Your data has numbers. Your AI needs meaning.
Chartcastr's semantic layer encodes what metrics mean, how they relate to your business, and what matters, so AI analysis is interpretation, not narration.
Why most AI analytics fail
The difference between “MRR grew 3%” and “MRR grew 3%, beating the Q1 target in your planning doc, driven by the campaign you launched Tuesday” is not a better model. It's better context.
Raw Data
Just numbers. No context. A chart that says revenue went up or down, with no explanation of why or whether it matters.
Generic AI
Narrates what the chart already shows. “Revenue increased by 3.2%, indicating positive momentum.” Adds nothing. Reads like a school report.
Context-Aware AI
Explains what happened and why it matters. References your targets, your team's decisions, and related signals from other sources. Actual analysis.
Five layers of meaning
Each layer adds context the AI uses during analysis. The first two work automatically. The rest build over time as you use the product.
Metric Definitions
Canonical definitions per provider. What AOV means, what a strong conversion rate looks like, the difference between Net Sales and Total Sales in Shopify. The AI starts from shared vocabulary, not guesswork.
Domain Expertise
Per-provider analysis instructions, common misinterpretations, and chart-type guidance. The AI knows Shopify metrics differently than HubSpot metrics, because they mean different things.
Your Business Context
Link your Google Docs, Notion pages, Granola meeting notes, Confluence, or Sheets tabs. Your planning docs, targets, and definitions become part of every analysis, automatically.
Institutional Memory
Thread conversations from prior pulses, team decisions, rejected suggestions. The AI remembers what was discussed last Tuesday and builds on it instead of repeating itself.
Cross-Source Intelligence
During analysis, the AI can search your Notion workspace, read Slack canvases, check Google Drive, query PostHog, or look up Linear tickets. It finds the why, not just the what.
What context-aware analysis looks like
Same data. Same model. Different context. The output is unrecognizable.
Shopify revenue drop
Without semantic layer
“Revenue decreased by 12% compared to the previous week, which represents a notable decline from the recent trend.”
With semantic layer
“Revenue down 12% WoW, consistent with the budget reallocation discussed in the March 25 marketing sync. Meta spend cut landed Thursday. This is the expected lag.”
HubSpot pipeline
Without semantic layer
“Pipeline value increased this week with 14 new deals added across all stages.”
With semantic layer
“Pipeline above the Q1 target outlined in your planning doc ($2.1M vs $1.8M goal), but average deal size declining for the third consecutive week. The new outbound sequence is filling top-of-funnel with smaller accounts.”
Google Sheets MRR
Without semantic layer
“MRR grew by 3.1% this month, continuing a positive trend.”
With semantic layer
“MRR grew 3.1%, with 4 new enterprise plans from the Product Hunt launch. Sarah asked about this last week in the thread. The spike she flagged has held.”
No warehouse required
Traditional semantic layers sit between your warehouse and your BI tool. Ours sits between your data and the AI that explains it. No ETL. No SQL. No warehouse.
Connect your sources. Add context if you want to (the built-in layers work automatically). Get analysis that understands your business delivered on schedule.
Common questions
What is Chartcastr's semantic layer?
It's the knowledge system between your raw data and the AI that explains it. Five components work together so the AI interprets your charts instead of just narrating them: metric definitions, domain expertise, your business context docs, institutional memory from past conversations, and cross-source search tools.
How is this different from dbt's semantic layer?
dbt's semantic layer translates business terms into SQL queries. It ensures "revenue" means the same thing in every query. Chartcastr's semantic layer ensures the AI understands what "revenue up 3%" means in the context of your targets, your recent decisions, and your business state. Both are valuable. They solve different problems at different points in the stack.
Do I need a data warehouse?
No. Traditional semantic layers sit between a warehouse and a BI tool. Ours sits between your data sources (Shopify, Sheets, HubSpot, whatever you use) and the AI that analyses them. No ETL, no SQL, no warehouse required.
What context sources can I connect?
Google Docs, Google Sheets (specific tabs or ranges), Notion pages, Confluence, Granola meeting notes, Linear projects, and free-text context you write directly. More providers are added regularly.
Can the AI access my documents during analysis?
Yes. When you link a document to a source or source group, its content is included in the AI's context during every scheduled analysis. The AI can also actively search your connected workspaces (Notion, Google Drive, Slack canvases) to find relevant context it wasn't directly given.
What happens if I don't add any context?
The semantic layer still works. The first two layers (metric definitions and domain expertise) are built in and activate automatically based on your data source. You get better-than-generic analysis out of the box. Adding your own context documents and letting the memory layer build up over time makes it progressively better.
Related docs
Setup guides and deeper context for this feature.
Your metrics have meaning. Let the AI use it.
Connect your sources, link your context, and get analysis that actually understands your business. Free to start. The semantic layer builds itself as you use it.
No card required. Setup in 3 minutes.
