code-ai-installer 2.5.0 → 4.0.0

package/domains/analytics/locales/en/.agents/skills/rfm-analysis/SKILL.md

@@ -1,314 +1,314 @@
----
-name: rfm-analysis
-description: Customer segmentation by recency, frequency, and monetary value
----
-# RFM Analysis — Customer Segmentation by Recency, Frequency, and Monetary Value
-
-## When to Use
-- When building a retention strategy — to understand who to retain and who to reactivate.
-- When optimizing a marketing budget — to direct resources to the most valuable segments.
-- When launching personalized campaigns — different messaging for different segments.
-- When analyzing a customer base — identifying Champions, loyal, at-risk, and lost customers.
-
-## Input
-
-| Field | Required | Description |
-|------|:-----------:|----------|
-| Business / Product | ✅ | What we are analyzing |
-| Business Model | ✅ | E-commerce / SaaS / Marketplace / Service |
-| Transaction Data | ⬚ | Table: customer, date, amount (if real data is available) |
-| Analysis Period | ⬚ | For what period (default — 12 months) |
-| Number of Customers | ⬚ | Base size to calibrate the scale |
-| Average Check | ⬚ | For modeling, if real data is missing |
-| Purchase Frequency | ⬚ | Average frequency for modeling |
-
-> If mandatory fields are not provided — **request them from the user** before starting the analysis. Do not make assumptions.
-
-> If real data is missing — the skill operates with **modeled data**, explicitly marked as hypothetical. The methodology remains applicable when real data becomes available.
-
-## Data Sources
-1. **Transactional data** — CRM, payment system, ERP (if provided).
-2. **Web search** — industry benchmarks (average check, frequency, retention rate).
-3. **Industry reports** — typical RFM distributions for the given business type.
-4. **Product analytics** — user activity data (for SaaS — logins, actions).
-5. **Data from the user** — exports from CRM, Google Analytics, payment provider.
-
-> When working with modeled data — specify benchmark sources. Data older than 12 months must be marked ⚠️.
-
-### Relationship with Other Skills
-| Skill | What We Take | When to Call |
-|------|-----------|----------------|
-| `web-research` | Industry benchmarks: avg check, frequency, retention | If there is no real data — for modeling |
-| `cohort-analysis` | Retention curves for calibrating Recency thresholds (step 2) | If cohort data is available |
-| `competitive-analysis` | Competitor benchmarks for retention and LTV | To compare with market standards |
-| `icp-buyer-persona` | Champions profile → ICP enrichment (step 8) | After segmentation: Champions = ICP standard |
-| `unit-economics` | LTV per segment for calculating reactivation budget | To evaluate ROI by segments |
-| `customer-journey-mapping` | Churn points for At Risk / About to Sleep | To understand exactly where we are losing customers |
-
-## Protocol
-
-### Step 0 — Gathering Context
-1. Check for the presence of mandatory data.
-2. Determine the mode: **real data** or **modeled**.
-3. Record the analysis period.
-4. Define the adaptation for the business model:
-
-   | Business Model | Recency = | Frequency = | Monetary = |
-   |---------------|-----------|-------------|------------|
-   | E-commerce | Days since the last order | Number of orders | Total order amount |
-   | SaaS (subscription) | Days since the last login/action | Number of active months | MRR or subscription plan |
-   | SaaS (usage-based) | Days since the last usage | Number of sessions/actions | Total payment amount |
-   | Marketplace | Days since the last transaction | Number of transactions | GMV (Gross Merchandise Volume) |
-   | Service (offline) | Days since the last visit | Number of visits | Total check amount |
-
-### Step 1 — Defining the Data Scope
-- Define the unit of analysis: customer / account / subscriber.
-- Define a "transaction" for the given business model (see table in step 0).
-- Define the period: start and end dates.
-- Exclude anomalies: test accounts, refunds, abnormal amounts (>3σ).
-
-### Step 2 — Recency Score (R)
-1. Calculate the days since the last transaction for each customer.
-2. Define thresholds based on the business cycle:
-
-   | Score | E-commerce | SaaS (subscription) | Marketplace |
-   |:-----:|:----------:|:---------------:|:-----------:|
-   | R=5 | 0-30 days | 0-7 days | 0-14 days |
-   | R=4 | 31-90 | 8-30 | 15-45 |
-   | R=3 | 91-180 | 31-60 | 46-90 |
-   | R=2 | 181-365 | 61-120 | 91-180 |
-   | R=1 | >365 | >120 | >180 |
-
-   > Thresholds are guidelines. Calibrate to the actual distribution (quintiles) if data is available.
-
-### Step 3 — Frequency Score (F)
-1. Count the number of transactions per customer for the period.
-2. Divide into 5 groups (quintiles or business logic).
-3. F=5: top-20% by frequency. F=1: one-time / minimal.
-
-### Step 4 — Monetary Score (M)
-1. Calculate the total amount per customer for the period.
-2. Divide into 5 groups (quintiles or business logic).
-3. M=5: top-20% by amount. M=1: minimal spending.
-4. For fixed subscription SaaS: replace M with an **Engagement Score** (depth of use: features, integrations, number of users in the account).
-
-### Step 5 — Segmentation
-Assign an RFM code (111-555) to each customer and group them:
-
-| Segment | RFM Range | Description | Typical Share |
-|---------|:------------:|----------|:-------------:|
-| **Champions** | R=5, F=4-5, M=4-5 | Best customers, buy frequently and a lot | 5-10% |
-| **Loyal Customers** | R=4-5, F=3-5, M=3-5 | Regular and valuable | 10-15% |
-| **Potential Loyalists** | R=4-5, F=1-3, M=1-3 | New with growth potential | 10-15% |
-| **Recent Customers** | R=5, F=1, M=1 | Just arrived, first purchase | 5-10% |
-| **Promising** | R=3-4, F=1-2, M=1-2 | Show interest, need activation | 10-15% |
-| **Need Attention** | R=3, F=3, M=3 | Average, at risk of leaving | 5-10% |
-| **About to Sleep** | R=2-3, F=1-2, M=1-2 | Losing interest | 10-15% |
-| **At Risk** | R=1-2, F=3-5, M=3-5 | Were active and valuable, leaving | 5-10% |
-| **Cannot Lose** | R=1, F=5, M=5 | VIPs who left — critical to win back | 1-3% |
-| **Hibernating** | R=1-2, F=1-2, M=1-2 | Long inactive, low value | 10-15% |
-| **Lost** | R=1, F=1, M=1 | Lost | 5-10% |
-
-### Step 6 — Evaluating Segment Value
-For each segment, calculate:
-
-| Segment | % of base | % of revenue | Avg LTV | Revenue/customer | Budget Priority |
-|---------|:------:|:---------:|:-------:|:--------------:|:-----------------:|
-| Champions | X% | X% | $XXX | $XXX | High (retention) |
-| At Risk | X% | X% | $XXX | $XXX | High (reactivation) |
-| Lost | X% | X% | $XXX | $XXX | Low (let go) |
-
-**Pareto Principle:** usually Champions + Loyal = 15-25% of the base, but 60-80% of revenue. If the distribution differs — record it as an anomaly.
-
-### Step 7 — Strategy and Personalization by Segments
-
-| Segment | Goal | Action | Channel | Tone | Offer | KPI |
-|---------|------|----------|-------|-----|-------------|-----|
-| Champions | Retention + advocacy | Exclusive access, referral program | Email, personal | Appreciative, VIP | Early access, referral bonus | NPS, referral rate |
-| Loyal | Upsell + loyalty | Cross-selling, loyalty program | Email, in-app | Friendly, expert | Upgrade, bundle | ARPU, retention |
-| Potential Loyalists | Engagement + onboarding | Guided onboarding, educational email series | Email, in-app | Supportive | Bonus for the 2nd purchase | F→2+, activation |
-| Recent | Activation | Welcome series, aha-moment | Email, push | Welcoming | Promo code for the 2nd order | Day-7 retention |
-| At Risk | Reactivation | "We miss you", special offer | Email, SMS | Caring, urgent | 20% discount, personal offer | Win-back rate |
-| Cannot Lose | Urgent reactivation | Manager call, personal offer | Call, email | Personal, VIP | Max discount, upgrade | Win-back rate |
-| Lost | Win-back or let go | Final email, churn reason survey | Email | Neutral | Survey + last chance | Churn reason data |
-
-### Step 8 — Migration Analysis
-Track the movement of customers between segments:
-
-1. Compare RFM segments for the current and previous periods:
-
-   | Migration | Count | % | Interpretation | Action |
-   |----------|:----------:|:-:|---------------|----------|
-   | Champions → At Risk | [N] | X% | Loss of best customers — critical | Immediate reactivation |
-   | Potential Loyalists → Loyal | [N] | X% | Healthy growth — onboarding works | Scale the approach |
-   | At Risk → Lost | [N] | X% | Reactivation failed | Revise strategy |
-   | Recent → Promising | [N] | X% | Activation works | Continue |
-
-2. Key migrations to monitor:
-   - **Negative:** Champions/Loyal → At Risk, At Risk → Lost, Potential → Hibernating.
-   - **Positive:** Recent → Potential → Loyal → Champions.
-3. Recommended recalculation frequency:
-
-   | Business Model | Frequency |
-   |---------------|---------|
-   | E-commerce (FMCG) | Weekly |
-   | E-commerce (high-ticket) | Monthly |
-   | SaaS | Monthly |
-   | Marketplace | Bi-weekly |
-   | Service (offline) | Quarterly |
-
-## Example — E-commerce: Sports Supplements Store
-
-**Context:** Online sports supplements store, B2C, 12,000 customers, 12-month period. Average check $45, average frequency 3.2 purchases/year.
-
-### Evaluation Scale
-
-| Metric | 5 | 4 | 3 | 2 | 1 |
-|---------|:-:|:-:|:-:|:-:|:-:|
-| **Recency** (days) | 0-30 | 31-60 | 61-120 | 121-240 | >240 |
-| **Frequency** (orders) | ≥6 | 4-5 | 3 | 2 | 1 |
-| **Monetary** ($) | >$300 | $200-300 | $100-200 | $50-100 | <$50 |
-
-### Segments
-
-| Segment | % of base | % of rev. | Count | Avg LTV | Strategy |
-|---------|:------:|:---------:|:------:|:-------:|-----------|
-| Champions | 8% | 35% | 960 | $520 | Referral program, VIP access |
-| Loyal | 12% | 28% | 1440 | $340 | Upsell to subscription, loyalty program |
-| Potential Loyalists | 11% | 8% | 1320 | $105 | Onboarding: "3rd order = free shipping" |
-| At Risk | 7% | 12% | 840 | $250 | "We miss you" + 15% discount on favorite category |
-| Cannot Lose | 2% | 5% | 240 | $360 | Manager call, personal offer |
-| Lost | 15% | 2% | 1800 | $25 | Final email + churn reason survey |
-
-**Pareto Check:** Champions + Loyal = 20% of base, 63% of revenue ✅ (healthy distribution).
-
-### Migration Analysis (MoM)
-
-| Migration | Count | % | Signal |
-|----------|:------:|:-:|--------|
-| Champions → At Risk | 45 | 4.7% | 🔴 Loss of VIPs — launch win-back |
-| Potential → Loyal | 180 | 13.6% | 🟢 Onboarding works |
-| At Risk → Lost | 120 | 14.3% | 🟡 Reactivation isn't effective enough |
-
-**Recommendation:** Champions → At Risk = 4.7% — above benchmark (3%). Investigate: price? quality? competitor?
-
-## Validation (Quality Gate)
-
-- [ ] All three metrics (R, F, M) are defined with clear thresholds adapted to the business model
-- [ ] Quintile thresholds are justified by the business cycle (or real data quintiles)
-- [ ] A minimum of 6 named segments with RFM ranges are outlined
-- [ ] For each segment: strategy, action, channel, tone, KPI
-- [ ] Segment value is evaluated: % of base, % of revenue, average LTV
-- [ ] Pareto check: Champions + Loyal ≈ 15-25% of base, 60-80% of revenue
-- [ ] Migration analysis: key transitions between segments are recorded
-- [ ] Recalculation frequency is recommended for the given business model
-- [ ] If the data is modeled — it is explicitly stated with benchmark sources
-- [ ] No logical contradictions in segmentation (RFM codes are correct)
-- [ ] Data sources are specified; data older than 12 months is marked ⚠️
-
-> If validation fails — iterate until it passes, rather than handing off.
-
-## Handoff
-The output of `$rfm-analysis` is an input for:
-- **Strategist / Mediator** — customer segments for strategic decisions.
-- **`unit-economics`** — LTV by segments, retention rates, reactivation budget.
-- **`cohort-analysis`** — cohort grouping within segments.
-- **`customer-journey-mapping`** — churn points for At Risk / About to Sleep.
-- **`icp-buyer-persona`** — Champions profile = ICP standard.
-
-Handoff format: table of segments (% of base, % of revenue, LTV) + strategies + migration analysis. When transferring — use `$handoff`.
-
-## Anti-patterns
-
-| Error | Why it's bad | How to do it right |
-|--------|-------------|---------------|
-| Identical thresholds for all businesses | SaaS and e-commerce have different buying cycles | Calibrate thresholds to the business model (table in step 0) |
-| Modeled data without labels | Decisions based on hypotheses, not facts | Explicitly mark, specify benchmark sources |
-| Too many segments (>11) | Impossible to create a unique strategy for each | 6-11 named segments |
-| Segmentation without a strategy | RFM for the sake of RFM, no actionable insights | Every segment = goal + action + channel + KPI |
-| Ignoring M for subscriptions | Monetary is less variable in SaaS | Replace M with Engagement Score for SaaS |
-| Static analysis without migration | Cannot see how customers move between segments | Migration analysis: negative and positive transitions |
-| No personalization | Identical message for Champions and At Risk | Different tone, channel, offer, frequency per segment |
-| Segments without value | Unclear where to allocate budget | % of base + % of revenue + LTV per segment + Pareto check |
-| Benchmarks without a source | Impossible to verify | Every benchmark — with a source and date |
-
-## Output Template
-
-```markdown
-### RFM Analysis — [Business / Product]
-
-**Business Model:** [Model]
-**Analysis Period:** [Date — Date]
-**Base Size:** [N customers]
-**Data Type:** Real / Modeled (Benchmark: [Source])
-**Date:** [Date]
-
----
-
-#### Metric Adaptation
-
-| Metric | Definition for the Given Business |
-|---------|-------------------------------|
-| Recency | [What is measured: days since the last X] |
-| Frequency | [What is counted: number of X over the period] |
-| Monetary | [What is counted: total amount of X / engagement score] |
-
-#### Assessment Scale
-
-| Metric | Score 5 | Score 4 | Score 3 | Score 2 | Score 1 |
-|---------|:-------:|:-------:|:-------:|:-------:|:-------:|
-| **Recency** | 0-[X] | [X]-[X] | [X]-[X] | [X]-[X] | >[X] |
-| **Frequency** | >[X] | [X]-[X] | [X]-[X] | [X]-[X] | [X] |
-| **Monetary** | >[X] | [X]-[X] | [X]-[X] | [X]-[X] | <[X] |
-
----
-
-#### Customer Segments
-
-| Segment | RFM Range | % Base | % Rev. | Count | Avg LTV | Priority |
-|---------|:------------:|:------:|:---------:|:------:|:-------:|:---------:|
-| Champions | R5, F4-5, M4-5 | X% | X% | [N] | $XXX | High |
-| Loyal | R4-5, F3-5, M3-5 | X% | X% | [N] | $XXX | High |
-| ... | ... | ... | ... | ... | ... | ... |
-
-**Pareto Check:** Champions + Loyal = X% of base, X% of revenue → [Healthy / Anomalous]
-
----
-
-#### Strategy and Personalization
-
-| Segment | Goal | Action | Channel | Tone | Offer | KPI |
-|---------|------|----------|-------|-----|-------------|-----|
-| Champions | [Goal] | [Action] | [Channel] | [Tone] | [Offer] | [KPI] |
-| At Risk | [Goal] | [Action] | [Channel] | [Tone] | [Offer] | [KPI] |
-
----
-
-#### Migration Analysis
-
-| Migration | Count | % of Segment | Signal | Action |
-|----------|:------:|:----------:|:------:|----------|
-| Champions → At Risk | [N] | X% | 🔴 | [What to do] |
-| Potential → Loyal | [N] | X% | 🟢 | [What to do] |
-| At Risk → Lost | [N] | X% | 🟡 | [What to do] |
-
-**Recalculation Frequency:** [weekly / monthly / quarterly]
-
----
-
-#### Recommendations
-1. **Immediate:** [At Risk / Cannot Lose — what to do now]
-2. **Medium-term:** [Potential Loyalists — how to activate]
-3. **Long-term:** [Champions — loyalty program / advocacy]
-4. **Recalculation:** every [N] days
-
----
-
-#### Sources and Assumptions
-
-| # | Fact / Assumption | Source | Date | Reliability |
-|---|-------------------|----------|------|:-------------:|
-| 1 | [Fact] | [Source] | [Date] | ✅ / ⚠️ / 🔮 |
-
-Legend: ✅ Verified (real data) | ⚠️ Estimated (benchmark) | 🔮 Assumed (modeling)
-```
+---
+name: rfm-analysis
+description: Customer segmentation by recency, frequency, and monetary value
+---
+# RFM Analysis — Customer Segmentation by Recency, Frequency, and Monetary Value
+
+## When to Use
+- When building a retention strategy — to understand who to retain and who to reactivate.
+- When optimizing a marketing budget — to direct resources to the most valuable segments.
+- When launching personalized campaigns — different messaging for different segments.
+- When analyzing a customer base — identifying Champions, loyal, at-risk, and lost customers.
+
+## Input
+
+| Field | Required | Description |
+|------|:-----------:|----------|
+| Business / Product | ✅ | What we are analyzing |
+| Business Model | ✅ | E-commerce / SaaS / Marketplace / Service |
+| Transaction Data | ⬚ | Table: customer, date, amount (if real data is available) |
+| Analysis Period | ⬚ | For what period (default — 12 months) |
+| Number of Customers | ⬚ | Base size to calibrate the scale |
+| Average Check | ⬚ | For modeling, if real data is missing |
+| Purchase Frequency | ⬚ | Average frequency for modeling |
+
+> If mandatory fields are not provided — **request them from the user** before starting the analysis. Do not make assumptions.
+
+> If real data is missing — the skill operates with **modeled data**, explicitly marked as hypothetical. The methodology remains applicable when real data becomes available.
+
+## Data Sources
+1. **Transactional data** — CRM, payment system, ERP (if provided).
+2. **Web search** — industry benchmarks (average check, frequency, retention rate).
+3. **Industry reports** — typical RFM distributions for the given business type.
+4. **Product analytics** — user activity data (for SaaS — logins, actions).
+5. **Data from the user** — exports from CRM, Google Analytics, payment provider.
+
+> When working with modeled data — specify benchmark sources. Data older than 12 months must be marked ⚠️.
+
+### Relationship with Other Skills
+| Skill | What We Take | When to Call |
+|------|-----------|----------------|
+| `web-research` | Industry benchmarks: avg check, frequency, retention | If there is no real data — for modeling |
+| `cohort-analysis` | Retention curves for calibrating Recency thresholds (step 2) | If cohort data is available |
+| `competitive-analysis` | Competitor benchmarks for retention and LTV | To compare with market standards |
+| `icp-buyer-persona` | Champions profile → ICP enrichment (step 8) | After segmentation: Champions = ICP standard |
+| `unit-economics` | LTV per segment for calculating reactivation budget | To evaluate ROI by segments |
+| `customer-journey-mapping` | Churn points for At Risk / About to Sleep | To understand exactly where we are losing customers |
+
+## Protocol
+
+### Step 0 — Gathering Context
+1. Check for the presence of mandatory data.
+2. Determine the mode: **real data** or **modeled**.
+3. Record the analysis period.
+4. Define the adaptation for the business model:
+
+   | Business Model | Recency = | Frequency = | Monetary = |
+   |---------------|-----------|-------------|------------|
+   | E-commerce | Days since the last order | Number of orders | Total order amount |
+   | SaaS (subscription) | Days since the last login/action | Number of active months | MRR or subscription plan |
+   | SaaS (usage-based) | Days since the last usage | Number of sessions/actions | Total payment amount |
+   | Marketplace | Days since the last transaction | Number of transactions | GMV (Gross Merchandise Volume) |
+   | Service (offline) | Days since the last visit | Number of visits | Total check amount |
+
+### Step 1 — Defining the Data Scope
+- Define the unit of analysis: customer / account / subscriber.
+- Define a "transaction" for the given business model (see table in step 0).
+- Define the period: start and end dates.
+- Exclude anomalies: test accounts, refunds, abnormal amounts (>3σ).
+
+### Step 2 — Recency Score (R)
+1. Calculate the days since the last transaction for each customer.
+2. Define thresholds based on the business cycle:
+
+   | Score | E-commerce | SaaS (subscription) | Marketplace |
+   |:-----:|:----------:|:---------------:|:-----------:|
+   | R=5 | 0-30 days | 0-7 days | 0-14 days |
+   | R=4 | 31-90 | 8-30 | 15-45 |
+   | R=3 | 91-180 | 31-60 | 46-90 |
+   | R=2 | 181-365 | 61-120 | 91-180 |
+   | R=1 | >365 | >120 | >180 |
+
+   > Thresholds are guidelines. Calibrate to the actual distribution (quintiles) if data is available.
+
+### Step 3 — Frequency Score (F)
+1. Count the number of transactions per customer for the period.
+2. Divide into 5 groups (quintiles or business logic).
+3. F=5: top-20% by frequency. F=1: one-time / minimal.
+
+### Step 4 — Monetary Score (M)
+1. Calculate the total amount per customer for the period.
+2. Divide into 5 groups (quintiles or business logic).
+3. M=5: top-20% by amount. M=1: minimal spending.
+4. For fixed subscription SaaS: replace M with an **Engagement Score** (depth of use: features, integrations, number of users in the account).
+
+### Step 5 — Segmentation
+Assign an RFM code (111-555) to each customer and group them:
+
+| Segment | RFM Range | Description | Typical Share |
+|---------|:------------:|----------|:-------------:|
+| **Champions** | R=5, F=4-5, M=4-5 | Best customers, buy frequently and a lot | 5-10% |
+| **Loyal Customers** | R=4-5, F=3-5, M=3-5 | Regular and valuable | 10-15% |
+| **Potential Loyalists** | R=4-5, F=1-3, M=1-3 | New with growth potential | 10-15% |
+| **Recent Customers** | R=5, F=1, M=1 | Just arrived, first purchase | 5-10% |
+| **Promising** | R=3-4, F=1-2, M=1-2 | Show interest, need activation | 10-15% |
+| **Need Attention** | R=3, F=3, M=3 | Average, at risk of leaving | 5-10% |
+| **About to Sleep** | R=2-3, F=1-2, M=1-2 | Losing interest | 10-15% |
+| **At Risk** | R=1-2, F=3-5, M=3-5 | Were active and valuable, leaving | 5-10% |
+| **Cannot Lose** | R=1, F=5, M=5 | VIPs who left — critical to win back | 1-3% |
+| **Hibernating** | R=1-2, F=1-2, M=1-2 | Long inactive, low value | 10-15% |
+| **Lost** | R=1, F=1, M=1 | Lost | 5-10% |
+
+### Step 6 — Evaluating Segment Value
+For each segment, calculate:
+
+| Segment | % of base | % of revenue | Avg LTV | Revenue/customer | Budget Priority |
+|---------|:------:|:---------:|:-------:|:--------------:|:-----------------:|
+| Champions | X% | X% | $XXX | $XXX | High (retention) |
+| At Risk | X% | X% | $XXX | $XXX | High (reactivation) |
+| Lost | X% | X% | $XXX | $XXX | Low (let go) |
+
+**Pareto Principle:** usually Champions + Loyal = 15-25% of the base, but 60-80% of revenue. If the distribution differs — record it as an anomaly.
+
+### Step 7 — Strategy and Personalization by Segments
+
+| Segment | Goal | Action | Channel | Tone | Offer | KPI |
+|---------|------|----------|-------|-----|-------------|-----|
+| Champions | Retention + advocacy | Exclusive access, referral program | Email, personal | Appreciative, VIP | Early access, referral bonus | NPS, referral rate |
+| Loyal | Upsell + loyalty | Cross-selling, loyalty program | Email, in-app | Friendly, expert | Upgrade, bundle | ARPU, retention |
+| Potential Loyalists | Engagement + onboarding | Guided onboarding, educational email series | Email, in-app | Supportive | Bonus for the 2nd purchase | F→2+, activation |
+| Recent | Activation | Welcome series, aha-moment | Email, push | Welcoming | Promo code for the 2nd order | Day-7 retention |
+| At Risk | Reactivation | "We miss you", special offer | Email, SMS | Caring, urgent | 20% discount, personal offer | Win-back rate |
+| Cannot Lose | Urgent reactivation | Manager call, personal offer | Call, email | Personal, VIP | Max discount, upgrade | Win-back rate |
+| Lost | Win-back or let go | Final email, churn reason survey | Email | Neutral | Survey + last chance | Churn reason data |
+
+### Step 8 — Migration Analysis
+Track the movement of customers between segments:
+
+1. Compare RFM segments for the current and previous periods:
+
+   | Migration | Count | % | Interpretation | Action |
+   |----------|:----------:|:-:|---------------|----------|
+   | Champions → At Risk | [N] | X% | Loss of best customers — critical | Immediate reactivation |
+   | Potential Loyalists → Loyal | [N] | X% | Healthy growth — onboarding works | Scale the approach |
+   | At Risk → Lost | [N] | X% | Reactivation failed | Revise strategy |
+   | Recent → Promising | [N] | X% | Activation works | Continue |
+
+2. Key migrations to monitor:
+   - **Negative:** Champions/Loyal → At Risk, At Risk → Lost, Potential → Hibernating.
+   - **Positive:** Recent → Potential → Loyal → Champions.
+3. Recommended recalculation frequency:
+
+   | Business Model | Frequency |
+   |---------------|---------|
+   | E-commerce (FMCG) | Weekly |
+   | E-commerce (high-ticket) | Monthly |
+   | SaaS | Monthly |
+   | Marketplace | Bi-weekly |
+   | Service (offline) | Quarterly |
+
+## Example — E-commerce: Sports Supplements Store
+
+**Context:** Online sports supplements store, B2C, 12,000 customers, 12-month period. Average check $45, average frequency 3.2 purchases/year.
+
+### Evaluation Scale
+
+| Metric | 5 | 4 | 3 | 2 | 1 |
+|---------|:-:|:-:|:-:|:-:|:-:|
+| **Recency** (days) | 0-30 | 31-60 | 61-120 | 121-240 | >240 |
+| **Frequency** (orders) | ≥6 | 4-5 | 3 | 2 | 1 |
+| **Monetary** ($) | >$300 | $200-300 | $100-200 | $50-100 | <$50 |
+
+### Segments
+
+| Segment | % of base | % of rev. | Count | Avg LTV | Strategy |
+|---------|:------:|:---------:|:------:|:-------:|-----------|
+| Champions | 8% | 35% | 960 | $520 | Referral program, VIP access |
+| Loyal | 12% | 28% | 1440 | $340 | Upsell to subscription, loyalty program |
+| Potential Loyalists | 11% | 8% | 1320 | $105 | Onboarding: "3rd order = free shipping" |
+| At Risk | 7% | 12% | 840 | $250 | "We miss you" + 15% discount on favorite category |
+| Cannot Lose | 2% | 5% | 240 | $360 | Manager call, personal offer |
+| Lost | 15% | 2% | 1800 | $25 | Final email + churn reason survey |
+
+**Pareto Check:** Champions + Loyal = 20% of base, 63% of revenue ✅ (healthy distribution).
+
+### Migration Analysis (MoM)
+
+| Migration | Count | % | Signal |
+|----------|:------:|:-:|--------|
+| Champions → At Risk | 45 | 4.7% | 🔴 Loss of VIPs — launch win-back |
+| Potential → Loyal | 180 | 13.6% | 🟢 Onboarding works |
+| At Risk → Lost | 120 | 14.3% | 🟡 Reactivation isn't effective enough |
+
+**Recommendation:** Champions → At Risk = 4.7% — above benchmark (3%). Investigate: price? quality? competitor?
+
+## Validation (Quality Gate)
+
+- [ ] All three metrics (R, F, M) are defined with clear thresholds adapted to the business model
+- [ ] Quintile thresholds are justified by the business cycle (or real data quintiles)
+- [ ] A minimum of 6 named segments with RFM ranges are outlined
+- [ ] For each segment: strategy, action, channel, tone, KPI
+- [ ] Segment value is evaluated: % of base, % of revenue, average LTV
+- [ ] Pareto check: Champions + Loyal ≈ 15-25% of base, 60-80% of revenue
+- [ ] Migration analysis: key transitions between segments are recorded
+- [ ] Recalculation frequency is recommended for the given business model
+- [ ] If the data is modeled — it is explicitly stated with benchmark sources
+- [ ] No logical contradictions in segmentation (RFM codes are correct)
+- [ ] Data sources are specified; data older than 12 months is marked ⚠️
+
+> If validation fails — iterate until it passes, rather than handing off.
+
+## Handoff
+The output of `$rfm-analysis` is an input for:
+- **Strategist / Mediator** — customer segments for strategic decisions.
+- **`unit-economics`** — LTV by segments, retention rates, reactivation budget.
+- **`cohort-analysis`** — cohort grouping within segments.
+- **`customer-journey-mapping`** — churn points for At Risk / About to Sleep.
+- **`icp-buyer-persona`** — Champions profile = ICP standard.
+
+Handoff format: table of segments (% of base, % of revenue, LTV) + strategies + migration analysis. When transferring — use `$handoff`.
+
+## Anti-patterns
+
+| Error | Why it's bad | How to do it right |
+|--------|-------------|---------------|
+| Identical thresholds for all businesses | SaaS and e-commerce have different buying cycles | Calibrate thresholds to the business model (table in step 0) |
+| Modeled data without labels | Decisions based on hypotheses, not facts | Explicitly mark, specify benchmark sources |
+| Too many segments (>11) | Impossible to create a unique strategy for each | 6-11 named segments |
+| Segmentation without a strategy | RFM for the sake of RFM, no actionable insights | Every segment = goal + action + channel + KPI |
+| Ignoring M for subscriptions | Monetary is less variable in SaaS | Replace M with Engagement Score for SaaS |
+| Static analysis without migration | Cannot see how customers move between segments | Migration analysis: negative and positive transitions |
+| No personalization | Identical message for Champions and At Risk | Different tone, channel, offer, frequency per segment |
+| Segments without value | Unclear where to allocate budget | % of base + % of revenue + LTV per segment + Pareto check |
+| Benchmarks without a source | Impossible to verify | Every benchmark — with a source and date |
+
+## Output Template
+
+```markdown
+### RFM Analysis — [Business / Product]
+
+**Business Model:** [Model]
+**Analysis Period:** [Date — Date]
+**Base Size:** [N customers]
+**Data Type:** Real / Modeled (Benchmark: [Source])
+**Date:** [Date]
+
+---
+
+#### Metric Adaptation
+
+| Metric | Definition for the Given Business |
+|---------|-------------------------------|
+| Recency | [What is measured: days since the last X] |
+| Frequency | [What is counted: number of X over the period] |
+| Monetary | [What is counted: total amount of X / engagement score] |
+
+#### Assessment Scale
+
+| Metric | Score 5 | Score 4 | Score 3 | Score 2 | Score 1 |
+|---------|:-------:|:-------:|:-------:|:-------:|:-------:|
+| **Recency** | 0-[X] | [X]-[X] | [X]-[X] | [X]-[X] | >[X] |
+| **Frequency** | >[X] | [X]-[X] | [X]-[X] | [X]-[X] | [X] |
+| **Monetary** | >[X] | [X]-[X] | [X]-[X] | [X]-[X] | <[X] |
+
+---
+
+#### Customer Segments
+
+| Segment | RFM Range | % Base | % Rev. | Count | Avg LTV | Priority |
+|---------|:------------:|:------:|:---------:|:------:|:-------:|:---------:|
+| Champions | R5, F4-5, M4-5 | X% | X% | [N] | $XXX | High |
+| Loyal | R4-5, F3-5, M3-5 | X% | X% | [N] | $XXX | High |
+| ... | ... | ... | ... | ... | ... | ... |
+
+**Pareto Check:** Champions + Loyal = X% of base, X% of revenue → [Healthy / Anomalous]
+
+---
+
+#### Strategy and Personalization
+
+| Segment | Goal | Action | Channel | Tone | Offer | KPI |
+|---------|------|----------|-------|-----|-------------|-----|
+| Champions | [Goal] | [Action] | [Channel] | [Tone] | [Offer] | [KPI] |
+| At Risk | [Goal] | [Action] | [Channel] | [Tone] | [Offer] | [KPI] |
+
+---
+
+#### Migration Analysis
+
+| Migration | Count | % of Segment | Signal | Action |
+|----------|:------:|:----------:|:------:|----------|
+| Champions → At Risk | [N] | X% | 🔴 | [What to do] |
+| Potential → Loyal | [N] | X% | 🟢 | [What to do] |
+| At Risk → Lost | [N] | X% | 🟡 | [What to do] |
+
+**Recalculation Frequency:** [weekly / monthly / quarterly]
+
+---
+
+#### Recommendations
+1. **Immediate:** [At Risk / Cannot Lose — what to do now]
+2. **Medium-term:** [Potential Loyalists — how to activate]
+3. **Long-term:** [Champions — loyalty program / advocacy]
+4. **Recalculation:** every [N] days
+
+---
+
+#### Sources and Assumptions
+
+| # | Fact / Assumption | Source | Date | Reliability |
+|---|-------------------|----------|------|:-------------:|
+| 1 | [Fact] | [Source] | [Date] | ✅ / ⚠️ / 🔮 |
+
+Legend: ✅ Verified (real data) | ⚠️ Estimated (benchmark) | 🔮 Assumed (modeling)
+```