Products · 02 · Business Intelligence
← All productsQrly
Ask. Build. Share. Self-service business intelligence, on your own infrastructure.
Connect a database, ask a question, and build a report, a chart or a dashboard — that is the whole system, until you need more. Qrly also carries a full spreadsheet engine, four cache and replication tiers, an approval-gated AI agent workforce, and a productised BCBS 239 risk-data governance module that no other BI platform ships.
40Connection types
13Visualisation types
500Worksheet functions
14Principles governed
The problem
Self-service analytics stops being self-service the moment it matters.
The tool is bought so business users can answer their own questions. Six months later every meaningful number still routes through two data engineers, the important reporting has migrated back into untraceable spreadsheets, and the AI features meant to help are the ones the risk function will not approve — because nobody can say what the model was told, or what it changed.
01
Engineer bottleneck
Every new question is a ticket. The tool's promise dies in the queue.
02
Shadow spreadsheets
The real reporting lives in files with hand-typed columns beside a stale export.
03
Unauditable AI
A model that writes queries and saves content, with no record of the prompt or the change.
04
Governance as a project
Data lineage, critical data elements and quality rules assembled by hand in documents.
The solution
A BI platform that takes governance as seriously as it takes charts.
01
Query without SQL
A point-and-click definition compiled into dialect-correct SQL for around forty database types, with drill-down, calculated fields, and a converter that turns hand-written SQL back into the visual builder.
02
Spreadsheet, properly
Not an export button — a real grid with around five hundred worksheet functions, four conditional-formatting families, and refresh-safe blocks that keep your hand-written columns beside their record when the query reorders.
03
Governed by construction
Every query, every AI prompt, every agent tool call and every schema change is written to an audit table. Schema authoring is off by default at three independent levels.
Capabilities
01Data Connectivity & Querying
- Around forty connection types — PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery, Redshift, Databricks, Trino, ClickHouse, DuckDB, Synapse, Vertica, DB2, Sybase and more.
- A visual builder covering tables, joins, filters, aggregations, grouping, sorting and limits — no SQL required — alongside a schema-aware SQL editor with completion, hover and formatting.
- A converter turns hand-written SQL back into the visual builder, so a hand-written query is not a dead end.
- Fourteen filter operators plus date helpers for week, month, year and rolling windows; calculated fields usable anywhere a column is, sanitised against anything that is not a read.
- Nightly schema introspection with cached columns and foreign keys, a data dictionary and an entity-relationship view, plus freshness probes for a 'data as of' status.
- Per-connection pooling, query timeout, row limits, a concurrency governor with a wait queue, SSH tunnels, TLS modes, and a guard against reaching internal hosts; every credential encrypted at rest.
- Drill-down in two modes — to underlying rows, or narrowed to the selected group — on visual and native queries alike.
02Dashboards & Delivery
- Thirteen visualisation types including pivot, funnel, gauge, progress and both regional and point maps, with a semantic model layer for dimensions, measures, rollups and period-over-period comparison.
- Twelve-column dashboards with drag and resize, dashboard-level filters, auto-refresh, and cascading filters that drive many cards with differing field names.
- Live dashboards pushed on an interval, or driven by a change notification from the source database itself; a morning briefing that diffs yesterday against today into a banner.
- Reports for side-by-side comparison or multi-card layout, with revocable public share links, report packages bundling many reports into one workbook, and report bursting per filter value.
- Alerts on rows returned, goal reached or below goal, delivered by e-mail, chat, webhook or SMS; scheduled deliveries as PDF or CSV with optional written commentary.
- Streaming results, chunked exports, and output as JSON, CSV, newline-delimited JSON or Parquet.
- Content marked 'Verified' as a trust signal, flippable only by curators and logged every time.
03Spreadsheet Engine
- A hand-written formula engine and virtualised grid — around 500 worksheet functions, cross-sheet references, whole-column ranges, lazy evaluation and cycle detection.
- Excel's two editing modes, click-and-drag reference picking with range outlining, fill-down with reference rewriting, and row/column reordering that relocates every reference.
- Four ways to bind data: a saved question, a visual query, ad-hoc SQL, or a natural-language request resolved to SQL once, after which refreshes cost nothing.
- Refresh-safe blocks: new source columns are appended, removed ones are cleared, your own rows and columns are never overwritten, and key columns return each record to its prior row.
- Conditional formatting by cell rule, colour scale, data bar and icon set; export keeps formulas as formulas.
04Agentic BI & Proactive Intelligence
- Four personas — Analyst (read-only), Composer (drafts questions and dashboards), Modeler (builds semantic models), Investigator (root-cause work) — running a reasoning loop over ten tools.
- Every AI write is a proposal a human must approve by default, shown as a card with accept and reject; row-level data access for AI is off by default.
- An audit table stores the exact prompt, message and injected context sent to the provider, so users can audit what the model was told.
- Seven provider families, one fully self-hosted; per-organisation token, session and step budgets; a built-in Model Context Protocol server exposing eight tools.
- An insights feed detects trend changes, milestones, new records, stale data, correlations, seasonality and anomalies, with severity and de-duplication.
- Anomaly detection over a rolling window excluding the point being judged; a badged change hands off to the Investigator persona in one click.
- Investigations produce findings and an executive summary, triggered manually, automatically above a severity threshold, or weekly.
05Performance & Governance
- Four performance tiers: in-memory, persistent storage that survives a restart, materialised tables written into a database you own, and replication with reload, watermark and change data capture.
- A typed lineage graph across connections, tables, questions, dashboards, alerts and deliveries, with a printable view and a downloadable report.
- The BCBS 239 module: fourteen principles, scored self-assessments with sign-off, a critical data element register, quality rules across six dimensions, a risk report register, and a remediation log.
- Schema authoring with import, dialect compilation and a plan preview, gated by three independent switches all off by default, with every executed statement recorded verbatim.
- A query audit log capturing actor, query type, full SQL, rows, duration, status and cache state, with retention and an auditor export.
- Two embed paths — a dashboard SDK with signed embeds, and a scoped embeddable assistant widget — plus an auto-generated database API on a widely used PostgreSQL REST convention.
- 240 help articles across twelve categories, including forty troubleshooting runbooks and nine administrative operational runbooks.
■Java 21 on Spring Boot 4 with virtual threads, native-image hints included
■PostgreSQL 14 or later, 86 tables, schema owned entirely by versioned migrations
■Server-rendered vanilla ES6 front end — no package manager, no bundler, no build step
■91 REST controllers with OpenAPI, an agent protocol server, and a database-compatible table and function API
■Single sign-on via OpenID Connect, LDAP and Active Directory — multi-factor authentication and SAML are not yet implemented
■Around 71 MB vendored assets, self-hosted fonts, zero content-delivery-network references
What it does
- 01A no-code query builder compiles to correct SQL across roughly forty database types — with a converter that turns hand-written SQL back into the visual builder.
- 02A real spreadsheet engine with around 500 functions sits on top of live, governed data — not a static export.
- 03Every AI write is a human-approved proposal, logged in a full prompt-and-context audit trail; BCBS 239 governance ships as a built-in module.
At a glance
For whomBanking · Insurance · Regulated enterprise · Embedded analytics
LicenceLicensed per tenant, not per viewer, with unlimited users. The annual licence is based on your own turnover ex VAT, charged in marginal tranches like income tax so there are no cliffs, starting at a €900 minimum and capped at a 25% yearly increase. Hosting, hardware and AI provider costs are yours; consultancy, setup, migration and training are quoted separately.