I get asked some version of "should we move off Metabase" or "should we move to Metabase" at least twice a month. The answer is almost always "depends on who is going to use it and what they already know". The follow-up is always "okay but what would you actually pick".
This post is the answer I usually give, written down so I can stop typing it into Slack DMs. It covers the seven tools that actually come up in those conversations: Metabase OSS, Metabase Pro/Cloud, Hex, Lightdash, Looker, Mode, and Preset.
I have used all of them in production for at least one client. None of them is the right answer for everyone. Most of the people I know who have switched between two of them ended up roughly equally happy on either side, and the switch cost them a quarter.
How to read this comparison
There are four questions that actually decide this:
- Who writes the queries? Engineers, data analysts, business users, or all three?
- What is your modeling layer? dbt, no formal layer, or a heavy semantic layer like LookML?
- How much do you care about embed and white-label? Selling dashboards to customers is a different game.
- What's your budget tolerance? "Free if I host it" and "$50k a year" are both fine answers, just very different worlds.
I'll come back to these four under each tool.
Quick verdict table
| Tool | Best for | Worst for | Starting price |
|---|---|---|---|
| Metabase OSS | Series A/B SaaS, mixed-skill team, self-host OK | Heavy modeling, embed at scale | Free (self-host) |
| Metabase Pro/Cloud | Same as OSS, want SSO and embed without ops | Same as OSS | $85/mo (Starter) |
| Hex | Analyst teams that live in notebooks, Python work | Non-technical users, simple dashboards | $24/user/mo (Team) |
| Lightdash | dbt-heavy teams, want a metrics layer in code | Teams without dbt, business users solo | Free OSS or $24/user/mo |
| Looker | Enterprise, semantic layer, data governance | Small teams, ad-hoc analysis | $5k+/mo (custom) |
| Mode | SQL-first analyst teams, narrative reports | Non-SQL users, dashboards-as-product | $349/mo (Team) |
| Preset (Superset) | Dashboard-heavy use cases, embed | Notebook work, lightweight teams | $20/user/mo |
Prices are 2026 list prices for the smallest paid tier, before negotiation. They will change. Confirm before you commit.
Metabase (OSS)
Open-source Java app (source on GitHub), self-host on a small VM, point it at your warehouse, give people accounts. The reason it shows up in every conversation is that it has the lowest "first useful dashboard" time of anything in this list, by a lot. A non-technical person can build a chart in 10 minutes by clicking through the visual query builder.
Where it shines. Mixed-skill teams. The CEO can build a count of users by source. The analyst can write SQL questions. Engineers can write SQL with template variables. Everybody ends up in the same Metabase. For Series A through early Series C SaaS this is almost always the right starting point.
Where it breaks. Two places. First, modeling: there is no real semantic layer. You can use the new "Models" feature, which is basically saved queries with column types, but it is not LookML. If your team wants metric definitions in code with version control, Metabase will frustrate you. Second, scale: once you have 500+ dashboards and 30+ active users, the lack of governance shows up. Permissions get tangled, dashboards multiply, and nothing prevents the "five MRR queries" problem.
Embed? Possible, both static (signed URLs) and interactive (with Pro). For a few embedded dashboards in your own app it works. For a serious data product, look elsewhere.
Verdict. Default starting point for SaaS until you have a specific reason to switch.
Metabase Pro / Cloud
The hosted version with SSO, advanced permissions, audit logs, and embed dashboards in interactive mode. Same product, fewer ops headaches, more money.
Where it shines. When you have outgrown self-hosted OSS but the underlying tool still fits. SSO, granular permissions, sandboxing for embed: all the things compliance teams ask for. Cloud removes the "we lost two days because the EC2 instance ran out of disk" problem that hits every self-hosted Metabase eventually.
Where it breaks. It is still Metabase. Modeling is still light. The semantic layer story is still weak compared to Looker or Lightdash. If you wanted dbt-native, this won't fix that.
Embed? Yes, this is where Pro starts to make sense. Interactive embed, sandboxing, parameter pass-through. Acceptable for a B2B SaaS that wants to ship a few dashboards to customers.
Verdict. Worth it once you have over ~10 active users and want SSO. Below that, OSS on a small VM is fine.
Hex
Notebook-first BI tool that combines SQL, Python, and a visual canvas. Think Jupyter but with shareable apps and warehouse connections built in.
Where it shines. Analyst teams where Python is a first-class citizen. The "notebook → app" workflow lets an analyst build a one-off analysis and ship it to a stakeholder as an interactive page in the same hour. Forecasting, ML scoring, anything where SQL alone is not enough: Hex is the cleanest tool for it.
Where it breaks. Non-technical users. The first time a CEO opens Hex they will not know what to do. There is no equivalent of Metabase's visual query builder. If your audience is "everyone in the company can self-serve", Hex is not it. Also: cost compounds. Per-user pricing on a 50-person team adds up fast, and cheaper viewer-only tiers do not always solve it cleanly.
Embed? Yes, Hex has gotten better at this, but it is not their main pitch. If embed is your top priority, look at Preset or Metabase Pro.
Verdict. Pick Hex if your data team is 5+ people who are happy in Python and you want them to ship work fast. Do not pick Hex as a self-serve dashboarding tool for a 100-person company.
Lightdash
dbt-native BI tool. The metrics live in your dbt project as YAML, and Lightdash exposes them as queryable dimensions and measures. Open-source core, paid hosted version.
Where it shines. Teams that already have a serious dbt project and want their BI layer to come from the same source of truth. When somebody says "MRR is defined here, in mart_finance.sql", and that definition shows up automatically in the BI tool, you eliminate a whole category of "five MRR queries" problems by construction.
Where it breaks. Two places. First, you need dbt. Lightdash without dbt is a thinner Metabase, and you should just use Metabase. Second, business users solo. Lightdash is friendlier than Looker but still mostly a tool for data people. A non-technical CEO will be lost.
Embed? Yes, in the paid version. Less mature than Metabase or Preset on this.
Verdict. If you have a strong dbt practice and you are choosing your first BI tool, Lightdash is one of the most defensible picks of 2026. If you don't have dbt, this is not the tool.
Looker
The big one. Now Google Looker, with a heavy semantic layer (LookML), enterprise pricing, and the most rigid governance story in this list.
Where it shines. Large companies where data governance matters more than speed. LookML forces you to define metrics in code, version them, review them. That sounds like overhead until you have 200 analysts and need to be sure they all calculate "revenue" the same way. It also has the best permissioning story for true multi-tenant data.
Where it breaks. Small teams. The setup cost is real, both in money and in months of LookML work before anyone gets a dashboard. The pricing starts well into five figures a year. For a 30-person Series B, Looker is overkill.
Embed? Yes, this is half of why Google bought it. Looker embed at the high end is the de-facto standard for enterprise embedded analytics.
Verdict. Pick Looker when your problem is governance, not speed. Most readers of this post are not at that scale yet. The ones who are know who they are.
Mode
SQL-first BI tool with strong narrative-report features. Originally aimed at analysts who write a lot of SQL and need to ship findings to stakeholders. Acquired by ThoughtSpot in 2023, still its own product as of 2026.
Where it shines. Analyst teams that produce reports more than dashboards. Mode's notebook-style "Reports" with mixed SQL, charts, and prose are excellent for one-off investigations and weekly updates. The Python integration is solid. The visual quality of the output is the highest of any tool in this list.
Where it breaks. Non-SQL users. Mode's visual query builder exists but is weaker than Metabase's. If half your team will never write SQL, it is harder to make Mode the primary tool. Also, dashboards-as-product is not Mode's strength. The reports are great, the live dashboards are fine but not exceptional.
Embed? Yes, but it has not been the main pitch in years. Possible, not the first thing I would choose for it.
Verdict. Mode is the tool for analyst teams who think of their work as "ship a report" more than "ship a dashboard". If that is not your team, you can probably skip it.
Preset (Apache Superset)
Hosted Apache Superset, the open-source dashboarding tool that came out of Airbnb. (Preset is the hosted commercial version.) Preset is run by Maxime Beauchemin (creator of both Superset and Airflow), so it is the "real" hosted version.
Where it shines. Heavy dashboarding workloads, especially when you need many dashboards for many tenants. Superset's flexibility is real, and the chart library is unusually broad. If you are building a customer-facing analytics product and you need to scale to thousands of embedded dashboards, Preset and Superset are a serious option.
Where it breaks. First-time setup. Superset is more complex than Metabase out of the box, and the visual query builder is rougher. The default UX is not as polished. For a small team trying to get a first dashboard live in a day, Metabase is faster.
Embed? Yes. Preset has invested heavily here.
Verdict. Pick Preset over Metabase Pro when embed at scale or chart variety is your top priority. Otherwise Metabase will be faster to live.
Decision framework
Strip away the marketing and the choice usually comes down to one or two questions.
- "We have dbt and we love it." → Lightdash, or Metabase with dbt-managed views.
- "We're at Series A/B and just need dashboards working this week." → Metabase OSS. Migrate to Cloud once SSO matters.
- "Our analysts live in Python notebooks." → Hex. Anything else will get fought.
- "We're building a data product for customers." → Preset (open-source friendly), Metabase Pro (faster to live), or Looker (enterprise).
- "We're a 500+ person company and metric drift is killing us." → Looker, or a real semantic layer (Cube, dbt Semantic Layer) feeding whatever BI tool you have.
- "We mostly write reports, not dashboards." → Mode.
The single biggest mistake I see in tool selection is picking based on what the data team likes, when the actual users are non-technical. The reverse mistake (picking the lowest-friction tool when the team needs serious modeling) is also common but usually less expensive to recover from.
What this post deliberately ignores
- Tableau, Power BI, Sigma. Different tier of product, different audience. Tableau and Power BI are usually inherited from a corporate environment, not chosen. Sigma is good but I do not see it picked over the tools above by SaaS teams under 200 people, in my client base. If you are choosing between Sigma and one of the above, the same framework above applies.
- Cube, dbt Semantic Layer, MetricFlow. These are semantic layers, not BI tools. Some of the tools above use them; Cube can sit under Metabase or Hex, for example. Worth a separate post.
- Notebook-only tools (Deepnote, Observable). Adjacent to Hex but more analyst-focused. If "share interactive analyses" is the job, evaluate them, but they are not really general BI.
Where MetaLens fits
MetaLens audits Metabase setups specifically. It is a Metabase tool, not a multi-BI tool. If you are reading this post because your Metabase has gotten messy and you are wondering whether to clean it up or migrate, the honest answer is almost always: clean it up first. Most of the reasons people give for migrating are reasons their current tool is full of cruft, not reasons the tool itself is broken. Migrating to a new BI tool with the same governance habits will produce the same mess in 18 months.
If you are on Metabase and 30% of the time you spend on data tooling is hunting for the right dashboard or reconciling MRR queries, that is the problem to fix first. MetaLens does that part. Then, with a clean baseline, you can decide if the tool itself is the problem.
If you are not on Metabase, this post is hopefully still useful, even if MetaLens is not the right next step for you.
