Enterprise leaders recognize synthetic data's potential for protecting sensitive information and accelerating development, yet adoption remains constrained. Recent research from Perforce Delphix reveals that data quality concerns—not awareness or compliance gaps—are the primary obstacle. Only 38% of leaders believe synthetic data achieves sufficient realism, and just 34% trust it to maintain referential integrity across systems.

This trust deficit has real consequences. Software engineering and testing teams deploy synthetic data at surprisingly low rates, leaving significant untapped opportunities on the table. Organizations struggling with synthetic data implementations often discover that generated datasets lack the realism or relational consistency needed to improve application development and testing workflows.

The appeal and the reality gap

Synthetic data addresses a longstanding organizational tension: teams need access to realistic test data while protecting sensitive information. Traditional solutions force a choice between restricted access that slows development or broad distribution that multiplies compliance and security risks. More than half of enterprises (51%) already use synthetic data alongside masking techniques to safeguard sensitive data.

The concept is compelling, especially as AI agents drive demand for readily available, high-quality datasets across organizations. Yet generating synthetic data is not equivalent to generating useful synthetic data. Many organizations have experimented with synthetic data only to find that it lacked the realism or relationship preservation needed to meaningfully improve testing outcomes. Adoption hinges on whether organizations believe that modern, AI-first approaches can address these legacy shortcomings.

What development teams actually need

Executives and development teams evaluate synthetic data through different lenses. While leadership focuses on governance, developers prioritize utility. Currently, 66% of software leaders report that their teams do not use synthetic data in development and testing.

Developers and testers don't primarily ask whether data is protected. They are asking whether it is useful.

Development teams ask whether synthetic data can accurately model customer behavior, support complex testing scenarios, preserve relationships across interconnected systems, and provide confidence that applications tested today will perform correctly in production. Synthetic data must go beyond being technically fake—it must retain the context, relationships, and realism that make data meaningful.

Consider a credit rating field: generating a random number may be technically valid, but it fails to reflect actual data distributions. Generating values within real-world bounds and ranges upholds both realism and quality. When synthetic data lacks realism or misrepresents system behavior, teams risk validating applications against scenarios that never occur in production, leading to defects, failed releases, and inaccurate test results.

Building trustworthy synthetic data

For synthetic data to support enterprise development at scale, it must behave like the systems it represents. Enterprise applications rarely operate within a single database. Customer information may reside in one system, orders in another, and payment details in a third. Testing succeeds only when those relationships remain intact across environments. Without cross-system consistency, teams test isolated components rather than real business processes.

Enterprise applications rarely operate within a single database. Testing only works when those relationships remain intact across environments.

Organizations evaluating synthetic data tools should treat these capabilities as essential:

  • Maintain relationships across databases and systems, not just within a single schema. Enterprise applications depend on cross-system dependencies that synthetic data must reflect—for example, investor and investment account records must always be linked.
  • Generate data in dependency order to keep parent-child relationships intact throughout the environment.
  • Validate realism and business logic by ensuring generated data follows expected formats, ranges, distributions, and rules. Test datasets should never produce impossible scenarios such as invalid identifiers or illogical business outcomes.
  • Continuously assess data quality as schemas evolve. If synthetic data requires extensive manual reconfiguration whenever a source system changes, quality and trust deteriorate over time.

Meeting these requirements allows synthetic data to remain useful for development while reducing data privacy risk.

AI acceleration and the data bottleneck

AI-assisted development is bringing a persistent challenge into sharp focus: software delivery can only accelerate as quickly as the data supporting it. Code generation has advanced dramatically, but access to high-quality test data has not kept pace. Synthetic data can bridge this gap by being generated on demand, unlike production data.

This capability proves especially valuable when teams need purpose-built data for new applications, features, edge cases, and AI-driven initiatives where production data may not exist or may be difficult to use. Synthetic data enables teams to construct realistic testing scenarios on demand and move validation earlier in the development lifecycle.

Software delivery can only move as fast as the data supporting it.

Synthetic data is not necessarily a universal replacement for all testing approaches. Organizations increasingly find value in combining both: using masked data for known workflows while deploying synthetic data to expand coverage and test entirely new scenarios. Together, these approaches deliver both realism and flexibility. The common denominator is trust—teams testing new features, validating integrations, or supporting AI-assisted development all depend on data that accurately represents real-world system relationships and context. Referential integrity has shifted from a technical convenience to a prerequisite for meaningful synthetic data adoption.

Trust as the adoption catalyst

The next wave of synthetic data adoption will be determined by whether engineering teams trust the data enough to build, test, and innovate with it. As software delivery accelerates toward agentic speed, trusted data becomes a competitive advantage. Organizations need data that can be generated quickly while accurately reflecting real-world relationships, business processes, and system behavior.

Ultimately, synthetic data adoption isn't an awareness problem. It's a trust problem, and trust comes from data quality.

Organizations that close this gap will not be those that simply generate more synthetic data. Instead, they will be those that generate realistic, referentially accurate, fit-for-purpose data that developers can confidently use to accelerate releases and keep pace with AI.