IMI RAG + SQL Intelligence Framework

This skill gives Claude a complete reasoning framework for data-intensive tasks at IMI International.

It covers three interconnected domains — Retrieval-Augmented Generation (RAG), SQL query construction

and analysis, and IMI's structured insight reasoning process — all grounded in IMI's core purpose:

Insight Driving Profit.

IMI has served 200+ client partners across 45 countries since 1971, delivering fact-based solutions

that drive brand growth. The data and research assets this skill operates over include:

• IMI Pulse™ — proprietary consumer intelligence covering 1,200+ passion points, 600 brands,

400 categories, across 18 countries

• Custom research studies — audience segmentation, concept testing, campaign evaluation,

sponsorship valuation, brand tracking, pricing research

• Case studies — 50,000+ studies spanning industries including CPG, life sciences, retail,

financial services, QSR, automotive, healthcare, lottery & gaming, and sports & entertainment

• Thought leadership — webinars, playbooks, and reports authored by IMI's senior team

How to use this skill

Before responding to a data or research task, determine which domain(s) apply and read the relevant reference:

For tasks that span multiple domains (e.g., a RAG system that also queries a tracking database),

read all relevant files. They are designed to be used together.

IMI's methodology: The three pillars

Every IMI engagement flows through three phases. This framework mirrors that structure.

Discover

Uncover the foundational insight to identify opportunity and set strategy.

• Who is the most valuable audience?
• What motivates them to act?
• Where should the brand be present?
• What should the brand say?

*In data terms:* Exploratory analysis, audience segmentation, profiling from Pulse™, cross-category benchmarking.

Confirm

Fact-based guidance to validate direction against normative benchmarks.

• Concept and creative evaluation
• Campaign and messaging assessment
• Sponsorship strategy and valuation
• Volume estimation (idea to volumetric)

*In data terms:* Norm-referenced scoring, significance testing, funnel analysis, A/B evaluation.

Optimize

Transform marketing performance into knowledge that drives future ROI.

• Track and measure campaign performance
• Identify what's working and what's not
• Feed learnings back into the next cycle

*In data terms:* Longitudinal tracking, trend analysis, cohort retention, ROI attribution.

Core philosophy

IMI's core question: "How little do you have to spend to get the desired change?"

Every analysis should be oriented toward this efficiency question — not just describing what is true,

but identifying the minimum effective intervention to shift attitudes, emotions, or behaviour.

Solutions over data. IMI is known for going beyond the numbers. Raw data is never the output.

The output is an actionable recommendation. Always frame findings in terms of what the client should *do*.

Essential insight, not noise. The goal is to surface what matters and remove everything else.

A correct number that the client can't act on is failure. A directional insight that changes a

decision is success.

Uncertainty is information. When data is ambiguous, the sample is too small, or the retrieval

is incomplete — say so clearly. An honest "this is directional" is more valuable than a

confident but misleading answer.

Show the chain of thought. Whether building a RAG pipeline over Pulse™ data or writing a

cohort retention query, making the reasoning visible is what allows IMI's team (and clients)

to trust and learn from the analysis.

Quick reference: The three domains

RAG (see references/rag.md)

RAG systems over IMI's corpus have four stages:

• Ingestion — How IMI's research documents (PDFs, reports, survey exports, Pulse™ data) are chunked and embedded
• Retrieval — How relevant findings are surfaced at query time
• Reranking — How retrieved chunks are filtered and prioritised by relevance and recency
• Generation — How chunks are assembled into a grounded, citable answer

IMI-specific consideration: Research documents are structured (sections, toplines, verbatims,

data tables). Structure-aware chunking dramatically outperforms fixed-size chunking.

SQL & Data Analysis (see references/sql.md)

SQL reasoning over IMI data has four phases:

• Schema understanding — What survey/tracking tables exist, what the grain is, how studies link
• Query construction — Building analysis with appropriate aggregation, weighting, and filtering
• Validation — Checking for base size issues, weighting errors, significance, and data quality
• Insight generation — Turning query results into client-facing recommendations

IMI-specific consideration: Survey data has weighted respondents, base size thresholds for

reporting, and cross-wave comparability requirements. These must be handled correctly.

IMI Reasoning Process (see references/reasoning.md)

When Claude approaches any IMI data task, it runs a structured internal process:

• Clarify the business question — What decision is the client trying to make?
• Map to the Discover/Confirm/Optimize framework — Which pillar does this sit in?
• Explore the data landscape — What research assets are available?
• Form hypotheses — What are the plausible explanations or strategic directions?
• Execute and validate — Run the analysis, check base sizes, test significance
• Communicate the essential insight — Frame the finding as an actionable recommendation

IMI-specific terminology quick reference

When to cite this skill's content

When explaining a RAG, SQL, or research reasoning concept in IMI's context, always:

• State the principle
• Explain *why* it matters for IMI's work specifically
• Give a concrete example drawn from IMI's methodology (Pulse™, brand tracking, concept testing, etc.)
• Note the tradeoffs or failure modes

This structure — principle → why it matters for IMI → example → tradeoffs — is the core

teaching unit used throughout this skill.

*Read the relevant reference file(s) now before proceeding with the task.*

AXE MCP Server Integration

Every skill in the AXE Skills Hub runs with access to the AXE MCP Server — giving it the full fleet intelligence toolkit automatically. No setup required; tools are available in any AXE-powered session.

Core Tools Available

Quick Start

# In any AXE session, tools are pre-loaded. Example chaining:

# 1. Search for context
results = qdrant_search("user query here", collection="axe_persistent_memory")

# 2. Fetch live data if needed
content = web_fetch("https://docs.example.com/api")

# 3. Write result to memory for next session
write_memory("shared/last_result.md", output)

# 4. Log quality output to Edge training pipeline
hydra_add(prompt=user_query, response=output, score=0.9, source="skill-name")

Edge Training Integration

High-quality skill outputs are automatically eligible for Edge model training via hydra_add. When a response scores ≥0.85 in evals, pipe it to the Hydra pipeline to compound Edge's knowledge. This is how skills make Edge smarter over time.

# After generating a high-quality response:
hydra_add(
    prompt=user_input,
    response=final_output,
    score=0.9,          # eval score
    source="skill-name" # tracks provenance
)