Decision intelligence, in plain words
Short definitions of the terms used around data, analytics and governance — written for the people who decide, not only for those who build.
Concepts
- Decision Intelligence
A discipline that connects data and analysis directly to the decisions an organization makes. Where traditional reporting ends at a view of the figures, decision intelligence carries on to what changed, why it matters and what the evidence is, so that a decision can be taken and defended.
In Rover Compass: Rover Compass answers a leadership question with a chart, a written analysis and the basis of the figure. Read more →
- Generative BIGenerative business intelligence
Business intelligence in which the chart and the commentary are composed at the moment a question is asked, rather than prepared in advance by an analyst. It widens access to data, and it makes governance more important, because the answer is new each time.
In Rover Compass: Every data answer in Rover Compass arrives with at least one chart and a written analysis. Read more →
- Conversational Analytics
Analysis carried out as a dialogue: a question, an answer, then a follow-up that builds on it. The conversation keeps its context, so the second question does not have to restate the first.
In Rover Compass: Sessions in Rover Compass retain context, and suggested follow-up questions are offered with each answer. Read more →
- Natural Language QueryNLQ · Text-to-SQL
Asking a database a question in ordinary language instead of in a query language. The system interprets the intent, composes the query and runs it. The quality of the result depends on how well the system knows the structure and the vocabulary of the data.
In Rover Compass: Questions are asked in Thai or English, and the query behind any answer may be reviewed. Read more →
- Self-service BI
An approach in which business users obtain their own figures without asking a data team for each one. It shortens the wait for an answer, provided that definitions and access remain governed centrally.
- Executive dashboard
A prepared screen that shows a fixed set of measures for leadership at a glance. It serves the questions that were anticipated when it was built; a question outside that set needs a new view, or another way of asking.
In Rover Compass: Answers worth keeping can be pinned to a gallery, and placed on a dashboard where dashboards are enabled in your plan. Read more →
Data foundations
- Data Catalog
An organized inventory of the data an organization holds: its tables, columns and what each one means. A good catalog lets people, and software, find the right data and understand it before using it.
In Rover Compass: Rover Compass keeps a catalog for each connection, in step with your database and searchable by your team. Read more →
- Metadata
Data about data: the name of a column, its type, its description, when it was last refreshed. Metadata is what turns a column called "amt_2" into something a reader can interpret correctly.
- Semantic Layer
A layer of business definitions that sits between raw tables and the people who ask questions. It states once what "revenue" or "active customer" means, so that every report and every answer uses the same definition.
In Rover Compass: Your organization defines and certifies its key measures, and questions that refer to them use those definitions. Read more →
- Certified MetricCertified measure
A measure whose definition has been reviewed and approved by the organization. Certification removes the familiar problem of two departments reporting two different figures under the same name.
- Data Quality
The degree to which data is complete, valid and current enough for its purpose. Missing values, stale tables and codes outside the expected set are typical quality problems, and each can quietly distort a figure.
In Rover Compass: The platform profiles connected data, checks rules your administrators set, and mentions relevant quality issues in the answer. Read more →
- Data Readiness
An assessment of whether a data source is prepared for analysis: whether it is described, current, profiled and reliable. It answers the question "can we rely on this before we decide?"
In Rover Compass: Rover Compass scores each connection on readiness, speed and accuracy. Read more →
- Schema changerelated: data lineage
A change to the structure of a database, such as a table added or a column removed. Unnoticed schema changes are a common cause of reports that silently stop being correct; lineage is the related practice of tracing where a figure comes from.
In Rover Compass: Administrators can be told by email when the structure of a connected database changes. Read more →
- Fiscal year and Buddhist era
A fiscal year is the twelve-month period an organization reports on, which need not begin in January. In Thailand, years are often written in the Buddhist era, which is the Gregorian year plus 543. Both affect what "this year" means in an answer.
In Rover Compass: Time zone, calendar, fiscal year start and reporting currency are set per organization. Read more →
Governance and security
- Data Governance
The policies and controls that decide who may use which data, for what, and how that use is recorded. Good governance lets an organization widen access to data without losing control of it.
- Role-based access controlRBAC
A way of granting permissions through roles rather than person by person. A member receives a role, and the role determines what they may see and do.
In Rover Compass: Roles in Rover Compass decide which parts of the product each member may use, and administrators can adjust what each role is shown. Read more →
- Table-level access
Access control that decides which tables a person or a group may query. Someone without access to a table cannot obtain answers drawn from it.
In Rover Compass: An administrator places members in groups that see only the tables assigned to them. Read more →
- Row-level security
Access control within a table: two people query the same table and each sees only the rows that concern them, such as their own region or branch.
In Rover Compass: Row-level filters on data access are enabled per plan. Read more →
- Multi-factor authenticationMFA · two-factor authentication
Signing in with more than a password, typically a code from an authenticator application on a phone. A stolen password alone is then not enough to enter the account.
In Rover Compass: Members can enable an authenticator application, and an organization can require it for everyone. Read more →
- Audit log
A record of significant actions: who signed in, who changed a setting, who exported data, and when. It is how an organization answers "who did this?" after the fact.
In Rover Compass: Organization administrators read their own audit log without raising a request. Read more →
- PDPAPersonal Data Protection Act
Thailand’s Personal Data Protection Act B.E. 2562, which governs how personal data is collected, used, disclosed and retained, and gives individuals rights over their data.
- Personal data masking
Replacing personal values, such as a name or an identification number, with placeholders before the data is passed on. The analysis still works on the pattern, while the individual remains unidentified.
In Rover Compass: Columns marked as personal data are masked before a result is passed to Rover AI for analysis. Read more →
- Read-only access
A connection that can read data and cannot change it. For analytics it is the safest arrangement, because no question, however it is phrased, can alter the source.
In Rover Compass: Rover Compass accepts read-only queries alone and validates each one before it runs. Read more →
- Data retention
The rule for how long data is kept before it is removed. A retention policy limits exposure and is often a compliance requirement.
In Rover Compass: Each organization decides how long its question and answer content is retained. Read more →
Analysis
- KPIKey performance indicator
A measure chosen because it reflects whether an objective is being met, such as revenue growth or occupancy rate. A KPI is only useful when everyone computes it the same way.
- Metric alert
A notification raised when a measure meets a condition, such as falling below a level or changing by more than a set percentage. It brings the figure to the person, instead of relying on someone to look.
In Rover Compass: Alerts can be set on a pinned figure by administrators, and by every member for themselves where the organization allows it. Read more →
- Threshold
The level at which a figure is treated as needing attention. A well-chosen threshold separates ordinary variation from a real change.
- Year-over-yearYoY
A comparison of a period with the same period one year earlier. It removes seasonal effects, which a comparison with the previous month would not.
- Pareto analysis80/20 analysis
Ranking items by their contribution and showing the running total, to reveal the few items that account for most of the result. It directs attention to where effort pays most.
In Rover Compass: Pareto charts are among the advanced charts enabled per plan. Read more →
- Quadrant analysis
Plotting items on two measures and dividing the chart into four areas by a threshold on each axis, for example high volume and low efficiency. Each area suggests a different action.
In Rover Compass: A question that asks for a quadrant is answered with one item per point and the four areas named. Read more →
- Cohort
A group that shares a starting point, such as customers who first bought in the same month. Following cohorts over time shows retention more honestly than a single overall average.
- Moving average
The average of the most recent periods, recalculated as each new period arrives. It smooths short-term noise so that the underlying direction is easier to see.
- Percentile
The value below which a given share of observations falls. The median is the 50th percentile; the 90th percentile describes the high end without being distorted by a single extreme value.