/Opportunity/Reason used to discuss insights and actions.
* **Conversation** - an ingested thread of messages (e.g., a support ticket) that Birdie can summarize and analyze.
* **Criteria** - the QA evaluation rules used to score and monitor service quality.
* **Criteria Catalog** - the admin area where QA criteria are created, edited, grouped into collections, and calibrated.
* **Critical Criterion** - a QA criterion that, if failed, makes the whole interaction fail regardless of other weights.
* **CSAT** - Customer Satisfaction Score; a common feedback source/metric used across Birdie.
* **CSV Export / Import** - a structured export format (multiple CSVs) used to analyze Birdie data outside the app and re-import into tools like Excel.
* **Custom Chart** - a saved chart configuration in Exploration (chart type, interval, metric, breakdown, top N, filters).
* **Custom Dashboard** - a user-built dashboard made of widgets to monitor metrics and share insights.
* **Custom Fields** - organization-specific fields used for filtering, segmentation, and joins.
* **Custom Table (Widget)** - a dashboard widget that renders a table with configurable columns and multiple metrics.

## D

* **Dashboard** - a curated set of widgets to track and communicate metrics and insights.
* **Dashboard visibility** - dashboards belong to the environment they were created in (Organization-level or Workspace-level).
* **Dashboard Widget** - a dashboard building block (Text, Chart, Number, Table, Custom Table, Prompt).
* **Data Lake / File-based ingestion** - an ingestion path where Birdie ingests prepared files from storage or a warehouse (high control and compliance).
* **DeBERTa** - a self-hosted language model Birdie uses for specific NLP tasks.
* **Digest** - a scheduled notification recap, usually sent to a channel.

## E

* **Elasticsearch query string** - the advanced keyword filter syntax Birdie supports for proximity search and complex queries.
* **Exploration** - the analysis experience where you explore feedback using filters, charts, segments, and tables.
* **Export / Import CSV** - the feature used to export Birdie feedback mappings and open them reliably in spreadsheets.

## F

* **Feedback** - a single user-submitted record (e.g., NPS response or app review) analyzed alongside conversations.
* **Feedback Details** - a filter field group that includes core text fields and org-specific attributes.
* **Filter** - a rule that scopes the dataset used by charts, tables, AI summaries, and Skye.
* **Foundation model** - a base LLM/SLM used as a building block for AI capabilities (summaries, classification, RAG).

## G

* **GenAI** - generative AI capabilities used for summaries, enrichment, and assistant experiences.

## H

* **Hallucination** - an AI failure mode where the model generates unsupported claims; Birdie mitigates this with data-grounding and controls.
* **Has Aspect** - a filter indicating whether a comment contains actionable, relevant content (as opposed to generic/noise).

## I

* **Impact Score** - a dissatisfaction-weighted metric that summarizes impact for a theme, using organization-defined source weights.
* **Indicator (Header)** - the metric cards shown at the top of navigation pages, including period-over-period variation.
* **Ingestion API** - Birdie's REST API for uploading conversations, messages, feedbacks, and optional audience data.
* **Initiative** - an action plan designed to activate a change based on a discovered opportunity.
* **Intention** - a classification that captures what the user is trying to do (e.g., Problem, Request, Question).
* **IP filtering** - an access control that restricts Birdie access to approved IP ranges (typically used with SSO).

## K

* **Keyword filter** - a text filter that supports AND, OR, parentheses, and proximity operators.
* **Kind** - the ingestion schema type that defines additional fields for a record's origin (e.g., support\_ticket, review, nps).

## L

* **LGPD** - Brazil's data protection law; Birdie supports anonymization and region handling to help with compliance.
* **LLM** - Large Language Model; used for generative features like summaries and assistant answers.
* **LLaMA** - an open model that Birdie can self-host for specific workloads.

## M

* **Manual Criterion** - a QA criterion flagged as requiring human evaluation (bypasses AI assessment).
* **Message** - one entry inside a Conversation (e.g., a ticket reply), uploaded via the ingestion API.
* **Metric** - a numeric measure computed over feedback (counts, scores, rates, impact); details live in the Metrics Glossary.
* **Metrics Header** - the top summary block on analysis pages showing key metrics for the current filter context.
* **MIPM** - Birdie's continuous improvement cycle: Monitor, Identify, Prioritize, Measure.

## N

* **Native connector** - a standard integration that pulls data from a well-known SaaS source via API credentials.
* **NLP** - Natural Language Processing; used for enrichment tasks like sentiment and intention detection.
* **Noise** - feedback that lacks actionable content and is filtered out by signal detection.
* **Notification** - an outbound message from Birdie (alerts, digests) delivered to configured channels.
* **NPS** - Net Promoter Score; a common feedback source/metric used across Birdie.
* **Number Widget** - a dashboard widget that displays a single KPI for fast monitoring.

## O

* **OpenID** - an authentication protocol supported for SSO.
* **Opportunity** - a specific insight derived from feedback that highlights trends, issues, or areas for improvement.
* **Opportunity Status** - labels that indicate an opportunity lifecycle state (e.g., Pending Calibration, Rejected, Processing).
* **Organization** - the top-level tenant boundary in Birdie; it controls base settings, metrics logic, and global defaults.

## P

* **PDCA** - a classic improvement loop (Plan, Do, Check, Act) that inspired Birdie's MIPM cycle.
* **PHI** - Protected Health Information; handled via anonymization/redaction to avoid sensitive data storage.
* **PII** - Personally Identifiable Information; removed/redacted during ingestion by Birdie or by your company pipeline.
* **Public / Private (Collections)** - visibility settings for collections (organization-wide vs creator-only).

## Q

* **Qualtrics** - an integration used to ingest Qualtrics feedback into Birdie.
* **Quality Score** - a QA score computed from criteria (weights and critical rules) across reasons, areas, and agents.

## R

* **RAG** - Retrieval-Augmented Generation; grounds AI answers in your data instead of free-form generation.
* **Reason (Contact Reason)** - a QA category that represents why customers contacted you, used to apply criteria and monitor agent performance.
* **REST ingestion API** - the programmatic ingestion method where your systems upload records directly to Birdie endpoints.
* **RoBERTa** - a self-hosted language model Birdie uses for specific NLP tasks.

## S

* **Segment** - a defined group of customers/feedback built from profile or behavior criteria, used to slice analysis.
* **Segment Collection** - a grouping mechanism for segments so you compare like-with-like (e.g., Lifecycle vs Firmographics).
* **Sentiment** - the emotional tone classification applied to feedback (Positive, Neutral, Negative, with levels).
* **Signal** - feedback that contains meaningful, actionable content (as opposed to noise).
* **Skye** - Birdie's AI assistant that answers questions and helps discover insights within the current filter context.
* **Source** - where feedback came from (e.g., NPS, tickets, reviews, social).
* **SSO** - Single Sign-On; integrates Birdie with your identity provider (often paired with IP filtering).

## T

* **Table Widget** - a dashboard widget for structured data exploration with sorting and filtering.
* **Template (Dashboard)** - a dashboard shared across workspaces as a starting point, usually controlled by organization admins.
* **Text Widget** - a dashboard widget that supports Markdown to add narrative, instructions, and context.

## U

* **Unit costs** - cost inputs used to compute potential savings metrics in Organization settings.
* **User (Audience)** - an optional entity used to segment feedback at the individual level (e.g., end-user profile).

## V

* **Voice of Customer (VoC)** - the practice and product surface focused on understanding and acting on customer feedback.

## W

* **Widget** - a component inside a Custom Dashboard (charts, tables, numbers, text, prompts).
* **Workspace** - a scoped environment inside an organization that controls data visibility, settings overrides, and access by role.

## Z