What the interviews show. Both are rated highly by their own buyers and very few describe replacing one with the other, so this is not a page about a winner. What the two buyer groups talk about is different. Databricks commentary concentrates on flexibility, notebooks and data-science work — several describe it as their core ETL and decision-making platform — with the recurring complaint that things have to be built rather than picked up. Snowflake commentary concentrates on reach: a warehouse buyers describe getting into through a web interface without additional tooling, and the environment the BI tools sit on. Its recurring complaint is compute spend. Separately, the buyers who describe moving describe shifting individual workloads rather than migrating off. Integration and pipelines draw disproportionately negative commentary on both sides, measured against each platform's own mix of praise and complaints — a shared complaint rather than a clean distinction between them.
| Strongest contrast in these interviews | Commentary emphasizes building — notebooks, pipelines and models the team assembles | Commentary emphasizes reaching — a warehouse buyers describe getting into without additional tooling |
|---|---|---|
| Buyer pattern in these interviews | Data-engineering-led teams, with the platform operated by specialists | Mixed teams, with analysts working alongside the engineers |
| Most-discussed job | Data engineering, ETL and machine learning | Cloud data warehouse |
| Also does | Dashboards and reporting buyers describe as workable rather than its strength | Sits underneath the BI layer — buyers name Power BI and Tableau on top of it |
| Praised most | Flexibility and the notebook model, for teams doing data-science work | Ease of reach — a web interface buyers describe using without additional tooling |
| Top complaint | Everything has to be built; buyers call it clunky next to a BI tool | Extracting data described as needing developer time and familiarity with the structure |
| Relative cost | Buyers describe cost getting out of hand as usage grows, and persistent upsell | Compute spend is the complaint that recurs; one buyer names it as the reason for moving a workload off |
| Where buyers say it fits | A team with the engineering capacity to build what it needs | A stack where reaching the data without additional tooling matters |
What buyers say about each platform
Every buyer interview spends its minutes somewhere — praising the thing that won the deal, or flagging the thing that still stings. Map where those minutes go for Databricks and Snowflake, and the two reputations are built out of different material:
Relative share of buyer commentary by theme in Alium's verified interviews, praise vs. complaint, through September 2026. Widths compare themes within this pair — they are relative, not counts. Valence is read against each platform's own mix of praise and complaints.
In their own words
The phrases buyers reach for, verbatim, when they describe each platform:
Databricks
“core ETL decision making platform” what it is used for
“very flexible and it's very user-friendly” data connections
“everything has to be built” the recurring complaint
“feels very clunky” next to a BI tool
“cost can get out of hand” as usage grows
Snowflake
“it's clear, it's clean, it's easy to use” reaching the data
“preferred by a lot of data and engineering teams” against the alternatives
“quite cost intensive” as a data lake
“a little limiting” support
“institutionalized” what makes leaving hard
Where Databricks buyers concentrate their praise
Databricks' standing rests on what teams build with it. Its buyers describe it as a core ETL and decision-making platform, heavily used for data preparation and machine learning, and they name the notebook model as the thing they reach for. Several describe strong connections to their data sources and a performant structure underneath, and one describes building dashboards directly against their data faster than they could elsewhere. The cautions are consistent and they are about the assembly: buyers describe everything having to be built, data cleaning that is not its strong suit, and integration complexity with non-native sources. One rating it low calls it clunky next to a BI tool, and a separate complaint runs alongside — that cost can get out of hand as usage grows, with one buyer describing persistent selling of additional services.
Where Snowflake buyers concentrate their praise
Snowflake's standing rests on ease of reaching the data. Buyers describe it as clear, clean and easy to use, reached through a web interface without extra tools on anyone's machine, and they name it as the environment their BI layer sits on. They praise how it organizes data into standardized formats and describe its compute favourably — one says they still rate its compute above the alternative even while moving workloads for budget reasons. The cautions are real and mostly commercial. Cost is the complaint that recurs, described as cost-intensive at volume and enough to cause delays; one buyer describes past struggles extracting data that needed developer time and familiarity with how it was structured; another calls its support a little limiting. One buyer who looked at consolidating elsewhere describes the obstacle as how much had become institutionalized on it rather than anything technical.
What buyers wish they'd known
Before picking Databricks
Budget the building, not just the platform. Buyers describe everything having to be assembled, data cleaning as a weak point, and integration complexity with non-native sources — and the ones rating it lowest are comparing its reporting with a dedicated BI tool. Cost is described as something that grows with usage rather than a fixed line.
Before picking Snowflake
Price the compute against the workloads you actually run. Buyers describe it as cost-intensive at volume, enough to cause delays and to move work elsewhere. And note what one buyer says made leaving hard: not the technology, but how much had become institutionalized on it once the reports and jobs were built.
Four questions the interviews raise
Its buyers describe specialists working in notebooks, and the ones rating it low describe it as not user-friendly for their team.
Buyers describe reaching it through a web interface with nothing extra installed, and name ease of use repeatedly.
Flexibility, notebooks and machine learning are the largest share of its commentary by a clear distance.
Buyers describe it as the warehouse the rest of the stack reads from, with the BI tools sitting on top.
Buyers describe cost growing with usage, and one describes persistent selling of additional services.
Compute spend is the complaint that recurs; one buyer names it as why they moved a workload elsewhere.
Buyers describing moves onto Databricks describe workloads shifting individually rather than a wholesale replacement.
One buyer describes the obstacle to consolidating elsewhere as how much had become institutionalized on it.
Don't see your exact situation?
Get the read for your brand →How the operating conditions differ
Recurring situations in the corpus, and where the evidence concentrates in each:
Pressures visible in these interviews
The complaints these interviews carry sit on both sides: Databricks' are that everything has to be built, data cleaning described as a weak point, reporting buyers compare unfavourably with a dedicated BI tool, and cost that grows with usage; Snowflake's are compute spend at volume, extracting data described as needing developer time, and support one buyer calls a little limiting. Integration and pipelines behave the same way on both — measured against each platform's own mix of praise and complaints, commentary there skews negative on each side, so it is a shared source of negative commentary across the two buyer groups rather than a distinction. The implication — ours, not a buyer's — is that the choice these interviews describe is less about which platform is better than about the work being done and how people need to reach the data, while the cost comparison depends on the workloads actually being run. For the wider market, see our Tableau alternatives read and Looker vs. Power BI.
Common questions
Databricks vs. Snowflake: which is better?
Both are rated highly by their own buyers in Alium's verified interviews, and very few describe replacing one with the other, so these interviews do not pick a winner. What they separate is what each buyer group talks about. Databricks commentary concentrates on flexibility, notebooks and data-science work, with the recurring complaint that things have to be built rather than picked up. Snowflake commentary concentrates on reach — buyers describe a warehouse reached through a web interface without additional tooling — with compute spend as the recurring complaint. Integration and pipelines draw disproportionately negative commentary for both platforms, measured against each platform's overall mix of praise and complaints, making integration a shared complaint rather than a clean distinction between them.
Are buyers moving from Snowflake to Databricks?
Some describe moving workloads rather than migrating off. One buyer says they still rate Snowflake's compute higher but shift workloads to Databricks to reduce the cost of compute; another describes consolidating onto Databricks as difficult because so much is already institutionalized on Snowflake. In the moves captured here, buyers describe shifting individual workloads rather than replacing one platform outright.
What is the difference between Databricks and Snowflake?
Buyers describe them doing overlapping jobs from different starting points. Databricks is described around data engineering, ETL and machine learning, with the notebook model as what its buyers reach for, and several describe using it as a core ETL and decision-making platform. Snowflake is described as the cloud data warehouse the rest of the stack sits on, reached through a web interface without extra tooling, and named as the environment underneath BI tools such as Power BI and Tableau. The practical difference in these interviews is how people get to the data.
Is Databricks or Snowflake more expensive?
These interviews do not order them on price, and buyers on both raise cost. Snowflake's version is compute spend: buyers describe it as cost-intensive at volume, and one describes moving workloads elsewhere specifically to reduce what compute costs. Databricks' version is that cost can get out of hand as usage grows, and one buyer separately describes persistent selling of additional services. The implication — ours, not a buyer's — is to compare the cost of the specific workloads you would run on each rather than treating either platform as categorically cheaper.
See which one fits your exact stack
Weighing Databricks against Snowflake, or running both and wondering which workloads belong where? Do a 15-minute interview about your stack and get this personalized — what peers your size say each one costs to run, how their teams actually get to the data, and where the pipelines bite.
Get my personalized read — 15-min interviewNo password · your interview is anonymized before it ever informs a page like this one.