BiG EVAL
Swiss data quality and observability platform combining deterministic rules, AI anomaly detection and captured human expertise. Deployed on-premises, in private cloud or hybrid, it validates data across any source and scales one rule to thousands of assets.
What is BiG EVAL?
BiG EVAL is a data quality and observability platform built by a Swiss software company for enterprises where being wrong about data has consequences. The current generation is branded BiG EVAL NX³, presented on the vendor's own site as the successor to BiG EVAL versions 1 and 2 rather than as a separate product.
Its central argument is that no single approach is sufficient, and that each new capability adds a layer instead of replacing the previous one. The first layer is hard rules: deterministic checks that return the same result every time, backed by a full audit trail and hundreds of pre-built validation methods covering completeness, comparisons, performance and schema drift. The second is AI: unsupervised anomaly detection, drift detection, context-aware plausibility checks, rule generation from natural language, and agents that read unstructured sources such as PDFs. Every AI-generated rule stays visible and editable, which is how the vendor answers the black-box objection. The third is human expertise, captured through a Briefbox for context notes, trust surveys, a Domain Memory that explains anomalies with business context, and a Business Hub where non-technical staff contribute without touching code.
The headline engineering claim is massive scaling: one rule definition is applied automatically across thousands of data assets through metadata discovery, instead of being copied and adapted per table or column. Helsana is cited running 17,000 automated checks a day, and a large retailer 3,000 auto-generated checks. Results aggregate into the Data Quality Radar, which maps scores onto whatever structure reflects the organisation, from processes and systems to governance domains, and lets a reader drill from a board-level view down to the root cause.
The platform is technology-neutral, with connectors for databases, warehouses, cloud platforms, files and APIs, and it fits into DevOps practice through regression testing across environments and quality gates in CI/CD. It is deployed on-premises, in private cloud or hybrid, and the AI itself can run locally, in a private cloud or against OpenAI-compatible APIs, with an MCP connector for driving tests from LLMs.
What it does
- Define a validation rule once and apply it automatically across thousands of data assets using metadata
- Map data quality scores to your own business structure and watch them update in real time on the Data Quality Radar
- Detect anomalies, drift and outliers that no fixed rule could describe, with explainability on every finding
- Turn a plain-language description of a quality expectation into an executable validation rule
- Block a deployment when data quality checks fail, through quality gates in CI/CD and data integration pipelines
- Capture what domain experts know and connect it automatically to detected findings
- Compare data across different systems and technologies within a single test definition
When to use BiG EVAL / When not to
A quick filter to help you decide if BiG EVAL is the right fit.
When to use BiG EVAL
- Data quality managers and data stewards in regulated sectors who must produce an auditable, board-level view of how reliable their data actually is
- Data engineers and ETL developers who refuse to hand-write one validation rule per table and want metadata-driven coverage across thousands of data assets
- Enterprises running data warehouse migrations or replatforming projects that need continuous before-and-after comparison, as Helsana and Swiss Life did
- Organisations bound by data sovereignty requirements that rule out SaaS and need the platform, and optionally the AI models, running entirely on their own infrastructure
- Data and compliance teams that want AI on data quality without a black box, since every AI-generated rule stays visible, auditable and editable
When not to use BiG EVAL
- Buyers who need a published price, as nothing is quoted publicly in either site language and every package goes through a sales conversation
- Individual analysts or small teams looking for a self-service tool, since the product is enterprise software installed on a Windows server with its own repository database
- Anyone expecting a permanent free tier or a clearly free trial, because evaluation runs through a negotiated proof of concept period
- Teams that want data automatically cleaned or corrected, as the platform detects, scores and documents defects rather than fixing them
- Organisations needing a product interface beyond English and German, which are the only two languages the user interface ships in
How to use BiG EVAL
A typical end-to-end flow, from setup to results.
- Start with the free Radar Assessment, an online self-evaluation that needs no signup and takes about five minutes
- Book a live demo with a product expert to walk through your own use cases, or request a quote for the modules you need
- Check the system requirements and prepare the server, then install BiG EVAL and its repository database
- Configure the instance: appsettings.json, SSL and HTTPS, proxy, register URL and global parameters
- Set up authentication and user management through LDAP, Active Directory, SSO or Microsoft Entra ID
- Create data source connections to your systems, from SQL Server, Oracle and Teradata to SAP R/3 ECC, Excel and CSV
- Build your first test case with the tutorial, or describe what you want checked in natural language and let the AI agent draft it
- Group test cases into test suites and organise them with tags and parameter lists
- Schedule runs, trigger them through the REST API, or drive them from the PowerShell module and your CI/CD pipeline
- Model your business domains in the Data Quality Radar and connect checks, metrics and trust surveys to each of them
Pros & Cons
Pros
- Over ten years of continuous production use with named, checkable references including Helsana, Swiss Life, Denner and Ifolor
- Metadata-driven scaling that avoids maintaining one rule per data asset, evidenced by two customers running thousands of checks
- Explainable AI by design, since every generated rule remains visible, auditable and editable rather than hidden in a model
- Genuine data sovereignty, with on-premises, private cloud or hybrid deployment and the option to run AI models entirely locally
- Covers both development-time testing and production monitoring using the same set of rules, instead of forcing a choice
- Thorough public documentation spanning user, administrator and developer handbooks, a REST API reference and a PowerShell module
- Modular commercial structure, where a mandatory Core module is extended only by the modules a customer actually needs
Cons
- No price is published in either site language, so no budget can be estimated without contacting sales
- Neither a permanent free plan nor a clearly free trial: the FAQ answers the free trial question with a proof of concept period whose cost is never stated
- No Data Processing Agreement is published or offered, which is a notable gap given the regulated audience
- No security certification such as ISO 27001 or SOC 2 is claimed anywhere on the site
- Server installation is required, so getting started needs Windows system administration skills and lead time rather than a signup form
- Probe queries are stored in clear text in the repository database, which the vendor documents openly but which needs handling if queries carry sensitive values
- Public release notes stop in 2023 while the product is marketed as a third generation, and the product interface exists only in English and German
Pricing & Plans
No pricing is published. The pricing page presents every package as available on request, in English as in German, and directs prospective customers to a tailored quote. Licensing is described as an annual subscription, with multi-year contracts arranged individually. The structure is modular: a mandatory Core module is combined with optional AI Assistance, Operations, Insights and Human Expertise modules, each offered at a Basic or Premium tier, and vendor-hosted AI is billed as usage-based credits. There is no permanent free plan. The vendor offers a proof of concept period rather than a free trial, and does not state whether that period is free of charge. The only genuinely free item is the Radar Assessment, a separate self-evaluation tool that requires no signup.
- full test engine with comparisons
- performance
- schema drift and scripted test methods
- unlimited tests and test instances
- scheduling
- API-triggered and parallel execution
- data models as an abstraction layer
- template gallery
- everything in Basic
- plus the Rules test method for record-level validation
- quality gates in CI/CD pipelines
- data profiling
- SSO and directory synchronization with Entra ID and Active Directory
- custom user roles
- external key vaults
- audit and audit reports
- AI coding agent for test and suite scripting
- test case suggestions from metadata
- AI-assisted test concept creation
- and vendor-hosted AI billed as usage-based credits. Pricing on request.
- everything in Basic
- plus an AI coding agent for validation rules
- bring your own AI through your cloud keys
- private cloud or fully self-hosted models
- and an MCP connector to create
- run and analyze tests through LLMs and agents. Pricing on request.
- Microsoft Power Automate integration
- webhook alerts with a standard payload
- and simple reaction scripts. Pricing on request.
- everything in Basic
- plus delivery guarantee with retries and delivery log
- custom payloads
- HMAC and OAuth authentication with key vault secrets
- conditional triggers
- routing and escalation chains
- alert aggregation
- deduplication and quiet hours
- data quality metrics derived from validation results at test
- suite and dataset level
- and a Data Quality Radar visualizing one domain model. Pricing on request.
- everything in Basic
- plus unlimited domain models with their radars
- manual and calculated metrics
- AI anomaly detection covering outliers
- drift and predictive alerts
- and AI-enhanced context-aware validation. Pricing on request.
- Briefbox for proactive context notes from domain experts
- and trust surveys measuring how far the business trusts its data. Pricing on request.
- everything in Basic
- plus Domain Memory and reasoning
- the Business Hub for business users
- an input workflow for experts to confirm
- revise or explain findings
- and issue tracking connected to tests
- findings and domain memory. Pricing on request.
- Test Automation Bundle
- Operations Bundle and Suite
- or a custom combination of modules. All priced on request.
Data, GDPR & hosting
A consolidated view of how BiG EVAL handles your data.
GDPR overview
The privacy policy, last updated June 2026, states that personal data is processed under the Swiss Federal Act on Data Protection and, where applicable, the GDPR, and it cites concrete legal bases: Art. 6(1)(f) legitimate interest, Art. 6(1)(b) contract and Art. 6(1)(a) consent. A representative under Art. 27 GDPR is formally designated: VGS Datenschutzpartner UG in Hamburg. Access, rectification, erasure, restriction, portability and objection rights are listed, with the FDPIC in Switzerland and local authorities in the EU named as complaint bodies. Transfers to the USA are covered by the EU-U.S. Data Privacy Framework and standard contractual clauses. Two gaps matter for a vendor targeting regulated buyers: no Data Processing Agreement is published or offered, and no data protection officer is named.
Who owns the data?
Because BiG EVAL is installed on the customer's own infrastructure, the customer keeps both the data under test and the repository database that holds every test definition, result and connection. The vendor publishes no claim of ownership over customer data, and its privacy policy covers website visitors only. The documentation is explicit about protection: data source usernames and passwords are encrypted and decryptable only by the BiG EVAL server itself, using the Windows server key, while protecting the file system holding appsettings.json is stated to be the customer's own system administrator's responsibility. One caveat is documented openly: probe queries are stored in clear text in the repository database, with no encryption of sensitive values a query may contain.
Reuse rights
No terms and conditions are published anywhere on the site, so there is no contractual grant or restriction to report on what an end user may do with data or results. What the documentation does state is a limit on outbound flows: when BiG EVAL triggers a third-party application such as Power Automate, Zapier or Make, it sends only the event and related statistics, and explicitly does not send data from the customer's data sources. The AI layer follows the same logic, since models can run fully self-hosted through Ollama with Mistral or Devstral, in which case the vendor states no data leaves the customer's network. Nowhere does the site address whether customer data is used to train models, in either direction.
Data retention & training
Hosting summary
There is no vendor-hosted offering to describe. BiG EVAL is installed by the customer on-premises, in a private cloud or in a hybrid arrangement, so the data under test and the repository database holding configuration, test definitions, results and credentials all reside in the customer's own infrastructure and jurisdiction. The vendor consequently declares no hosting country or region for customer data, and full data sovereignty is presented as a selling point rather than an option, with Swiss banks cited as running everything on-premises. The AI layer follows the same principle and can be self-hosted through Ollama with Mistral or Devstral, in which case the vendor states no data leaves the customer's network. One distinction is worth keeping clear: the bigeval.com website itself is hosted on Cloudflare infrastructure in the USA and its product analytics run through PostHog's EU cloud, but that concerns website visitors only and tells you nothing about where the platform's data lives.
Things to keep in mind
Risks and trade-offs to weigh before adopting BiG EVAL.
- No pricing is published in either site language, so evaluating cost and comparing vendors requires committing to a sales conversation first
- No Data Processing Agreement is published or offered, and no security certification such as ISO 27001 or SOC 2 is claimed, on a product sold to regulated industries
- Probe queries are stored in clear text in the repository database, so any sensitive value written into a query is stored unencrypted and needs a review of who can read that database
- The publisher appears under three different names across its own site, and the footer copyright name matches no entity in the Swiss commercial register; the registered company is Bolt Technology Consulting GmbH
- The legal entity was registered in March 2010 under a different name and a different manager, and only renamed in 2011, while the product itself is dated from 2015 on the site; company age and product age are not the same thing here
- Market statistics quoted prominently, including the BARC 2026 ranking and the 77, 60 and 12 percent figures, come from BARC, Gartner and Precisely/Drexel and measure the market, not this tool; the 320 percent ROI and sub-four-week break-even claims carry no published source
- Third-party services listed in the privacy policy, including Cloudflare, Google, PostHog, LinkedIn, Apollo.io, Ahrefs and Zoho, process website visitor data only and say nothing about the platform installed on customer infrastructure
Setup & Integrations
Technical difficulty
High for the initial deployment. This is server software, not a signup: you check system requirements, prepare a Windows server, install the application and its repository database, then work through appsettings.json, SSL, proxy and global parameters before wiring up LDAP, Active Directory, SSO or Entra ID and creating data source connections. Expect a Windows administrator plus someone comfortable with SQL and scripting in C# or PowerShell. Against that, the vendor claims a first quality check within three minutes once running, quotes two to four weeks for an enterprise rollout with its support, and lists around twenty certified implementation partners.
Deployment
Integrations
Supported languages
Behind BiG EVAL
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
How much does BiG EVAL cost?
Is there a free plan or a free trial?
Can BiG EVAL run on our own infrastructure?
Does BiG EVAL provide an API?
Where does the AI run, and can we keep it local?
Can we drive tests from an LLM?
Which data sources are supported?
How long does it take to get started?
What languages does the product support?
Who is behind BiG EVAL, and how do we get support?
Should you pick BiG EVAL?
BiG EVAL is one of the more coherent propositions in the data quality space, and its coherence is structural rather than rhetorical. The argument that rules, AI and human knowledge each solve a problem the other two cannot runs consistently from the homepage to the pricing matrix to the documentation, and the product is organised along exactly those lines. That is rarer than it sounds.
The maturity is real and checkable. Over a decade of continuous operation, named customers in Swiss insurance and retail, and a public documentation set covering server administration, scripting, the REST API and the PowerShell module all point to software that has been deployed rather than demonstrated. The metadata-driven scaling claim is the most distinctive part of the offering, and it is the one backed by the most concrete customer numbers. The AI is functional rather than decorative: it writes validation rules, scores anomalies, reads documents and exposes an MCP connector, and it can run entirely inside the customer's network, which matters for the regulated buyers this platform targets.
The reservations are mostly about what the vendor does not publish. There is no price anywhere, in either language, so the platform cannot be shortlisted on budget without a sales conversation. There is no Data Processing Agreement and no security certification on display, which is a conspicuous silence for a product sold into insurance and banking. Evaluation runs through a proof of concept whose cost is left unstated. Public release notes stop several years before the generation currently being marketed.
For an enterprise data team that already knows it needs auditable, sovereign data quality management at scale, this is a serious candidate. For anyone wanting to try before talking, it is not.
- Choosing a selection results in a full page refresh.
- Opens in a new window.