Keeping Watch on the Data Seas: 10 Rules for Effective Data Monitoring by Data Captains

Keeping Watch on the Data Seas: 10 Rules for Effective Data Monitoring by Data Captains

February 23, 2024

Table of contents

As a concept, Data Monitoring isn’t new, but is ever evolving. As organizations navigate increasingly complex data ecosystems with decreasing resources, ensuring data integrity and reliability has never been more crucial. While any ‘list’ of rules for effective monitoring of data is hardly comprehensive, let’s explore some guidelines. 

Enterprise Data Teams Need Data Monitoring and Data Review

Why even monitor data at all? Over the past decades, those in the data space referred to “data as the new oil”, or recently in Matthew McConaughey television spots as gold. Aligning data with such valuable commodities, we would be doing a disservice to ourselves if we didn’t monitor our data as we would our oil or gold reserves. Enterprise data teams serve as custodians of valuable organizational data, entrusted with maintaining its integrity and reliability. Data monitoring and review serve as the bedrock of this responsibility, enabling teams to proactively identify and rectify data quality issues, ensure compliance with regulations, and optimize data processes.

Aspects of Data Monitoring to Consider

When embarking on data monitoring initiatives, several key aspects warrant consideration:

  • Frequency: Regular monitoring intervals tailored to the criticality of data assets.
    How often should you monitor your data? That will depend on the use of the data within your organization. Consider how often data is loaded or refreshed. Is it batch or real time? Who uses it and when? What quirks or other milestone events lie in usage of your data - maybe an uptick around certain times of the year such as holidays. Taking these factors into account will help you set the right balance.
  • Metrics and What to Look For: Key performance indicators (KPIs) such as data freshness, query performance, and error rates.
    What KPIs are important to your business, and how do they affect outcomes? Freshness of your data matters, but may vary. When do processes start and finish? What times are analytics and reports produced? While you’re at it, are there places where freshness doesn’t matter in your data? Even seldomly updated tables should be “fresh” by the appropriate standard. What you do need to look at is the rotten data - data never updated or updated frequently but never used should be discarded as it has no value. KPIs associated with performance and errors, likewise are important. Don’t waste valuable resources on bad queries that suck up credits or slots. What other metrics are important to your business?
  • Mode (Automated/Manual): Striking a balance between automated alerts and manual review for comprehensive coverage.
    In an ideal world, we would have a big fat red “easy” button to monitor, review, and solve our data issues. This is not that world. Though with increasing use of AI, Machine Learning, and proactive solutions, we can approach to somewhere north of 90% automation. Automatically surfacing the “big rocks” in your data landscape with meaningful insight and information can help Data Engineers get a handle on what is happening. Automation alone would be foolhardy. It’s the job of automated monitoring to inform Engineers to perform educated executions. An expensive query may be worth the cost of business. A spike in usage may be appropriate and expected. Great observation tools surface the known knowns, the known unknowns, and the unknown unknowns. Manual review helps you make the right decisions.
  • Results/Follow-ups: Prompt action and follow-up to address identified issues and prevent recurrence.
    As GI Joe taught us: Knowing is half the battle. Knowing about a data issue, be it a stopped pipeline, runaway query, or performance problem is useless unless you take action. Once the issue is surfaced and vetted, you need to make a fix, and circle back to prevent this from happening again with the newfound knowledge. This virtuous circle of issue understanding, remediation, and prevention makes for a strong and efficient data landscape and team.

These aspects help us write the “rules” of data monitoring for the modern enterprise; what we think is a good start to building and maintaining an effective data operations program.

Transform your data observability experience with Revefi
Get started for free

10 Rules of Enterprise Data Monitoring

  1. Identify Key Metrics: Establish clear metrics and KPIs to gauge data health and quality, including data freshness and query performance.

    Before you start monitoring, understand what matters to your organization. Is it spend, quality, performance, usage, all of the above? Most likely, but the values will vary, and it is important to understand that.
  1. Define Monitoring Frequency: Tailor monitoring frequency to the criticality of data assets, ensuring timely detection of issues.

    It makes sense to start out with a blanket monitoring frequency across your data landscape. Daily is best - think of it as the Morning Data News. You won’t fully know what to monitor at which intervals until you’ve established a baseline. Tweaking this baseline over time and review will lead to the most effective design.

  2. Automate EVERYWHERE Possible: Leverage automation to streamline monitoring processes and minimize manual effort, reducing the risk of oversight.

    As much as we can POSSIBLY automate, we should. Something automated is something a person need not do; especially lower level repeatable tasks. Applied correctly, automation will yield consistent results and make fewer errors than manual processes. Time and resources are certainly conserved by a well implemented solution. Infuse machine learning into that process that allows the system to narrow in on thresholds and provide more precise insights.

  3. Combine Automation with Manual Review: Employ a hybrid approach that integrates automated alerts with manual review for thorough analysis and resolution.

    Automated insights are a powerful tool for data engineers. By bubbling up issues, trends, anomalies, and such, data workers have at their fingertips pointed outcomes with supporting information. These insights are ripe for review and execution by those that understand the landscape.

  4. Monitor Data Source Health: Regularly assess the health and reliability of data sources to maintain data integrity and reliability.
    Blindly applying a strategy and letting it go in perpetuity is seldom a path to success. Automation is absolutely an important facet of effective data operations, but it needs to be reviewed, tweaked, and maintained - like an engine. The goal again is to reduce the manual process by automatically informing data workers who can make the right adjustments and decisions.
  1. Track Data Lineage Maintain visibility into data lineage to trace data origins and ensure compliance with governance standards.

    Garbage in, garbage out. It’s as true as ever for your data pipelines and downstream assets. A solid understanding and mapping of your data pipeline combined with effective oversight and alerts safeguards against poor decisions based on bad data. Proactively understanding issues upstream of tables, views and reports allow downstream stakeholders to avoid a known problem in advance.

  2. Address Long-Running Queries: Identify and optimize long-running queries to enhance system performance and resource utilization.

    Long running queries are the perennial bugbear of data team budgeting. Through understanding which queries run long, when, and why, proper action can be taken. Perhaps a query can be rewritten to be more efficient with a second set of eyes. Maybe the query should be removed from scheduling, or run less frequently. All of these have budgetary impact. Not all expensive queries are bad. Sometimes they simply cannot be made more efficient but provide value exceeding their cost. Would you axe a $30K query that fed millions in revenue? Better to know why and ask for the right budget.

  3. Review and Improve Processes: Continuously evaluate and refine data monitoring processes based on feedback and evolving business requirements to drive continuous improvement.

    Mixing metaphors further, data is like water it flows in streams, rivers, lakes, and oceans. Streams can change direction, rivers overflow their banks, lakes grow and shrink, oceans throw up tidal waves; data acts the same way. The right data monitoring solution for your data infrastructure today may be completely wrong down the line. Constant monitoring and correction ensure a strong and efficient observation solution.

Why Revefi Data Operations Cloud is a Must for Enterprise Data Monitoring

Having explored the aspects important to data monitoring, and drafted some rules to implement it effectively, we have baked these ideas into our design at Revefi. Revefi Data Operations Cloud emerges as the most complete solution for enterprise data monitoring, offering unparalleled features and capabilities around automation, learning, and review. With Revefi, organizations gain comprehensive access to real-time insights, and actionable intelligence in data monitoring data quality, spend, performance, empowering data teams to uphold data integrity, drive informed decision-making, and unlock the full potential of their data assets.

Effective data monitoring is a must for enterprise data teams striving to safeguard data integrity and reliability amidst the complexities of modern data ecosystems. By embracing the principles and rules outlined above and harnessing the power of Revefi Data Operations Cloud, organizations can embark on a journey of data excellence, poised for success in an increasingly data-driven world.

Black-and-yellow aim icon