IMI Client Deliverable Framework

This skill teaches the AI to write research deliverables that meet IMI's exacting standards:

every sentence calibrated to evidence, every insight actionable, every recommendation

grounded in data.

IMI's Communication Standard

Three non-negotiable principles:

• Never overclaim. Do not state something as definitive when the evidence is directional.
• Never over-hedge. Do not bury a clear finding in unnecessary caveats.
• Always recommend. Data without a recommendation is a failure. The client pays for a decision, not a dashboard.

IMI's refund guarantee means every deliverable must meet the agreed objective. A deliverable

that provides accurate data but fails to answer the client's business question has failed.

The Insight Hierarchy

Not all findings are equal. IMI uses a three-tier hierarchy:

Tier 1: Essential Insight

The single most important finding that changes the client's decision.

• There is usually only ONE essential insight per study
• It answers the original business question directly
• It reframes how the client thinks about the problem
• Example: "Your brand's consideration decline is not a brand problem — it's a category

relevance problem. Consumers are leaving the category, not choosing competitors."

Tier 2: Supporting Evidence

The 3-5 data points that substantiate the essential insight.

• These prove the essential insight is correct
• They should be presented in a logical narrative flow
• Each one should be expressed as a finding + implication (not just a number)
• Example: "Consideration dropped 6pts among 25-34s (from 48% to 42%, sig. at 95%) —

this cohort is the growth engine and their decline accounts for 80% of the total drop."

Tier 3: Context

Everything else — the broader data landscape that provides background.

• Important for completeness but not for the recommendation
• Should be available in an appendix or data tables, not in the main narrative
• The client should never have to read Tier 3 to understand the recommendation

The golden rule: If the client reads ONLY Tier 1 and Tier 2, they should have everything

they need to make a decision. Tier 3 is for the detail-oriented.

Language Calibration

The Evidence-Strength Scale

Every finding must be expressed with language calibrated to the evidence:

Language Dos and Don'ts

DO:

• "The data shows X, which means the client should Y" (finding + implication)
• "Brand consideration is 42% (down 6pts, significant at 95%)" (number + context + significance)
• "This is the most important finding because..." (hierarchy signalling)

DON'T:

• "Interestingly, X" (vague — why is it interesting? what should the client do about it?)
• "It could be argued that..." (over-hedging — if the data supports it, say so)
• "42% of consumers said..." (raw data without interpretation)
• "Clearly, definitively, X" when the base is n=47 (overclaiming)

Deliverable Structures

Structure 1: Topline (2-5 pages)

The topline is delivered within 48 hours of fieldwork completion. It is the first read

and often the most important deliverable.

## [Study Name] — Topline
**Client:** [name] | **Category:** [name] | **Fieldwork:** [dates]
**Base:** n=[n] | **Methodology:** [online/CATI/in-person]

### The Business Question
[What the client needed to know — restate in plain language]

### Essential Insight
[ONE paragraph — the single most important finding. Bold the key sentence.
This paragraph should answer the business question directly.]

### Key Findings
1. **[Finding headline]** — [1-2 sentences with data points, significance, and implication]
2. **[Finding headline]** — [1-2 sentences]
3. **[Finding headline]** — [1-2 sentences]
[Maximum 5 key findings]

### Recommendation
[What the client should DO — specific, actionable, tied to findings]

### Next Steps
[What IMI recommends as the follow-up — specific IMI capability or action]

### Data Note
[Base sizes, weighting methodology, any caveats — brief]

Structure 2: Full Report (15-40 pages)

## [Study Name] — Full Report

### 1. Executive Summary (1-2 pages)
- Business question
- Essential insight
- Key findings (3-5 bullets)
- Recommendation

### 2. Methodology (1 page)
- Sample design and size
- Fieldwork dates and mode
- Weighting approach
- Significance conventions used in the report

### 3. Context (2-3 pages)
- Category landscape
- Competitive context
- Client's strategic context

### 4. Findings (8-25 pages)
- Organised by theme, NOT by question number
- Each section: Finding → Evidence → Implication
- Visuals (charts, tables) with clear titles and callouts
- Significance flags on all comparisons

### 5. Synthesis & Recommendations (2-4 pages)
- How findings connect to each other
- The essential insight (expanded)
- 3-5 specific recommendations, each tied to evidence
- Prioritisation: what to do first, second, third

### 6. Appendix
- Full data tables
- Questionnaire
- Sample profile
- Statistical notes

Structure 3: Strategic Recommendation

## Recommendation: [Title]

### The Situation
[Brief context — what the client is facing]

### What the Data Shows
[3-5 key data points with calibrated language]

### What This Means
[Interpretation — connect the dots between data points]

### What We Recommend
[Specific action(s) — with rationale for each]

### What We Expect
[Expected outcome — what metrics should move, by how much, over what timeframe]

### Risk
[What could go wrong and how to mitigate]

Chart and Table Standards

Chart Title Standard

Every chart must have:

• A descriptive title that states the finding (not just the metric)
• Base size noted
• Significance indicators

Good: "Brand A leads on consideration (+8pts vs. nearest competitor, n=500, sig. at 95%)"

Bad: "Consideration by brand"

Table Standards

| Brand | Awareness | Consideration | Preference | NPS |
|---|---|---|---|---|
| Brand A | 89% | 48%▲ | 22% | +32▲ |
| Brand B | 85% | 41% | 19% | +28 |
| Brand C | 72%▼ | 35%▼ | 15% | +18▼ |
| **Norm** | **82%** | **40%** | **18%** | **+24** |

▲ Significantly above norm (95%) | ▼ Significantly below norm (95%)
Base: n=500 per brand | Weighted data

The "So What?" Test

Every sentence in an IMI deliverable must pass the "So What?" test:

Statement: "42% of consumers in the 25-34 age group consider Brand X."
So what? → "This is 6 points below the category norm and represents the brand's
            weakest demographic — the cohort that drives 35% of category value."
So what? → "If the brand does not address this, it risks losing the next generation
            of category buyers to competitors who are investing in this age group."
So what? → "We recommend a Discover phase using Pulse™ to understand what this cohort
            values and a targeted concept test for repositioned messaging."

The THIRD "so what?" is the recommendation. That is what goes in the deliverable.

Evidence-Strength Mapping Template

For complex reports with multiple findings at different evidence levels:

### Evidence Map

| Finding | Evidence Level | Key Data Point | Base | Sig? | Confidence |
|---|---|---|---|---|---|
| PI above norm | STRONG | 62% vs 55% norm | n=300 | Yes (95%) | High |
| Gen Z over-index | DIRECTIONAL | Index 135, n=67 | n=67 | Marginal | Low |
| Trust erosion | MODERATE | -4pts wave-on-wave | n=250 | Yes (90%) | Medium |

**Reporting guidance:**
- STRONG findings: State with confidence, use in headline recommendations
- MODERATE findings: Include in supporting evidence, flag significance level
- DIRECTIONAL findings: Report as indications, recommend further investigation
- INSUFFICIENT: Note in methodology caveats, do not include in findings

Common Pitfalls

• Leading with data, not insight. "42% said X" is data. "The brand's growth depends

on solving the under-35 relevance gap" is insight. Lead with insight.

• Burying the recommendation. If the recommendation is on page 30, the client may

never read it. The topline and executive summary must contain the recommendation.

• Presenting all findings equally. Not every data point matters. The insight hierarchy

exists for a reason — use Tier 1/2/3 ruthlessly.

• Inconsistent language calibration. Using "the data clearly shows" for both a sig.

result (n=500) and a directional result (n=45) destroys credibility.

• Forgetting the business question. Every deliverable must explicitly answer the

question the client asked. If the data doesn't answer it, say so and recommend how to find out.

Cross-Skill References

• For all analytical skills that produce findings → all other imi-* skills
• For SQL to extract data for deliverables → imi-rag-sql-intelligence
• For pitch/proposal versions of deliverables → imi-pitch-intelligence

*Built for IMI International's Local AI — grounded in IMI's 50+ year standard of

purposeful advisory: reveal what the client couldn't see before.*

*Purpose: Insight. Method: Rigour. Outcome: Profit.*

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
)