Improve data reliability by shifting left

How to Shift-Left Data Reliability

The Data Observer
6 min readMay 9, 2023


As organizations rely more heavily on data analytics for decision-making, the amount of data being captured and fed into analytics data stores has increased substantially. Reliable data is crucial to ensure that enterprises can make informed decisions based on accurate information. Data comes from various sources, including both internal applications and repositories, as well as external service providers and independent data producers.

For companies that produce data products, external sources typically provide a significant percentage of the data. Since the end product is the data itself, it is critical to bring together high-quality data that can be trusted. Shifting left in the approach to data reliability is essential, but it is not a switch that can be turned on immediately. Data Observability plays a crucial role in shaping data reliability, and the right platform is necessary to ensure that only good, healthy data is entering the system.

High-quality data can help organizations gain a competitive advantage and deliver innovative, market-leading products continuously. Poor quality data, on the other hand, can lead to bad outcomes and broken businesses. Data pipelines that feed and transform data for consumption are increasingly complex and can break at any point due to data errors, poor logic, or insufficient resources to process the data. Therefore, the data team’s challenge is to establish data reliability as early in the data journey as possible and create optimized data pipelines that can perform and scale to meet an enterprise’s business and technical needs.

Data reliability operates in one of three areas within the data pipelines that manage data supply chains, including the data landing zone, the transformation zone, and the consumption zone. Traditionally, data quality tests were only applied in the final consumption zone due to resource and testing limitations. Modern data reliability checks data in any of these three zones and monitors data pipelines moving and transforming data.

Why is Data Reliability Essential?

The cost of fixing problems in data pipelines follows the “1 x 10 x 100 Rule,” which applies to the cost of fixing problems at different stages of the process. It is far more cost-effective to fix data problems as early as possible. For every $1 it costs to detect and fix a problem in the landing zone, it costs $10 to detect and fix it in the transformation zone, and $100 to detect and fix it in the consumption zone. Early detection of data incidents in the supply chain helps data teams optimize resources, control costs, and produce the best possible data product.

Data supply chains have become increasingly complex, which is evident in the increasing number of sources feeding data, the complexity of logic used to transform data, and the number of resources required to process the data. Data reliability checks were traditionally performed only in the consumption zone. Today’s best practices require data teams to “shift-left” their data reliability checks into the data landing zone to effectively manage data and data pipelines.

Shift-left Data Reliability

By shifting-left data reliability, data incidents can be detected earlier, leading to faster correction. This approach also prevents bad data from reaching downstream stages where it could be consumed by users and result in poor decision-making. Applying the 1 x 10 x 100 rule, early detection allows for efficient correction at the lowest possible cost ($1). Conversely, if data issues spread downstream, they can impact more data assets and result in far greater costs to correct ($10 or $100).

The Path to Shifting-Left Your Data Reliability

For a data reliability solution to effectively shift-left requires a unique set of capabilities and features, which include:

  • Perform data reliability checks before data enters the data warehouse and data lakehouse:
  • Keeps bad data out of the transformation and consumption zones.
  • Support for data-in-motion platforms:
  • Supports platforms like Kafka
  • Monitors data pipelines in Spark jobs or Airflow orchestrations
  • Allows data pipelines to be monitored and metered.

It should also include:

Support for files:

  • Files often deliver new data for data pipelines
  • Checks on various file types
  • Capturing file events to know when to perform incremental checks is important.


  • APIs that integrate data reliability test results into data pipelines
  • Allows pipelines to make decisions to halt data flow when bad data is detected
  • Prevents bad data from infecting other data downstream.

Data isolation:

  • When bad data rows are identified

Prevents continued processing:

  • Isolates bad data
  • Ability to have further checks run to dig deeper into the problem.

Data reconciliation:

  • Ability to perform data reconciliation
  • Keeps the same data in multiple places in sync.
Data Pipeline Monitoring in Acceldata

Continuous Data Pipeline Monitoring

Continuous monitoring of data pipelines is essential for detecting issues early and maintaining healthy and properly flowing data. In order to achieve this, a consolidated incident management and troubleshooting operation control center is necessary to provide data teams with continuous visibility into data health and enable them to respond rapidly to incidents. To further support continuous monitoring, data reliability dashboards and control centers should have the capability to:

  • Instantaneously offer 360-degree insights into data health.
  • Provide alerts and incident information as they occur.
  • Integrate with popular IT notification channels such as Slack.
  • Allow data teams to drill down into incident data to identify the root cause.
Data Health Alerts and Incident Management in Acceldata

Identify / Prevent Data Issues

To quickly identify the root cause of data incidents and remedy them, data teams need as much information as possible about the incident and what was happening at the time it occurred. Acceldata offers correlated, multi-layer data on data assets, data pipelines, data infrastructure, and the incidents at the time they happened. This enables data teams to:

  • Perform root cause analysis of any incident and make adjustments to data assets, data pipelines, and data infrastructure accordingly.
  • Automatically re-run data pipelines when incidents occur, which allows them to quickly recover.
  • Eliminate bad or erroneous data rows to keep data flowing without the low quality rows.
  • Compare execution, timeliness, and performance at different points in time to see what’s changing.
  • Perform time-series analysis to determine if data assets, pipelines, or infrastructure is fluctuating or deteriorating.

Prepare Your Data Environment for Shift-Left Data Reliability

Shifting left your data reliability helps your data teams detect and resolve issues earlier in a data pipeline and prevents poor-quality data from flowing further downstream. This approach can offer a range of benefits, including:

  • Frees data teams from firefighting data issues, allowing them to focus on innovation.
  • Lowers the cost of fixing data issues according to the 1 x 10 x 100 rule, reducing overall data engineering costs.
  • Ensures that low-quality data does not reach the consumption zone, increasing business teams’ trust in the data they use.
  • Keeps data pipelines flowing properly to maintain timely and fresh data, facilitating agile business processes.

Get a demo of the Acceldata Data Observability platform and learn how to shift-left data reliability for smooth and effective data reliability processes.

Photo by Markus Spiske on Unsplash