What do users say?

An auditor asked us to reproduce a report exactly as it looked on one date three months ago. In the past, that meant rebuilding old data and rerunning pipelines for days, and still wondering if the result matched. With Delta tables in Databricks, I queried the table as it was on that date and had the same numbers in minutes. I also use the lineage view a lot. When a figure looks wrong, I can see which upstream tables feed it, instead of asking five people. Review collected by and hosted on G2.com.

Time travel only goes back as far as you keep history, so I had to change our retention settings before it covered the period the auditors cared about. Clusters left running also made our bill jump once, and now I set them to shut down on their own. Review collected by and hosted on G2.com.

I like Databricks because it provides a flexible environment for building and maintaining data transformation pipelines at scale. I can use PySpark and Spark SQL to handle complex transformations, process large datasets, and implement incremental data loads. Delta Lake is particularly useful for managing reliable tables and handling updates without relying on multiple disconnected process tools. It helps ETL development by bringing data processing transformation and pipeline management into a unified workspace. I can consolidate data from multiple sources, apply business rules, perform data quality checks, and prepare curated datasets for downstream analytics. Its features like PySpark, Spark SQL, Delta Lake workflow scheduling, and Medallion architecture make Delta tables useful for reliable data updates. It provides a convenient way to develop, test, and ETL logic. Notebook supports collaboration, and job monitoring helps track pipeline executions. Its integration with cloud storage, orchestration tools, and data warehouse, combining Azure Data Factory, Azure Data Lake Storage, and Databricks, is useful for creating end-to-end ETL pipelines. Review collected by and hosted on G2.com.

It has a learning curve, especially when working with advanced cluster configuration, access control, and performance tuning. Troubleshooting inefficient Spark jobs can take time when dealing with skewed data or memory issues. Compute costs also require attention when workspaces are not configured efficiently. Review collected by and hosted on G2.com.

I like Databricks because it gives me a practical environment to work with data processing and transformation using Python, PySpark, and SQL. I can explore datasets, clean inconsistent records, remove duplicates, and prepare structured data for reporting. I also find the notebook environment helpful because I can test my code step by step and understand the result immediately. I like notebooks, PySpark DataFrames, Spark SQL, Delta Lakes, and job scheduling. Notebooks make development easier, while Delta Lake helps to manage reliable tables and supports operations such as updates and incremental processing. These features are useful for learning real-world data pipeline development. The workspace is well-suited for writing and testing code in notebooks; running individual cells and viewing output makes debugging easier. However, it works well with cloud storage, data orchestration tools, and other data platforms. Connecting it with Azure Data Lake Storage and Azure Data Factory is useful for building end-to-end data pipelines. Review collected by and hosted on G2.com.

The platform has many features, so it can take time to understand the workspace compute settings, permissions, and job configuration. Clustering starting times can sometimes interrupt development, and I need to be careful with compute usage to avoid unnecessary costs. Review collected by and hosted on G2.com.

I like that Databricks lets me apply complex transformations and rely on a single source of truth. I value being able to share data with my team so they understand my business logic. Automatic scaling is valuable at the end of each quarter for in-depth customer churn and upsell analysis, and it helps high-volume processes run quickly without the system crashing. Setup was very easy in our environment. Review collected by and hosted on G2.com.

I think Databricks’ interface could be improved because it seems too focused on machine learning and generic analytics. I also have to keep a close eye on costs, especially toward the end of the billing cycle, because usage can increase significantly. Review collected by and hosted on G2.com.

The best thing that I admire about Databricks is the strong data expertise that it has provided us, also they have helped us to bridge the gap between our engineering and. ML along with BI tracks. It's has helped us to speed up the delivery timelines by 39.89% Also With the ETL pipelines has helped us to deploy without upfront infra provisioning. We have seen a change in collaboration since teams are able to collaborate better by 58% since it has vc integrations and multi language support such as (Pyspark, SQL & Scala). Review collected by and hosted on G2.com.

There are things which I feel that could be improved in Databricks, first thing is the consumption based model can escalate quickly and it is something not suitable for small and mid businesses. Second is learning curve it is hard to learn and it has complex architecture involved in it. Third, the integrations and dashboards can be better and more intuitive. Review collected by and hosted on G2.com.

I like that Databricks brings our finance exports, timesheets, pipeline, and Revit quantity schedules into one governed layer, so I can connect fee performance with resourcing without the weekly reconciliation chore, and our post-occupancy energy models use those same tables. Autoscaling on Delta Lake work is what sold me. With Photon, aggregations across years of project and utilization history run fast enough that I’ll actually iterate on a model instead of kicking it off and going to make coffee. Unity Catalog is the other piece I’d fight to keep: partners see exactly the tables they should and nothing else, and the lineage view helped settle a disagreement about a margin figure that had been going in circles for a fortnight. Review collected by and hosted on G2.com.

Git integration for notebooks leaves a good deal to be desired. When two of us touch the same notebook, merge conflicts take real effort to untangle, so I have taken to claiming files outright. I’d like cell-by-cell conflicts with a proper side-by-side diff and the option to resolve them in the workspace. Large result sets in workspace cells can also lock up the tab, so I sample before displaying anything substantial. Setup was steep at the start, mainly getting Unity Catalog structured properly. Review collected by and hosted on G2.com.

What I like most about Databricks is how its user friendly UI brings the entire data engineering workflow into one platform. In my project, data lands in the Bronze layer through snaplogic, and we use databricks to transform it into Silver and Gold data products. The serverless compute option has significantly reduced the effort of managing infrastructure, while Unity Catalog makes governance and access control straightforward. I also find AI Genie and the built-in monitoring features useful for investigating pipeline failures, checking job runtimes, and debugging issues much faster.From a pricing perspective, the pay-for-what-you-use model works well for us, especially with serverless compute and auto-scaling, as we avoid paying for idle infrastructure. The documentation is comprehensive, onboarding new team members is relatively straightforward, and there is a strong knowledge base and community that helps resolve issues quickly. Overall, it has made our ETL development and day 2 day operations more efficient. Review collected by and hosted on G2.com.

One downside is that troubleshooting can sometimes be difficult when a pipeline fails, as the error messages aren't always detailed enough and you may need to dig through multiple logs to find the root cause. The platform also has a lot of features, so it can take some time for new users to become comfortable with everything. While serverless is convenient, costs can increase if compute usage isn't monitored properly. I'd also like to see faster UI responsiveness in some areas, especially when navigating large job histories or catalogs. Review collected by and hosted on G2.com.

The best thing I like about Databricks is its overall end-to-end experience which makes it different from the other solutions. As a project manager, we had used Databricks in our organization for almost 3 years. In those 3 years, we have used it for its unified data lake for data warehouse. Also, it has helped us to reduce the time to market for our projects by almost 45%. With its unified environment, it has helped our teams to reduce friction and to move at a faster pace. Review collected by and hosted on G2.com.

There are no particular things that I dislike, but they may improve on auto-scaling compute since it can get small companies in trouble as it may create a budget overhead. Also, it is a code-heavy platform, which also means that non-tech users can't use it much. Review collected by and hosted on G2.com.

I've been using Databricks as part of our data engineering workflow to build and maintain ETL pipelines, analyze large datasets, and support reporting requirements. One of the things I like most is that it brings data engineering, analytics, and notebooks into a single workspace. Instead of switching between multiple tools, I can write PySpark code, validate transformations, collaborate with teammates, and schedule jobs from the same platform. This has made day-to-day development more organized, especially when working on multiple data pipelines.

Another feature I rely on frequently is the notebook environment. It's convenient for developing and testing transformations before moving them into production. During development, I often use notebooks to inspect sample data, troubleshoot failed transformations, and validate business logic with SQL and PySpark. The ability to mix code, markdown documentation, and query results in one place also makes it easier for team members to understand the implementation during code reviews or knowledge transfer sessions.

I also appreciate the platform's scalability. Some of our data processing jobs involve millions of records, and Databricks handles distributed processing efficiently without requiring us to manage the underlying infrastructure directly. Features like cluster management, job scheduling, and integration with cloud storage reduce operational overhead. That said, cluster startup times can occasionally delay quick debugging sessions, and managing compute resources carefully is important to avoid unnecessary costs. Overall, Databricks has helped simplify large scale data processing while giving enough flexibility for both development and production workloads. Review collected by and hosted on G2.com.

While Databricks has been reliable for our data engineering workloads, there are a few areas where I think it could be improved. One challenge I've experienced is cluster startup time. When I only need to test a small code change or validate a transformation, waiting for a cluster to start can interrupt the development flow. It's not a major issue for scheduled production jobs, but during active development and debugging, those extra minutes add up.

Another limitation is cost management. Since compute resources are tied to cluster usage, it's important to monitor cluster configurations and ensure they are shut down when not needed. We've had situations where development clusters remained active longer than expected, resulting in higher cloud costs. The platform provides tools to manage this, but it still requires teams to establish good governance and usage policies. I also found that some configuration settings for jobs, permissions, and clusters have a learning curve, especially for new team members who are unfamiliar with the Databricks environment.

From a day to day perspective, debugging distributed Spark jobs can sometimes be challenging. While the logs provide useful information, identifying the exact cause of failures often requires navigating through multiple execution logs and Spark UI details. For straightforward issues this isn't a problem, but troubleshooting more complex pipeline failures can take time. Despite these limitations, none of them outweigh the benefits of the platform, and most challenges can be managed with proper cluster configuration, monitoring, and team practices. Review collected by and hosted on G2.com.

I use Databricks to run ingestion, transformation, and reconciliation pipelines for several clients’ data on one platform, and to prepare feature tables for fraud detection. The setup was smoother than I expected: my workspaces, cluster policies, and Unity Catalog were ready within the first sprint. Genie Code drafts and debugs my PySpark, so I can spend more time on edge cases than boilerplate, and it often reads a job’s stack trace faster than I do. It also helped me pinpoint a provider file-format mismatch, so I could add an expectation and get reconciliation reports to the client’s finance team on time. Separate workspaces keep client projects apart, cluster policies help prevent oversized clusters on client budgets, and Unity Catalog gives me one place to manage table permissions and trace lineage when a source change breaks something downstream. Databricks helps me fix schema changes and failed jobs quickly so clients receive reliable data without delays. Review collected by and hosted on G2.com.

My main reservation is the pipelines dashboard. Since I deliver for several clients, their pipelines end up in one flat list and blur together. Assigning permissions pipeline by pipeline is tedious and, with client data involved, leaves too much room for error. Folders with inherited access, tags for client and environment, filters, and a summary of failed or delayed runs would make it easier to check one client’s pipeline health at a glance. Review collected by and hosted on G2.com.