SAS Data Maker in Microsoft Marketplace: Privacy‑Preserving Synthetic Data for Enterprises

SAS has quietly — and now publicly — brought a synthetic-data generator into the Microsoft cloud ecosystem, a move that stitches together SAS’ recent acquisition of UK startup Hazy with the company’s long-running Viya analytics platform and Microsoft’s Marketplace distribution model. The new product, SAS Data Maker, is listed in the Microsoft Marketplace and presented by SAS as a low-code/no-code, enterprise-grade synthetic data solution that preserves statistical, relational and temporal properties of real datasets while adding privacy protections and auditable quality checks.

Background​

Synthetic data generation has emerged as a practical privacy-first strategy for model training, software testing and analytics when access to production data is constrained by regulation or corporate confidentiality. SAS’ acquisition of Hazy’s software assets in November 2024 gave the established analytics vendor proprietary capability in differential-privacy-aware synthetic data — a capability Hazy built its reputation on for financial services and healthcare customers. SAS publicly confirmed the acquisition and explicitly linked Hazy’s technology to the development of SAS Data Maker. SAS announced SAS Data Maker’s commercial availability in the Microsoft Marketplace in late November 2025, describing the product as a generator that can produce multi‑table and time‑series synthetic datasets, offering built‑in quality evaluation tools and privacy‑enhancing technologies (PETs). SAS positioned the Microsoft Marketplace entry as an initial channel, with plans to support other cloud providers and deeper integration into SAS Viya over time.

---

What SAS Data Maker is claiming to deliver​

• Enterprise-grade synthetic data generation that preserves statistical, relational and temporal structure across tables and time series while protecting identifiable information. SAS frames the product as suitable for regulated industries such as finance, healthcare and government.
• Privacy-enhancing technologies including differential-privacy mechanisms integrated into the synthesis process to reduce re‑identification risk and to help meet privacy‑regulation requirements. Hazy’s platform historically emphasized configurable differential privacy, and SAS states it has integrated that IP into Data Maker.
• No‑code / low‑code graphical interface that aims to make synthetic-data generation accessible to analysts and business users, not only to developers and data scientists. SAS’ product messaging repeatedly highlights GUI-based workflows and visual quality dashboards.
• Built‑in quality control and evaluation so users can visually compare synthetic outputs with original data and inspect metrics before promoting synthetic datasets into training or test pipelines. SAS advertises visual evaluation and statistical fidelity tools as part of the offering.
• Marketplace distribution (initially Microsoft) with integration paths into SAS Viya and other partner ecosystems (Snowflake, for example) anticipated later; SAS’ messaging emphasizes broad connector support and multi-cloud intent.

These are the core features SAS is promoting; independent coverage and pre‑acquisition materials from Hazy corroborate many of the technical claims (differential privacy, time‑series support and multi‑table relational synthesis).

---

Why this matters for enterprises and WindowsForum readers​

Synthetic data is no longer an experimental curiosity. For organizations wrestling with regulatory constraints (GDPR/UK GDPR, HIPAA), internal confidentiality or cross‑border data access rules, an enterprise-grade synthetic data product can:

• Remove legal and operational friction for model training and testing by providing privacy-safe stand‑ins for real data.
• Shorten development cycles by enabling parallel experimentation and large-scale augmentation without exposing production PII.
• Democratize access to testable datasets across product, analytics and engineering teams without constantly provisioning masked views of production systems.

SAS’ move is significant because the company brings decades of analytics trust to the market and because packaging the capability into a Marketplace app lowers procurement friction for Azure‑centric customers. Microsoft Marketplace (and related partner frameworks such as Snowflake’s native app support) makes it easier to buy, onboard and govern partner software inside a cloud tenant — a crucial operational advantage.

---

Technical reality check: what SAS says versus what independent sources show​

SAS’ marketing claims are specific in several important ways; independent sources and Hazy’s historical documentation validate most of the technical cornerstones — but with important details and caveats to note.

Verified claims​

• SAS acquired Hazy’s software assets in November 2024 and explicitly built Hazy capabilities into SAS Data Maker. This acquisition is documented in SAS press materials and PR distribution.
• SAS Data Maker is available in the Microsoft Marketplace as of November 24, 2025; the product page and SAS press release describe multi-table, time‑series and privacy‑preserving features.
• Hazy’s technology historically included configurable differential privacy controls, visual evaluation dashboards, multi‑table relational synthesis and time‑series preservation; these capabilities are well documented in Hazy materials and vendor comparisons prior to acquisition.

Claims that need operational verification​

• “Differential privacy” is a broad term: SAS and Hazy state DP mechanisms are available, but the exact privacy budget management (epsilon defaults, per‑use policy, implementation details) and whether those guarantees meet a specific regulator’s anonymization threshold must be validated in each deployment. Vendors commonly provide DP controls, but the practical privacy/utility trade‑offs depend on hyperparameters and usage patterns. Treat any DP claim as configurable and environment‑dependent rather than absolute.
• The advertised no‑code experience lowers the technical barrier, but complex, multi‑table schema design and privacy policy definition will still require data‑owner oversight. No‑code GUIs remove friction, but they do not remove the need for governance, lineage and subject‑matter review. Independent analyst coverage has described Data Maker as low‑code/no‑code, but real deployments require collaboration across IT, legal and data science.
• Performance, throughput, pricing and tenant‑level resource consumption (compute billing inside Azure, marketplace metering and any SAS licensing or managed‑service fees) are not fully published in the product press release. Enterprises must obtain sample invoices, SLOs and billing models for realistic TCO calculations. SAS’ announcements signal intent but not precise commercial terms.

---

Key risks and practical limitations (what procurement and security teams must insist on)​

• Privacy is not binary. Differential privacy mechanisms require setting epsilon and other parameters. Higher privacy (lower epsilon) typically reduces data utility. Teams must negotiate safe defaults, meet the legal team's anonymization criteria, and run independent re‑identification testing. Vendors’ DP implementations vary in scope — e.g., whether DP is applied at model‑training time, per‑field, or to summary statistics.
• Synthetic leakage and overfitting. Models that overfit can inadvertently reproduce real records. Quality reports and generation‑time audit logs are necessary to detect inadvertent leakage. Ask for formal leakage test results and automated checks in the vendor demo. Hazy’s approach combined generative modeling with DP to reduce this risk, but operational tests are the only reliable verification.
• Measurement of utility. Synthetic datasets must be evaluated for downstream model performance: accuracy, bias, fairness, calibration and drift. Visual dashboards are helpful, but quantitative metrics and test harnesses (A/B with real data where allowed) are required. SAS advertises visual and statistical evaluation tools; buyers should demand reproducible evaluation artifacts for procurement.
• Governance and lineage. Synthetic data workflows must be auditable, with lineage linking generation runs to source inputs, algorithm versions, privacy budgets and approvals. Regulatory audits will look for provenance, not just a CEO’s claim that data is synthetic. SAS highlights lineage and audit logs, but customers must validate exportability and integration with existing governance tools.
• Vendor and platform lock‑in. Deploying marketplace apps tightly integrated with a cloud ecosystem can increase data gravity and create migration friction. SAS states Data Maker is initially in Microsoft’s shop with plans for other clouds — a practical path for Azure shops, but multi‑cloud customers should insist on portability, exportable models, and documented migration procedures.
• Cost and billing transparency. Marketplace listing does not substitute for detailed pricing scenarios: Azure capacity consumption, container orchestration costs, and SAS licensing or managed-service fees add up. Procurement should model 1/3/5‑year TCO for expected generation volumes and pilot-to-production scale.

---

How to pilot SAS Data Maker responsibly: a practical 8‑step plan​

• Define the use case and success metrics. Choose a single, high‑value, low‑risk use case (for example, a non‑customer‑facing model training pipeline or software test dataset) and set measurable goals: model MAE, AUC, fairness metrics, and desired privacy thresholds.
• Inventory the source data and legal controls. Map sensitivity, residency, retention policies and regulatory constraints for the datasets you intend to synthesize.
• Establish privacy budgets and acceptance criteria. Work with legal and privacy officers to define acceptable epsilon ranges and re‑identification test criteria.
• Run a small, reproducible pilot. Generate synthetic datasets for one table or a small multi‑table join, measure utility against a holdout real dataset and record privacy metrics.
• Verify leakage testing and audit logs. Require vendor-supplied leakage tests, access to generation logs, and proof of lineage for the pilot artifacts.
• Integrate evaluation into CI/CD. Treat synthetic data generation as part of your ML pipeline: automate generation, validation, and promotion steps with versioning and approvals.
• Model deployment and monitoring. Deploy models trained on synthetic data behind production monitors for drift, fairness, and performance; establish rollback triggers.
• Negotiate exit and portability terms. Before scaling, ensure you can export generation models, synthetic artifacts, and associated metadata for migration or independent audit.

This stepwise path minimizes surprises and clarifies whether synthetic data — and specifically SAS Data Maker — delivers the expected business results under your governance rules.

---

Where SAS Data Maker fits in the broader vendor landscape​

SAS is not the only vendor in this space. Other players (Hazy pre‑acquisition, Gretel, Mostly AI, Syntho, Tonic, Gretel and others) offer synthetic data capabilities, each with trade‑offs around privacy guarantees, ease of use, supported data types and enterprise readiness. What differentiates SAS is:

• Legacy trust and vertical credibility in regulated industries. SAS’ brand and long enterprise relationships matter for procurement and compliance reviews.
• Marketplace distribution and native integration paths into enterprise data clouds (Azure, Snowflake) reduce onboarding friction for customers standardizing on these platforms.
• Breadth of analytics and governance tooling via Viya, which may make end‑to‑end data → model → decision pipelines easier to govern inside a single vendor ecosystem.

That said, niche vendors often move faster on novel features (finer‑grained privacy controls, open APIs, SDKs for iterative research) and may offer more flexible licensing for startup or product teams. Enterprises should evaluate trade‑offs between the SAS enterprise package and smaller vendors’ agility.

---

Practical checklist for IT, security and procurement teams​

• Request demonstration of DP controls including default epsilon, per‑run auditing and adjustable privacy budgets.
• Require sample generation logs, lineage exports and a runbook for reproducing synthetic datasets.
• Insist on a short proof‑of‑value pilot with both privacy and utility metrics.
• Get modeled billing scenarios for expected workloads and ask for any additional SAS support or managed‑service fees.
• Validate integration points with your MLOps stack, data catalogs and SIEM / observability platforms.
• Negotiate contractual language on portability, exportability and data handling (including obligations in the event of supplier insolvency).

---

Critical perspective: strengths, but don’t ignore the trade‑offs​

SAS Data Maker’s arrival in the Microsoft Marketplace is a technically credible and commercially sensible step: it couples established synthetic‑data technology with a well‑known enterprise analytics vendor and delivers a marketplace‑friendly procurement path. The combination of Hazy’s differential‑privacy‑aware algorithms with SAS’ Viya platform and Microsoft distribution makes the offering attractive for Azure‑centric enterprises and regulated sectors that value vendor assurances.
However, synthetic data is not a turnkey anonymization bullet. The real work is operational: selecting privacy budgets, validating synthetic dataset utility, building lineage and auditability into pipelines, and maintaining governance discipline across teams. Marketplace availability reduces procurement friction, but it does not remove the need for rigorous engineering and legal controls. The exclusivity window (SAS’ initial listing in Microsoft Marketplace) appears to be an initial channel strategy rather than a long‑term lock — SAS says it plans to expand to other clouds and integrate Data Maker into Viya — but timelines and commercial terms for other clouds remain subject to confirmation. Therefore, multi‑cloud customers should verify portability and migration paths before deep adoption.

---

Bottom line and recommended next steps​

SAS Data Maker is now a viable option for enterprises that need privacy‑aware synthetic data and prefer an Azure‑native procurement path. The buying case is strongest when:

• You operate in regulated industries with strict controls on production data access.
• You already use Microsoft Azure / Microsoft Marketplace and want a marketplace-managed onboarding model.
• You need integrated evaluation, lineage and an enterprise vendor you can hold accountable.

Recommended immediate actions:

• Initiate a scoped pilot that compares models trained on synthetic data from SAS Data Maker against a small real‑data baseline.
• Demand demonstrable privacy budget settings, leakage tests and exportable lineage for regulatory audit purposes.
• Obtain clear TCO modeling for expected generation volumes and productionization scenarios.
• Plan governance: decide who owns privacy budgets, approves generation runs, and signs off on production promotion.

SAS’ entry into the Microsoft Marketplace with Data Maker marks a maturing synthetic‑data market and gives enterprises a familiar vendor option. The technical foundations (differential privacy, relational and time‑series synthesis, visual evaluation) are credible, but the responsibility for ensuring safe, lawful and useful synthetic data rests with customers and their integration practices. Practical pilots, transparent metrics and contractual guardrails will determine whether synthetic data becomes a productivity multiplier — or an under‑governed source of risk.

---

SAS’ announcement and marketplace entry consolidate a clear vendor story: acquire proven synthetic‑data IP (Hazy), productize it into a low‑friction offering (Data Maker) and make it available where enterprise buying happens (Microsoft Marketplace), while promising multi‑cloud reach and Viya integration later. For organizations building AI on Windows and Azure platforms, Data Maker is worth evaluating now — just not without the measurement, governance and legal checks that modern AI systems demand.

---

Source: Brandsit SAS brings synthetic data generator to Microsoft ecosystem - Brandsit