Concept card

BARC — Benchmark: How fast is your BI platform, really?

As of 2026-10-06

Edited by Cédric Mary, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is BARC?

The four dimensions that separate enterprise BI platforms — query latency, rendering, concurrency at 10-500 users, and 10M-1B row scaling — are what fail on a laptop demo but decide the RFP.

What it is

BARC's 'Benchmark — How fast is your BI platform, really?' is a lab-style performance study comparing BI platforms on representative enterprise workloads. The URL slug references Qlik and Power BI as the primary subjects, which is instructive: the benchmark is designed to address the Qlik vs. Power BI vs. the field performance debate that dominates large-enterprise BI evaluations.

The benchmark methodology covers four dimensions: query latency (time to first row under defined data volumes), dashboard rendering time (full-screen time-to-interactive under concurrent user load), concurrency behaviour (how latency degrades as simultaneous users increase from 10 to 100 to 500), and data-volume scaling (how each platform performs as the fact table grows from 10M to 1B rows). These four dimensions are what discriminate BI platforms at the enterprise tier — any platform is fast on a 1M-row test dataset on a developer laptop.

Why it matters

Concurrency behaviour (10 to 500 simultaneous users) is the dimension that decides operations-analytics deployments, not raw query speed.
Citing BARC's neutral lab methodology beats trading marketing claims when a client's IT team makes an unverifiable competitor performance claim.
The benchmark's four dimensions double as the acceptance-test checklist for scoping a concurrency-sensitive SAC deployment.

Key points

BARC lab-style performance benchmark of BI platforms.
URL slug references Qlik and Power BI — likely primary subjects.
Methodology categories: query latency, dashboard rendering, concurrency, data-volume scaling.
Use when defending SAC Live in a multi-vendor RFP — gives an independent comparator.
Numeric results and vendor rankings paywalled.
None of BARC's four dimensions was designed to measure AI-agent query traffic (Joule 'just ask', MCP-connected LLM clients) — a platform sized for human concurrency can still be stressed by a much smaller number of agents firing several queries per question.
Microsoft's own DirectQuery documentation states a one-million-row intermediate result limit per query and a default 10-connection concurrency cap per data source for Power BI Pro — concrete, sourced constraints worth checking before citing a Power BI benchmark number.
A benchmark's headline number is only as transferable as its test configuration — schema complexity, concurrency curve (not just headcount), and sponsorship disclosure all change what the number is evidence of.

Terms used on this page

In-memory BI: BI platform that pre-loads data into RAM for sub-second query response — the Qlik and SAC Live architecture. Contrast with Direct Query / Pass-through, where every dashboard interaction fires a database query.
Direct Query: Power BI's architecture for connecting live to a database (e.g. Azure SQL, Fabric, SAP HANA) without import — lower latency on fresh data, but concurrency and size constraints apply.
Concurrency test: Benchmark scenario simulating multiple simultaneous users querying the same dashboard — the most discriminating test for enterprise BI performance because it reflects production load.
Intermediate result-set limit: A cap on the number of rows a single query (or intermediate operation) may return before failing — Power BI DirectQuery enforces one million rows per Microsoft's own documentation, a concrete constraint independent of any lab benchmark.
Query folding: The process by which a BI tool's transformation steps are translated into a single native query pushed to the source, rather than being executed client-side — required for DirectQuery to scale; breaking folding is a common cause of a 'slow DirectQuery' complaint.
Agent query pattern: The distinct load shape generated by an AI agent (Joule, an MCP-connected LLM client) answering a single natural-language question with several rapid exploratory queries — structurally different from the human click-and-wait pattern BARC's concurrency dimension was designed around.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks · the facts worth quoting.

Open in the app →

Guides that answer with this page

These guides cite this page as one of the sources their answer rests on.

SAP Analytics Cloud vs Qlik: What Actually Differs