Azure Databricks Reviews & Product Details

Having everything—ETL, SQL analytics, and machine learning—in one platform has saved our team a massive amount of time. The Delta Lake integration is super reliable, and Photon makes running heavy Spark jobs significantly faster. I also love that our data engineers and data scientists can collaborate in the exact same notebooks using PySpark and SQL without constantly exporting data back and forth. Unity Catalog has also made managing permissions across datasets much easier. Review collected by and hosted on G2.com.

The pricing can scale up surprisingly fast if you don't keep a close eye on compute clusters and auto-scaling rules. It also has a bit of a learning curve for team members who aren't as comfortable with Spark or cloud infrastructure. Occasionally, the workspace UI can feel a little sluggish when navigating through lots of folders and old notebooks. Review collected by and hosted on G2.com.

Excellent ROI, with qualitatively defined metrics Review collected by and hosted on G2.com.

Lengthy implementation with some risks on the customer Review collected by and hosted on G2.com.

What I like most about Azure Databricks is how it makes working with data feel straightforward without me having to overthink the setup.

From my experience, I mostly use it for querying, transforming, and validating data, and it handles large datasets really well without slowing me down. I don’t have to worry too much about performance — I just write what I need, and it runs.

I also like the flexibility of switching between SQL and PySpark depending on what I’m doing. It makes it easier to explore data and troubleshoot issues quickly without being stuck in one approach.

The notebook environment is another big plus for me. I use it to organize my queries and logic in one place, so I can always go back, adjust things, or reuse parts without starting from scratch.

Overall, it just makes my workflow cleaner and more efficient, especially when I’m working with large volumes of data and need quick, reliable results. Review collected by and hosted on G2.com.

What I dislike about Azure Databricks, based on how I’ve used it, is mostly tied to day-to-day usability.

When I’m working with files (especially around /dbfs), I sometimes run into seemingly random errors that aren’t very clear. It takes extra time to figure out what actually went wrong, which is frustrating when I’m just trying to get quick results.

Debugging is another area that can slow me down. If a query or transformation doesn’t behave as expected, it isn’t always obvious where the issue is, so I end up spending more time tracing and narrowing things down than I’d like.

The notebook environment is useful, but as a single notebook grows, it can get messy and harder to manage. If I’m not careful, it’s easy to lose structure and organization.

Cost is also something I’ve had to keep an eye on. Even when I’m only testing or running queries, usage can add up quickly if resources aren’t managed properly.

Overall, it works well, but there are still moments where it feels less intuitive than it should—especially when something goes wrong. Review collected by and hosted on G2.com.

I love Azure Databricks most for its seamless collaborative notebooks—teams can code, visualize, and iterate together in real-time without hassle. The autoscaling Spark clusters handle massive data effortlessly, so no more babysitting resources. Integration with Azure services like Data Lake feels native and smooth. Delta Lake keeps everything reliable and versioned. Plus, built-in ML tools make model building a breeze. It's like a powerhouse that just works for data pros. Review collected by and hosted on G2.com.

The primary downsides of Azure Databricks are its high and often unpredictable costs, as the combination of DBU units and underlying Azure VM fees can escalate quickly without strict governance. Additionally, the long cluster cold-start times can be frustrating for developers used to the instant responsiveness of serverless environments, often leading to wasted time or expensive "always-on" configurations. Finally, the steep learning curve required to manage Spark optimizations and complex security integrations can feel overkill for smaller teams that just need simple data processing. Review collected by and hosted on G2.com.

I'm a Senior Data Scientist at a mid-sized company where I work a lot with Azure Databricks to connect data in our production machine learning models to the raw data lakes. My team uses it every day to clean huge amounts of user behaviour data, build complex features and develop predictive models, our Customer Churn risk and Product recommendation models.I love how black and white it's (almost) the headache of big data infrastructure is eliminated. Prior to adopting Databricks, creating and maintaining a Spark cluster had been a tedious project, which necessitated significant data engineering support. Now I simply choose my compute size from a dropdown and head straight to writing code in PySpark, Python or SQL in a collaborative notebook. My team's dream has also come true with the native integration of MLflow. Previously, recorded model version, parameters and metrics was stored in handwritten spreadsheet files and pickle files. This is now the case for every experiment, and automatic logging puts comparisons of model runs or reverting back to an earlier version a breeze. It's also cool that it integrates with our Azure Data Lake storage without having to go through security hurdles to mount data onto it. Review collected by and hosted on G2.com.

The worst thing is that the bill can add up quickly if you're not managing your compute usage. Auto-scaling is a godsend for performance but we do have some serious billing shock-and-tumbles when clusters were scaling up feverishly with some badly written query left running, that could have been avoided. Another one of the frustrations is the StartUp Time for Clusters. The minutes required to spin up a cluster just to run an ad-hoc data check, and then wait for the results can really bring you out of the flow. Lastly, the Spark UI can be a bit confusing and unwelcoming for newer, less technical members of the team who are simply looking to run some simple SQL queries, instead of getting into spark configuration or compute policies. Review collected by and hosted on G2.com.

I use Azure Databricks to build and manage data pipelines. It provides all required services in a single place, like data engineering, SQL, and ML features. It helps me simply process large-scale data for enterprise projects, making Azure Databricks a valuable tool for me. The SQL features make it easy to query and analyze data quickly, and the ML capabilities support experimenting with models on the same platform. The initial setup is very easy; you just need to create a resource on the Azure portal by entering the resource group and Databricks workspace name with the rest of the default settings. Review collected by and hosted on G2.com.

Cost optimization: it can be more optimized by providing the single cost monitoring dashboard by default for the workspace admins, as they have this budget feature in the preview for the account console only. Review collected by and hosted on G2.com.

It makes data easy and simple to understand even if we are not from technical background, i dont need to swap or switch different app or software now for data eng , analytics or data science all can be done in once now. The interface is very good and user friendly easy to understand tabs given , i have tried uploding a large set of data , uploading experience was very smooth and easy Review collected by and hosted on G2.com.

Mostly, I got confused during the cluster setup. It was very difficult for me, and even with the settings I’m still struggling with it. Review collected by and hosted on G2.com.

What I like best about Azure Databricks is how well it combines scalability, speed, and collaboration in a single environment. It makes it easy to work with large datasets, build and run data pipelines efficiently, and support both engineering and analytics tasks without switching between too many tools.

I also appreciate how smoothly it integrates with the broader Azure ecosystem, which makes it especially useful for end-to-end data processing and analytics workflows Review collected by and hosted on G2.com.

One thing I dislike about Azure Databricks is that it can feel complex and not always immediately intuitive, especially at the beginning. The environment is powerful, but that also means there are many concepts, configurations, and moving parts to get used to before it feels really smooth.

Another drawback is that, for some tasks, the setup and navigation can feel heavier than expected, which slows down simple workflows. In short, it is a very capable platform, but the learning curve and operational complexity can make it less straightforward than I would like. Review collected by and hosted on G2.com.

What I like best about Azure Databricks is how it simplifies large-scale data processing while still giving flexibility to engineers. From my experience, the biggest advantage is the unified platform I can do data engineering, transformations, performance tuning, and even analytics in one place without jumping across multiple tools. The integration with Spark is seamless, and things like auto-scaling clusters, job scheduling, and notebook collaboration make day-to-day work much more efficient. I also appreciate features like Delta Lake handling ACID transactions, schema evolution, and time travel directly on data lakes makes production pipelines much more reliable. On top of that, optimizations like Adaptive Query Execution, auto-optimize, Z-ordering, and caching really help when working with large datasets. Another thing I like is how well it integrates with the Azure ecosystem whether it’s ADLS, ADF, Key Vault, or Unity Catalog for governance. It reduces a lot of setup overhead and makes deployments smoother across environments. Overall, it lets me focus more on solving data problems and performance tuning rather than worrying about infrastructure management. Review collected by and hosted on G2.com.

One thing I dislike about Azure Databricks is that cost management can get tricky if clusters and jobs aren’t monitored closely. Because it’s so easy to spin up clusters and run large workloads, costs can increase quickly especially with auto-scaling or multiple parallel jobs running. So it requires good governance and monitoring in place. Another area is debugging and troubleshooting. While notebooks are great for development, debugging production job failures especially intermittent Spark or infrastructure issues can sometimes take time. Logs are available, but tracing the exact root cause across cluster events, Spark UI, and job runs isn’t always straightforward. I’ve also noticed that handling CI/CD and deployments (like moving notebooks, workflows, configs across environments) isn’t as smooth out of the box compared to traditional code repos. It’s improving with Databricks Asset Bundles and Repos, but still needs careful setup. That said, most of these are manageable with best practices cost controls, monitoring, and proper DevOps processes. Review collected by and hosted on G2.com.

Lakebase and API gateways. We use Lakebase as our primary database, and it has very strong capabilities for AI workloads. It’s also easy and flexible to work with because it’s a Postgres database. I think the addition of autoscaling databases is a really good improvement; instead of having static Compute Units assigned to each database, they can now scale automatically. I also like that with autoscaling you can set both the minimum and maximum CU, which gives you more control while still keeping things flexible. Review collected by and hosted on G2.com.

Pricing is still not very clear, things are still measured in Compute units which is really hard to get down for pricing Review collected by and hosted on G2.com.