This note covers the BI tool landscape in 2026, with a focus on four selection criteria and a comparison of Lightdash, Looker, and Metabase. Roughly 15 serious contenders exist beyond legacy incumbents; the four questions below are the primary filters.
The Four Questions
1. How Invested Are You in dbt?
This is the most decisive filter. If dbt is your transformation layer and you want your BI tool to read from it directly, the landscape narrows significantly.
Deep dbt investment: Lightdash is the strongest fit — it reads directly from dbt project YAML, auto-generating dimensions and metrics from model metadata. No separate modeling layer, no drift between transformation and presentation. Evidence works if you prefer code-based reports (SQL + Markdown deployed as static sites). Omni is worth evaluating if you want Looker-like governance with better dbt integration.
No dbt: Metabase or Preset/Apache Superset are better starting points. They don’t assume a specific transformation layer.
dbt but not all-in: Looker has a dbt Semantic Layer connector. Tableau shipped one in Tableau Cloud 2025.2. These tools maintain their own modeling layers but can consume dbt metrics.
2. What Are Your Governance Requirements?
Governance separates tools that work at 5-person teams from tools that work at 500-person organizations.
Enterprise governance: Looker still has the most comprehensive permission model (Permission Sets + Model Sets + Content Access), detailed audit logging, and mature row-level security. If your compliance requirements demand granular access control and audit trails, Looker’s decade of enterprise refinement is hard to match.
Engineering-workflow governance: Lightdash offers Git-native governance aligned with engineering workflows — metric definitions reviewed in pull requests, changes tracked in version control. User attributes handle row-level security. This appeals to teams where the metrics-as-code pattern is the governance mechanism.
Tiered governance: Metabase gates most governance features (audit logging, sandboxing, advanced permissions) behind Pro/Enterprise tiers. Free and Starter tiers are permissive by design.
3. What’s Your Budget?
This is where differences become stark. Pricing models vary so much that the same team might pay 4x more with one tool than another.
| Tool | Pricing Model | 50-Person Team Estimate |
|---|---|---|
| Lightdash | Unlimited users, flat rate | ~$28,800/yr (Cloud Pro $2,400/mo) |
| Metabase | Per-user or self-hosted free | ~$0 (self-hosted) to ~$10,200/yr (cloud) |
| Looker | Enterprise negotiated | ~$84K-120K/yr |
| Preset | Per-user | ~$12,000/yr ($20/user/mo) |
| Sigma | Per-user, tiered | ~$36K-60K/yr |
| Evidence | Open source + cloud tiers | $0 (self-hosted) to custom pricing |
Lightdash’s unlimited-user pricing at $2,400/month makes the economics dramatically different from per-seat tools at scale. But Looker’s $150K average deal (Vendr data across 355 transactions) includes a governance layer that cheaper tools don’t match out of the box. Gartner estimates organizations spend 40-60% of total Looker investment on LookML development and maintenance — a real cost that doesn’t appear on the license invoice.
4. Who Are Your Users?
User sophistication determines adoption more than feature sets.
Non-technical business users who need to self-serve without SQL: Metabase’s visual query builder or Sigma’s spreadsheet-like interface will get adoption faster than tools requiring understanding of dbt concepts or YAML. Sigma’s interface generates SQL against the warehouse in real-time while looking like Excel — popular with finance teams for exactly this reason.
Analytics engineers and data analysts comfortable with YAML and SQL: Lightdash or Evidence will feel natural. These tools assume a technical user base and optimize for that workflow.
Mixed audiences: Omni offers multiple exploration modes for different skill levels. Hex blends SQL, Python, and no-code in a notebook-style interface. These tools try to serve both technical and non-technical users, though the tradeoff is typically more complexity in the overall system.
The Player Map
Tools grouped by design philosophy:
dbt-Native Tools
Lightdash (MIT license, ~5,600 GitHub stars). The BI tool most tightly coupled to dbt. Auto-generates dimensions and metrics from dbt YAML. Cloud Pro at $2,400/month with unlimited users. In 2025: AI Agents for natural language querying, React SDK for embedding, Vega-Lite custom charts, write-back promoting ad-hoc SQL into governed dbt models.
Evidence (MIT, ~4,000-5,000 GitHub stars). BI as code: SQL + Markdown reports deployed as static sites. Version-controlled alongside your dbt project. Evidence Studio cloud launched June 2025, embedded analytics following in December.
Omni (ex-Looker engineers, $69M Series B at $650M valuation, March 2025). YAML-like modeling with Git versioning, multiple exploration modes. Acquired Explo (September 2025) for embedded analytics.
Accessibility-First Tools
Metabase (AGPL, 40,000+ stars, 60,000+ companies). Visual query builder for non-technical users. Free self-hosted, $85/month cloud for 5 users. The tradeoff: no central semantic layer. Metrics defined per-question, creating consistency problems at scale.
Preset / Apache Superset (Apache 2.0, 70,000+ stars — most-starred open-source BI project). 40+ visualization types, SQL Lab IDE, plugin architecture. Preset managed SaaS at $20/user/month. AI Assist for natural language to SQL shipped in 2025.
Enterprise and AI-Forward Tools
Looker (Google Cloud). LookML semantic modeling refined over a decade. Gemini-powered conversational analytics (2025). Average annual deal ~$150K (Vendr).
Sigma Computing ($100M ARR, $1.5B valuation). Spreadsheet-like interface generating live SQL. Snowflake BI Partner of the Year three consecutive years. Databricks BI Partner of the Year 2025.
Hex ($70M Series C, $172M total funding). Notebook-style: SQL + Python + no-code. Acquired Hashboard (May 2025) for semantic modeling. Customers include Reddit, Figma, Anthropic.
ThoughtSpot (acquired Mode for $200M, July 2023). AI-powered search + Mode’s code-first analytics = Analyst Studio. Combined ARR exceeds $150M.
Worth Watching
Steep (Stockholm). Integrates with dbt Cloud semantic layer. Notion/Figma-like UX. Free for up to 3 users.
Streamlit (40,000+ stars, acquired by Snowflake for ~$800M). Python-native data apps. More app framework than BI tool, but fills a similar need for technical teams.
The Head-to-Head: Lightdash vs Looker vs Metabase
These three represent the most common real-world comparison for teams evaluating modern BI.
| Dimension | Lightdash | Looker | Metabase |
|---|---|---|---|
| Semantic layer | dbt YAML native | LookML (proprietary) | None (per-question) |
| Self-hosting | MIT, free | Paid license only | AGPL, free |
| dbt integration | Built on dbt | Parallel layer + connector | Third-party plugin only |
| Entry cost | Free / $2,400/mo | ~$36K-60K/yr | Free / $85/mo |
| Non-technical access | Moderate | Moderate (after LookML setup) | Excellent |
| API | Cloud Pro+ | Excellent (REST 4.0) | Solid |
| Embedding | React SDK, $0.05/load | Embed Edition (expensive) | Mature SDK |
| Row-level security | User attributes | Native, mature | Pro/Enterprise only |
| AI features | AI Agents (NLQ) | Gemini conversational | MetaBot |
Choose Lightdash when dbt is your foundation, your users are comfortable with analytics engineering workflows, and you want unlimited-user economics.
Choose Looker when enterprise governance is non-negotiable, you have budget for LookML development, and your organization needs the most comprehensive permission model available.
Choose Metabase when non-technical self-service is the priority, budget is constrained, or you need to get a BI tool running this week rather than this quarter.
Market Context
The global BI market sits at roughly $41B in 2026, projected to reach $62B by 2031 (Mordor Intelligence). Market share leaders remain Power BI (~20%), Tableau (~16.4%), and Qlik (~10%). The modern tools covered here collectively represent a smaller share, but they’re where growth and innovation are concentrated.
The Fivetran-dbt Labs merger (late 2025, combined ~$600M ARR, 10,000+ customers) is the defining event. It signals that the “modern data stack” is consolidating around fewer, larger platforms. The transformation layer — dbt — is becoming the foundation that BI reads from, not a separate concern.
Hybrid Deployment Pattern
A common migration path: retain the legacy tool for financial reporting and board decks, add a modern tool for ad-hoc analysis. This avoids a full migration while getting most of the benefit of a newer tool.
73% of BI implementations fail to deliver ROI in the first year (SR Analytics). Poor data modeling, missing governance, and unclear metric definitions are the common failure modes — not tool selection itself.