Impact of Poor Data Quality
Picture a VP opening Monday morning’s executive dashboard and making a six-figure call off a number that has been quietly wrong for three weeks. Every pipeline ran green. No alert fired. The data was simply bad – and nobody knew. Decision-makers lean on dashboards, ML model outputs, and conversational analytics agents to grow the business, but trusting that data blindly is costly. A 2020 Gartner study estimates that poor data quality costs companies $12.9 million a year on average (Data Quality: Best Practices for Accurate Insights, n.d.). Without any insight into the quality of the data being used, the tools that these decision-makers rely on are ticking time bombs that can severely hurt the bottom-line.
What Data Monitoring Functions Actually Do
The typical way to monitor data quality is to bolt-on an external tool, such as Great Expectations or Monte Carlo, or to write custom checks, which take time to define, write, and test. A more recent Snowflake feature solves these problems with a workflow that enables fast prototyping and rapid customization of a data quality solution. Data monitoring functions (DMFs) are a Snowflake-native solution with many built-in metrics that keep everything on the Snowflake platform. No need to bring in external tools that add complexity to the data stack, and no need to manually write and test common DQ checks. Now everything that’s needed is built-in.
A yet even more recent feature of Snowflake is CoCo, a rebranded Cortex Code with a new suite of features announced at Snowflake Summit. CoCo is a Snowflake-native agentic coding tool, available as a standalone IDE or wherever it is you already do your work in Snowsight, VS Code, or a terminal CLI. CoCo comes with built-in awareness of its Snowflake environment and governance, so its context will already include existing data and infrastructure within a Snowflake account. Better yet, it has Snowflake’s own best practices and patterns built-in as well in the form of skills, so out of the box, it is well familiar with the ideal way to build in Snowflake.
The Business Challenges DMFs Solve
Most data teams are already sitting with some version of these problems when they start looking at native monitoring:
Tool sprawl and stack complexity. Bolt-on monitors like Great Expectations or Monte Carlo are one more system to integrate, secure, and license. Native DMFs keep quality checks inside Snowflake, where the data already lives.
Slow time-to-coverage. Hand-written checks have to be defined, written, and tested before they catch anything. The Cortex Data Quality wizard profiles a table and recommends checks in minutes, collapsing days of setup into a single session.
Blind spots in the data feeding AI. Conversational analytics and ML outputs are only as trustworthy as their inputs. DMFs attach measurable quality signals – null counts, freshness, uniqueness – to the very tables those agents read from.
Governance fragmentation. External tools sit outside Snowflake’s access and governance model, creating a second place to manage permissions and a gap auditors notice. DMFs and CoCo operate inside the account’s existing governance.
Initial DMF Setup
A first-pass setup of DMFs is as simple as a few clicks within Snowsight. After navigating to the “Data Explorer” page and selecting a table, there will be a “Data Quality” tab. Here you’ll find an option to add checks, either manually or with the Cortex Data Quality wizard. This is an AI-powered feature that profiles and analyzes your data, then provides suggestions on what DMFs to apply along with explanations of why those suggestions would be helpful. A human-in-the-loop then looks over these recommendations and selects which ones to apply and which to discard. This allows for the rapid defining of common types of data quality checks – such as null counts or expected values – with no external tools and no SQL. In a matter of minutes, you’ve gone from flying blind to scheduled monitoring checks.
The Next Level: Cortex Code and Your Business Needs
The next step might normally be designing, writing, and validating custom SQL DMFs and checks based on your own business rules and requirements, but no more. CoCo comes equipped with Snowflake’s own best practices for implementing both built-in and custom DMFs. Because of this, it will outperform more generic agentic coding tools, which are much more general-purpose tools that can write code for a Snowflake environment, but seem amateurish compared to CoCo and its encoded knowledge of how best to utilize Snowflake’s vast feature-set.
This workflow can further be customized to specific business needs, as Cortex Code allows for creating a custom skills directory to expand the Snowflake built-in skills with patterns and architectures that follow your own standards and practices. This, combined with existing functionality, allows engineers and analysts to build an agentic coding workflow that is tightly tailored to the needs of the data pipeline and business. All of this connects straight back to Snowsight, which shows the results of the monitoring and checks visually.
Three Snowflake capabilities make this practical at production scale: native scheduling, so DMFs run on a set cadence without external orchestration; built-in system metrics that cover the most common checks – null counts, freshness, row-count drift – out of the box; and CoCo’s environment-aware governance context, so generated code already respects the account’s data and access model.
Where This Fits in the 7Rivers Data Native® Model
Data quality monitoring lives at the Foundation stage of the 7Rivers Data Native Model® – the layer everything else is built on. Clients typically arrive here with pipelines already running but no systematic way to know whether the data moving through them is correct; quality is assumed until a bad number surfaces in a board deck. As a Snowflake Elite Partner, 7Rivers establishes that Foundation by standing up native DMFs and CoCo-driven checks, so the Insights and Actions stages that follow are built on data the business can trust.
Trusted Data, Trusted Insights
With full visibility into the quality of your data, users will have full trust in the insights that they see from whatever agent is using that data as context. Traditional dashboards and Snowflake CoWork alike provide these insights based on the semantic layer built with the data model as its foundation. This allows users to focus on actions instead of double and triple checking their insights, hoping that they can trust them. Snowflake CoCo and DMFs provide the assurance of quality data, so businesses can grow with confidence. Native data quality monitoring is fast becoming the default expectation for any serious Snowflake deployment, and the teams that adopt it now are the ones whose AI agents will be trusted tomorrow.
References
Data Quality: Best Practices for Accurate Insights. (n.d.). Retrieved from Gartner: https://www.gartner.com/en/data-analytics/topics/data-quality
