Teams define business logic inside BI tools, creating vendor lock-in and making it hard to switch tools or maintain consistency across dashboards
A version-controlled semantic layer that sits between the data warehouse and any BI tool, paired with an AI engine that auto-generates dashboard code from natural language or metric definitions
freemium - free for small teams/limited metrics, paid tiers for enterprise features like governance, audit trails, and multi-BI-tool sync
BI vendor lock-in is a well-documented, deeply felt pain in data teams. Enterprises spend months migrating between BI tools because business logic is trapped inside them. The Reddit signal ('BI vendors desperately want you to define business logic in them to stay relevant') reflects widespread frustration. However, it's a 'slow burn' pain — teams live with it until a migration forces the issue.
TAM is substantial — the BI market is ~$30B+ globally, and semantic layer / metrics layer is a growing sub-segment. Target of mid-size to enterprise data teams using dbt narrows the addressable market but increases willingness to pay. Estimated SAM: $500M-1B for semantic layer tooling. However, the specific intersection (AI + multi-BI) may initially serve a niche within that.
Data teams already pay $50-150/user/mo for BI tools, and enterprises pay $50K-200K/year for tools like AtScale and Looker. A tool that reduces BI migration costs or eliminates vendor lock-in has clear ROI. But the buyer (data engineering lead) needs to justify budget, and 'insurance against lock-in' is harder to sell than 'solve today's urgent problem.' Freemium for small teams is smart — land with data engineers, expand to enterprise.
This is the critical weakness. Generating native Tableau (.twb XML), Power BI (.pbix — proprietary compressed format), and Looker (LookML) files from a single semantic definition is extremely hard. Each BI tool has underdocumented, frequently-changing file formats. Keeping dashboards in sync across tools is a complex state management problem. AI generating 'good' dashboards (correct chart types, layouts, filters) requires significant UX intelligence. A solo dev cannot build a credible multi-BI-output MVP in 4-8 weeks. A single-BI-output MVP (e.g., just Metabase or Superset via API) is feasible.
No existing tool combines all three pillars: semantic layer + AI dashboard generation + multi-BI output. Cube comes closest on the semantic layer, Hashboard on dashboards-as-code, but nobody generates native dashboards across multiple BI tools from a single definition. This is a genuine gap, not just a feature delta.
Natural subscription product. Metrics evolve, dashboards need updating, new BI tools get adopted, team members change. The sync/governance layer is inherently ongoing. Enterprise features (audit trails, RBAC, multi-BI sync) command premium recurring pricing. Usage-based pricing on metrics or dashboard syncs is viable alongside seat-based pricing.
- +Genuine whitespace — no tool combines semantic layer + AI generation + multi-BI output today
- +Strong pain signal backed by the entire 'modern data stack' fragmentation problem
- +Natural enterprise upsell path (governance, audit trails, multi-tool sync)
- +Positioned at the intersection of two hot trends (semantic layers and AI-powered analytics)
- +dbt ecosystem provides a clear distribution channel and community to build within
- !Technical complexity of generating native dashboard formats for multiple BI tools is severely underestimated — each tool's format is a multi-month engineering effort
- !Cube.dev or dbt could add AI dashboard generation as a feature, instantly commoditizing the differentiation
- !Enterprise sales cycles are 6-12 months — time to revenue is long and capital-intensive
- !BI tool format specs are underdocumented and change with updates, creating ongoing maintenance burden
- !The 'multi-BI output' value prop assumes companies want to maintain multiple BI tools rather than just picking one — some will view this as enabling a problem rather than solving it
Open-source semantic layer platform that exposes metrics via REST/GraphQL/SQL API to any downstream BI tool. Includes caching and pre-aggregation engine.
Metrics defined in YAML within dbt projects, exposed via GraphQL/JDBC API through dbt Cloud. Acquired Transform/MetricFlow in 2023.
Dashboards-as-code platform where metrics and dashboards are defined in YAML config files, version-controlled, and deployed via CI/CD. Combines semantic layer with visualization.
Enterprise BI platform with LookML, a Git-based semantic modeling language. Most mature semantic layer in market with 10+ years of development.
Code-first BI tool where you write SQL + Markdown and get auto-generated dashboards deployed as static web apps. BI as a static site.
Scope down aggressively. MVP = semantic layer (YAML, dbt-compatible) + AI generation of dashboards for ONE open-source BI tool (Metabase or Apache Superset — both have well-documented APIs). Skip Tableau/Power BI native format generation entirely for v1. Use LLM to generate dashboard JSON via the target tool's API from natural language prompts. Ship as a CLI tool that reads your dbt/YAML metrics and pushes dashboards to Metabase. Add a second BI tool output only after the first one works well and customers demand it.
Free CLI for single-BI-tool output with ≤10 metrics → $49/mo Pro for unlimited metrics + AI generation + dashboard versioning → $199/mo Team for collaboration + audit trails → $499+/mo Enterprise for multi-BI-tool sync + governance + SSO + SLA. Revenue accelerator: managed cloud service that handles sync/scheduling.
3-4 months to MVP (single BI tool output), 5-6 months to first paying customer (likely a design partner from the dbt community). 12-18 months to meaningful recurring revenue ($10K+ MRR). Enterprise deals start closing at 18-24 months. The conditional GO depends on scoping the MVP to a single BI output — trying to launch with multi-BI output will push first revenue to 12+ months.
- “have your BI tool hit a version controlled semantic layer for deterministic SQL generation so that you aren't locked into any particular BI tool”
- “BI vendors desperately want you define your business logic in them to stay relevant”
- “explore dashboards-as-code tools and have AI generate them”