Data Quality Statistics & Insights From Monitoring +11 Million Tables in 2025: What Monte Carlo’s Numbers Reveal
We’ve distilled Monte Carlo’s latest data-quality findings into a practical digest, no fluff, just what matters.
You can check out the full article here.
Key Takeaways
-
On average, data estates now see roughly 1 data-quality incident per 10 tables per year, a notable shift from the ~1 per 15 tables seen in prior years.
-
If you assume the typical resolution time is about 15 hours per incident, you can approximate expected data-downtime with: (tables ÷ 10) × 15 hours, emphasising how even small pipelines can carry significant business risk at scale.
-
The root causes of data-quality issues are widely distributed, revealing a fragile ecosystem rather than a single failure mode. Breakdown:
-
26.2% - pipeline execution failures (missed schedules, failed tasks, permission issues) Monte Carlo Data
-
20% - real-world data variation (i.e. shifts or changes in incoming data distribution) Monte Carlo Data
-
14.2% - intentional changes or backfills (meaning not all “incidents” are errors) Monte Carlo Data
-
Additional contributors include ingestion problems (16.6%), platform instability (15.2%), and schema drift (7.8%). Monte Carlo Data
-
This distribution shows that a considerable portion of “incidents” are not unintended bugs, but planned changes - which suggests risk comes not just from mistakes, but from complexity and change itself.
Why This Still Hurts
Modern data ecosystems are more than just tables and jobs. They’re sprawling networks: ingestion streams, batch pipelines, transformations, schema changes, manual backfills, third-party inputs, all tied together. With that scale comes fragility.
A single failed task, a subtle shift in incoming data, or an unplanned schema alteration can ripple through dashboards, analytics models, and even machine-learning workflows. That’s not just a technical hiccup, it’s a business risk: delayed decisions, incorrect conclusions, and eroded trust in data.
Worse: detection often lags. Without continuous, end-to-end monitoring, many errors are only found when an analyst or stakeholder hits a problem, meaning hours or even days of flawed business decisions may have already been made.
How Teams Are Fighting Back: What Work Experience Suggests
From Monte Carlo’s data (and what more mature data organisations report), these practices help convert data from a liability into an asset:
-
Full-lifecycle observability - from ingestion to final outputs. By monitoring at each layer (raw ingestion, pipeline, intermediate stages, published data), teams catch issues early and trace root causes fast.
-
Smart alerting and routing - avoid blast-radius alert floods. Route notifications to appropriate channels (e.g. incident management tools, not inboxes), classify by severity, and avoid overwhelming teams with noise.
-
Clear ownership & escalation paths - assign each alert or incident a responsible owner, empower “power users,” and ensure leadership maintains visibility. This keeps accountability high and response structured.
-
Automated detection over brittle tests - static SQL tests or manual checks don’t scale. Using anomaly detection, schema-lineage tracking and automated monitors reduces maintenance overhead and increases signal-to-noise.
-
Building a culture of reliability - data quality must be treated as infrastructure, not a one-off project. Teams that bake data health reviews into governance, and treat data reliability as mission-critical, tend to weather failures better.
Why Data Quality Matters Beyond Technicalities
As organisations rely more on analytics, AI, real-time decisioning and automation, data is no longer just a backend asset: it’s a central nervous system. Data quality isn’t a “nice-to-have”, it’s infrastructure hygiene.
When data is brittle, trust decays. Reports, dashboards and models built on shaky data lead to flawed decisions. At worst, stakeholders lose faith.
Treating data quality with the same discipline as uptime, performance, or security (investing in observability, clear ownership, governance) turns data into a dependable asset, not a hidden liability.