Growth Intelligence System / Quattr
Turning expertise into
repeatable growth programs.
I designed the analytical skills and built the system around them so a customer’s growth question could become a repeatable method, an inspectable recommendation and a useful deliverable.
- My contribution
- Skill design · Analytical workflows · System design
- Working context
- Enterprise search and customer delivery
- Evidence
- Implementation excerpts and archived reports
01 / The recurring problem
The question repeats.
The account context changes.
Which pages deserve attention? Where do paid and organic signals overlap? What should we investigate when performance changes?
Those questions recur across customers. But the business priorities, data availability, conversion definitions and reporting needs differ. Each answer can become another custom project unless the method has a consistent structure.
The opportunity was to preserve customer context while making the analytical work reusable.
02 / The design decision
Keep the method reusable.
Make the context explicit.
I designed the analytical skills and built the Growth Intelligence System around them. Each skill needed a clear purpose, declared data requirements and a structured result.
Account differences belonged in a registry and client ontologies: configuration describing the customer, the relevant datasets and the definitions the analysis needed. Shared runbooks explained how agents should use that context.
@dataclass
class SkillResult:
success: bool
data: Any = None
explanation: Optional[str] = None
errors: List[str] = field(default_factory=list)03 / How it works
A small set of responsibilities,
connected deliberately.
Define the question, consult account context and keep the analysis within its supported scope.
Name the relevant datasets and preserve client-specific definitions and availability.
Resolve configured data references rather than guessing table names.
Give agents a consistent interface to analytical capabilities and their results.
data_catalog:
gsc:
bulk_export:
database: PROD
status: available
alias: searchdata_url_impression
primary_table:
GOOGLE_SEARCH_CONSOLE_BQ_<CLIENT>.SEARCHDATA_URL_IMPRESSIONExplore the implementation and its limits
The implementation combines Python analytical skills, a registry and YAML ontologies, Snowflake data access, and an internal MCP interface. DatasetGuard and ClientResolver provide dataset-resolution paths.
Skill contracts include required and optional datasets, execution metadata and structured results. Infrastructure verification exists as an explicit path; legacy and cached paths have documented limits. Source code establishes implementation, rather than universal enforcement or deployment-wide adoption.
04 / The useful output
Give the customer
a place to start.
A related prioritization workflow brought organic demand, business-event signals and paid investment together around a common page identity. It applied configurable scoring and retained the reasons for review.
One archived export contained 812 pages and a top-100 shortlist. The result gave a team a defined set of priorities to inspect and discuss.
| Page | Clicks | Impressions | Queries | Priority score |
|---|---|---|---|---|
| Page 01 | 822 | 338,594 | 5,805 | 90.75 |
| Page 02 | 88 | 101,932 | 2,424 | 25.09 |
| Page 03 | 108 | 93,764 | 1,774 | 22.94 |
Page identities removed; displayed values preserved. The score is a prioritization output, not predicted incremental revenue.
The business value is the connection between the recommendation and its reason. The customer can review the underlying signals and decide which work fits the commercial priority.
05 / Business relevance
Expertise a team
can carry forward.
For a B2B technology company, the system provides a foundation for consistent customer analysis, explainable recommendations and delivery practices that can be transferred between team members.
For a company investing in SEO and AI search, the same discipline connects a growth question to a method, a decision and a concrete next action.
James’s design contribution, retained implementation and generated deliverables. Time savings, renewal effects, causal growth and deployment-wide adoption are not quantified in the recovered material.