My Analytics & BI Quadrant
A very opinionated assessment
This week, Gartner released its latest BI and Analytics Quadrant and I’m always curious to check the chart and read the Pros and Cons.
However, I often find that these reports lack actionability and are clearly targeting the Enterprise space I’m not really dealing with. So, I thought I'd share my own version of the magic quadrant, focusing on tools that are more relevant for Start and Scale-ups. :)
DISCLAIMER: This is my personal opinion and experience. Yours can differ.
You can build pretty and ACTIONABLE reporting with ANY of these tools. Depending on your use case, budget, team’s know-how, the company’s infrastructure and the stakeholders’ tech literacy, some options make more sense than others. It will always be a case-by-case decision.
Tools I have worked with*
*Sometimes a few years ago, but I try to stay up to date with feature development
Tableau: The Visualization King
Tableau is amazing for visualizations and the option which has been around the longest. Back in the day, their weak area was data transformation and preparation. Nowadays, they have dedicated tools for this, but — as I mentioned in a previous article—you don't need this logic in your BI tool anyway. It should reside in your data warehouse or semantic layer.
Tableau is great for analysts who don't have access to the data warehouse or aren’t that comfortable with SQL. It’s amazing to explore any random dataset thrown at you! But when you want to build production dashboards, you will very quickly find yourself on Google, Stack Overflow, or ChatGPT, building custom fields and parameters to filter by start and end date.
Things like this shouldn’t be hard and on top of that it can also be pricey quite quickly once you want to roll it out across the company and add Creator and Explorer Seats.
To wrap up my experience:
You can visualize ANYTHING in Tableau, but you need to build EVERYTHING.
Looker: Former State of the Art
Looker has long been considered a state-of-the-art BI tool, thanks to its back-then-groundbreaking Semantic Layer based on LookML and the ability to define everything as code. But its high cost and the rise of dbt have somewhat diminished its appeal.
The necessity to define everything as code means that Analysts and Business Users often need to rely on your data engineering team, which slows down decision-making and makes the tool less agile for business needs. (Or it incentivizes Shadow BI in another Analytics Tool or Spreadsheets)
Nowadays, often a lot of transformations, definitions and documentation
(e.g. of fields) happen within dbt, there’s less of a need for the powerful LookML.
And since they provide their own Semantic Layer, you will find them integrated into the Semantic Layer of AtScale, but e.g. not with dbt or Cube that often.
It’s a more closed eco-system and the incentive to integrate towards dbt is not there right now as Google also has its own alternative.
So you can argue that its biggest strength has developed into one of its weaknesses nowadays.
Power BI: The Oldschool feeling
“Nobody gets fired for buying IBM” they said in the ‘80s, and I think the same applies to Microsoft’s BI solution nowadays. It’s cheap because it’s heavily subsidized (like MS Teams), integrates well with MS Excel, where any piece of data sooner or later ends up, and you can build amazingly flexible reports with it.
So what’s wrong with it?
In my opinion, the workflow feels outdated:
You build a modern version of an OLAP cube, collaboration is a pain, especially for Mac users who need a virtual machine to run it. You need to build dashboards on your local machine and push them to the cloud, but e.g. only one person (the Owner) can see the config and transformations of a Dataflow at a time. At least with the new PBI projects, some sense of version-control will enter.
Then, I have the feeling that PBI wants to cater to everybody.
You can do a transformation and filter
in DAX
or in the PowerQuery of a Semantic Model
or in the PowerQuery of a Dataflow
or in a Datamart (soon “Fabric Warehouse”)
or in Factory
or in …
(You get the idea)
Eventually, you need a dedicated lineage tool for Power BI itself, which is available from third parties, to find where “Business Unit” is mapped.
Similar to Tableau, Power BI engine is made to import the data to deliver the best dashboard performance. This is because it obviously reduces the query load on the Data Warehouse and back in the days when these solutions were built Data Warehouses were mostly run on-premise and not in a scalable cloud.
I understand that for very large implementations compute cost of queries can be a concern, but I have also seen enough examples where the internal caching solved that pretty well.
Additionally, having the SQL-code generated from Charts and Dashboards available makes debugging so much easier.
Then, because of data import AND local development, you can end up with these 10GB .pbix files, you need to deal with (good luck having two of them open to compare/debug).
To tame performance you see yourself applying fancy hacks (here or here) or deal with things, I thought wouldn’t be necessary anymore in 2025.
While you can build VERY good dashboards with Power BI—there's no question about that—you can literally do the same with any other tool out there and have a better experience. Unless you're fully committed Azure, I see hardly any compelling reason to choose Power BI. Sorry.
Looker Studio: Free but Limited
Looker Studio is great at visualization and is free, making it very powerful for its price point (it’s free!). It's pretty intuitive, yet powerful, and anyone who dedicates half an hour to it can build a dashboard. I have found its time intelligence is unmatched (in terms of what you can achieve straight out of the box), especially comparing it to the solutions you’d need above. However, as your company grows, you'll need more governance and administrative features, which Google will happily upsell you into Looker Studio Pro or eventually Looker.
Buggy visualizations, slow browsers and little things like not being able to separate the database columns and the “labels” shown in the dashboards are things that can drive you crazy with a certain complexity.
But still, if you’re running on BigQuery, you won’t do anything wrong using Looker Studio as your first Dashboarding tool. If you’re on Snowflake or something else, you should think twice.
Metabase: Easy but Chaotic
Metabase is easy to get up and running (either Cloud or self-hosted Open-Source), but you are a little bit limited in terms of what you can visualize and how you can customize it. I don’t find that bad, because you should stick to boring tables, line- and bar-charts anyway. But especially the restricted visual tweaking might turn some people off.
Its modern data stack readiness might be a little bit better than Looker Studio, but just thanks to definitions as code with a very nice API, which enables third-party Python libraries or packages or some advanced deployments.
Their Question-based Chart-Builder approach makes a ton of sense first, but also leads you to re-invent the wheel (or “model”) very quickly without even noticing.
So be aware of chart and definition sprawl when using Metabase.
Also, at least when I worked with it in Q2 2024, it was not possible to build a Chart or Report with your development table from dbt and then later move it to production easily. (This might have changed in the meantime)
Generally, it's already a good starting point, but it could be better if you want to leverage the benefits of something like dbt or SQLmesh.
Superset: Techy but Clunky
Apache Superset (Open-Source) has a lot of great features to offer, but not very user-friendly.
It's a little bit clunky to use, and I wouldn't feel comfortable giving it to a business user to build a dashboard with it.
As often, it’s in the little things:
Performing more advanced time intelligence, such as comparing consecutive weeks or setting a custom start day for the week (e.g., Monday), can be surprisingly complex or impossible the last time I tried (2023 to be fair).
The same is true for sorting in pivot tables (e.g. nested product category and products in a dimension) and conditional formatting, which can be both time-consuming and frustrating to implement.
Their paid offering, Preset, doesn’t really solve that, but at least has a good dbt integration, which was one of the first to offer a bilateral information exchange!
However, if you are technical and want to tinker a little bit, you can get it done in Superset. If you’re goal is to let Analysts from the business teams serve themselves, I would not feel too comfortable.
Lightdash: My Current Favorite
Lightdash is my current favorite for startups and scale-ups, that have an Analytics Engineer who can leverage it. The UI is Looker-like, but the “backend” for definitions is more or less outsourced to dbt instead of writing LookML.
You define dimensions and metrics in your schema YAMLs next to your dbt tests, and therefore benefit from things like version control and batch editing.
You can self-host their open-source version or pay for the managed service if you want features like an AI Assistant or simply cannot or don’t want to deal with the infrastructure.
In terms of features, there are so many things that make so much sense:
You can automatically export data to Google Sheets - using definitions from your semantic model - while having certain governance, and see what is pushed where. Since it’s just a matter of time until stakeholders will ask you "can I have this Dashboard as a CSV or spreadsheet?", this is great.
Same goes for the Metric Canvas, dbt Write-Back or the Python SDK, which are still a bit raw, but show you where the direction goes.
Also, having different environments (e.g. prod, dev or PR1, PR2, PR3) —which is a pain in Power BI, Tableau, and Metabase. Being able to prototype and iterate based on the dev-version and then natively “promote” it to Production is great.
Would I recommend it for a 100,000-person enterprise? Not, right now.
But I think for startups and scale-ups, it's one of the tools to have a look at, if you use dbt.
Tools I haven’t touched yet*
*But look very intriguing for their very different reasons
ThoughtSpot: Search-Based and AI-Heavy
ThoughtSpot takes a unique approach with its search-based and AI-driven interface. Although I haven't had hands-on experience with it, I heard first-hand from cases where companies tested Looker vs Thoughtspot for a few weeks and decided on the latter. The reasons for that were the better dbt integration (at least dbt → Thoughtspot, but not the other way around) and easier workflow to explore and analyze data for Business Users.
Inserting some keywords and getting charts as a return looks magical in “canned” demos and video clips. The skeptic in me will only shut down once he sees it work out in real-life scenarios. But others seem to have it implemented well for them. Also, the options to customize charts are reportingly limited, but as said with Metabase, depending on your use case it might not be a deal-breaker.
So, without personal experience, I can't provide a detailed assessment, but it's certainly a tool to watch for that unique approach.
Sigma: Spreadsheet-Based and User-Friendly
Another unique approach is from Sigma.
They stand out with their spreadsheet-like interface, which is intuitive for business users familiar with Excel. This makes it accessible and user-friendly, and more flexible than traditional dashboards. Also, they seem to address something which is often a limitation of traditional dashboards: It’s read-only.
For more interactive applications, non-dashboard solutions such as Streamlit or Retool often served as an alternative to work with user input. Sigma allows you write back data (details).
Community sentiment seems to be on the positive side (1, 2), but I’d expect some immaturities in the UI / Dashboarding side.
Besides that, same as for ThoughtSpot, I’ll keep an eye on it.
Omni: Looker 2.0
Last but not least: Omni, founded by former Looker team members, aims to build on Looker's strengths while addressing some of its weaknesses.
One of the key improvements is reducing the dependency on engineering by allowing analysts and business users to define metrics and custom fields using Excel syntax (so you can join tables with a VLOOKUP). I personally like the idea of Sigma and Omni to drag Business Users not too much out of their familiar territory and incorporating familiar Syntax and Workflows into their tools.
(while the “owner” of Excel then came up with DAX or the M language :D)
This flexibility, combined with tight integration with dbt and features like Git branches for testing, makes Omni a promising tool, which seems to do a lot of things right so far.
Conclusion
In summary, tools like Sigma, ThoughtSpot, Omni, and Lightdash are leading the way from my perspective.
Things which are important to me:
dbt integration (to e.g. push descriptions or pull exposures)
The option to switch between the different dbt targets to let Analysts build Dashboards in dev instead of prod (which is unfortunately still the norm)
Definitions as code
The ability to see the SQL-statements from the charts to debug metrics more easily
They don’t lock you into their ecosystem with their own formula languages
You can bring your own AI (like Snowflake Cortex) and trigger it from the dashboards, which opens up a lot of interesting use cases
Since you can build pretty charts in all of the tools, the differentiators are rather in the workflow and philosophy for me. And this might be very different based on your preferences. If you want to build your transformations with SSIS, Azure Data Factory or Fabric: go ahead! Then, Power BI is probably an amazing fit.
To add some context to my Quadrant:
For the two axes I chose the more tangible “User friendliness” and “MDS Readiness” over the original “Ability to Execute” and “Vision”.
User friendliness covers the Dashboard Experience and Deployment, and Governance aspects.
MDS Readiness covers more technical things like dbt integrations and definitions as code.
It’s rather “ball-parking”. So, please do not nail me down if a tool should be a bit more left/right or down/up.
Also, just because a tool is not in the top right of the quadrant does not mean that I wouldn’t recommend it in certain situations. There are a lot of situations where, even right now, I would suggest a Metabase or Looker Studio without hesitation. Just that I prefer the direction of others more.
Before you switch or implement any BI tool you need to answer questions like:
Where should transformations and definitions live?
Who builds dashboards and how many people will do it?
Who will be the user of the tool? Tech-savvy? Non-technical?
Self-host or Cloud?
What’s your budget in terms of price and time?
…
I'd love to hear your thoughts and experiences with these tools.
Do you see anything similar or different to how I see it? I would be very curious to hear what you like or dislike about certain solutions.