The knowledge layer

Every analysis makes the next question cheaper.

Reports don't pile up — they compound. Ask one question across everything you've ever analyzed and get a synthesized answer with citations back to the exact reports.

How it compounds ↓ The account library →

ask_libraryacross all your reports
Has churn improved since the pricing change?
Churn fell from 6.1% to 4.8% in the two quarters after the change [S1], and the drivers shifted from price to onboarding [S2].
[S1] Q3 churn cohorts · [S2] churn drivers deck
answers cite the reports they came fromyour library answers

What Is the Knowledge Layer?

The Knowledge Graph visualizes connections. The Knowledge Layer is the infrastructure
that makes those connections possible in the first place.

Traditional Analytics
  • Reports live in folders
  • Search by filename only
  • No connection between analyses
  • Insights get buried and forgotten
  • Teams work in silos
Transform
Knowledge Layer
  • Reports become searchable objects
  • Search by meaning and intent
  • Automatic cross-analysis linking
  • Insights surface when relevant
  • Cross-team intelligence

How it actually works

No magic objects — three real steps, each one shipping today.

1  Every delivered report is embedded — report- and card-level vectors, computed at delivery.
2  Search speaks your language — reports_list with a semantic query finds past findings by meaning, not keywords.
3  ask_library synthesizes — one question, answered across everything you've analyzed, with [S1]-style citations back to the exact reports.

From Analysis to Searchable Objects

Every analysis produces standardized objects that become part of your knowledge base.
Here's how it works.

Step 1

Analysis Completes

// Churn Prediction Results
churn_rate: 4.2%
top_driver: support_tickets
at_risk_users: 847
model_accuracy: 0.89
Step 2

Object Extraction

metric segment insight model

Each finding typed and tagged

Step 3

Embedded & Indexed

Semantic vectors created
Related objects linked
Now queryable forever

Standardized Object Types

Metric
KPIs, rates, scores
Trend
Patterns over time
Segment
Groups & cohorts
Insight
Findings & drivers

Every object becomes a point in a high-dimensional vector space. Related analyses cluster together automatically — regardless of when they ran or who created them.

See the technical deep dive: Knowledge Graphs & Embeddings →

Cross-Sectional Queries in Action

One question. Insights from everywhere.

Query

"What do we know about profitability?"

The Knowledge Layer doesn't search for the word "profitability." It finds semantically related concepts across all your analyses—margin trends, cost drivers, pricing impacts, efficiency metrics.

Finance 3 months ago
Gross Margin: 42.3%
From Q4 Financial Analysis
Sales 6 weeks ago
Average Deal Margin: 35.2%
From Deal Analysis Report
Operations 2 weeks ago
Cost per Unit: $12.40 (down 8%)
From Efficiency Analysis
Product 1 month ago
Premium tier 3x more profitable
From Product Mix Analysis

Why This Matters

The Knowledge Layer solves three fundamental problems
that plague every data-driven organization.

Institutional Memory Loss

Organizations repeatedly analyze the same questions because past insights are buried. The Knowledge Layer ensures nothing gets forgotten—every analysis becomes permanent, searchable memory.

Siloed Intelligence

Marketing knows about churn. Finance knows about revenue impact. Product knows about feature usage. But no one connects the dots. The Knowledge Layer automatically links related insights across teams.

Discovery Friction

Finding relevant past work requires knowing exactly what to search for. The Knowledge Layer enables semantic discovery—ask questions in natural language and find conceptually related analyses.

Every analysis makes your organization permanently smarter.

Start Building Your Knowledge Layer

Every analysis you run becomes searchable organizational intelligence. The more you use it, the more valuable it becomes.

WHERE NEXT
cite itCitations in every report →Four citation formats, ready to paste into your paper or deck. methodHow reports stay reproducible →Pinned code, versioned data, and re-runnable results. howHow the pipeline works →From your question to a checked, delivered report.