Part 1: Retention Framework: Defining Success and Measuring Value
1.1 Strategic Need: Shift from Acquisition Funnel to Retention Loop
Many e-commerce businesses are currently facing ‘High new influx, low repurchase rate’ The problem is not just a marketing challenge; it represents significant financial inefficiency. It is a proven fact that the cost of acquiring a new customer (CAC) is much higher than the cost of retaining an existing customer.
💡 KEY STATS:
- Existing customers compared to new customers Up to 5x higher conversion ratesIt looks like
- step 5% increase in customer retention rateThis company's profits Increase by up to 95%I can do it
- Email marketing in e-commerce and retail is average Amazing ROI of 4500%The fundamental reason for achieving this is to build effective relationships with an already established customer base.
The current business model does not maximize the lifetime value (LTV) of customers acquired at great cost, which is essentially wasting acquisition costsSame as: So a low repeat purchase rate means you're giving up this potential ROI and staying in the expensive, one-time acquisition game.
proposed in this report ‘CRM Retention Engine’is not just a communication tool. This optimizes the ROI of all customer acquisition costs executed in the past and establishes a sustainable profit structure by systematically nurturing customer assets. Key growth driversno see.
1.2 KPI Dashboard: Define, Calculate and Execute Key Performance Indicators
To accurately measure and optimize the performance of this retention engine, define a key performance indicator (KPI) dashboard. It acts as a measurement layer that monitors the health of the engine.
| KPI | Definitions and Formulas | strategic purpose | industry benchmark |
|---|---|---|---|
| Within 30/60/90 days Repurchase rate |
definition: Percentage of customers who make two or more purchases within a specific period (e.g. 30 days) (Number of customers who purchased more than twice during the period / Total number of customers who purchased) × 100
|
Measure the conversion speed of the ‘first purchase → repeat purchase’ loop | 30/60/90 day metrics are useful for cohort analysis, but the key is to fit the product life cycle |
| Dormant Conversion Rate (DCR) |
definition: Among customers who were dormant for a certain period of time, the percentage of customers who were reactivated (purchased) through a campaign (Number of dormant customers who purchased within the period / Total number of dormant customers at the beginning of the period) × 100
|
Directly measures the success of your reaction flow | Need to set internal goals |
| Customer Lifetime Value (LTV) |
definition: Total value created by one customer over the life of the relationship Formula 1 (Convenience Meal): (Average purchase amount) × (Purchase frequency) × (Average customer retention period)
Formula 2 (revenue-based):
(Total revenue in period / Total number of customers in period)
|
This is the metric that measures the ultimate success of your retention strategy. Increasing LTV allows you to afford a higher CPA, giving you an edge in the race to acquire new customers | Goal: LTV > CAC × 3 |
| customer retention rate (CRR) |
definition: Percentage of customers who remain with a business over a specific period of time [(Number of customers at the end of the period - Number of new customers in the period) / Number of beginning customers] × 100
|
It is a macro indicator that measures the stickiness of the entire customer base. | Need to set internal goals |
| CRM message Open rate/click rate |
definition: Open or click rate compared to messages sent | It is a micro indicator that diagnoses the effectiveness of message channels and content. |
• Average open rate (email): 26.6% • Average CTR (email): ~1.3% • Average CTR (Notification Talk): ~5.2% • Personalized CTA conversion rate: +202% |
🎯 KPI causal chain: These KPIs form a causal chain that is closely linked to each other. Personalized CTAs drive high CTRs, which leads to increased repurchase rates. A higher repeat purchase rate maximizes LTV, and higher LTV justifies more aggressive customer acquisition cost (CPA) execution. This means that optimizing a single CTA phrase in an automated message can have a ripple effect on a company’s entire growth strategy and budget structure.
Part 2: RFM Segment Model: ‘Who’ to Focus on
Retention engines shouldn't work the same for all customers. 'To whom' Customer segmentation, which determines what messages to deliver, acts as the brains of the engine. For this purpose, the most powerful and universal RFM modelUse .
2.1 RFM Methodology: Three Dimensions of Customer Value
RFM consists of three key metrics that measure value based on customer behavior:
- Recency (R - recency): How recently did the customer make a purchase? (The strongest predictor of future purchases)
- Frequency (F - frequency): How often did the customer purchase? (Measure of brand loyalty)
- Monetary (M - purchase amount): How much did the customer spend? (Measure of guest price and value)
2.2 Expert-level scoring: applying a hybrid model
Typically, each indicator is divided into 5 grades (1 to 5 points) and given a score, but the simple quintile method can lead to data distortion.
⚠️ IMPORTANT: In many e-commerce businesses, approximately 80% of all customers may have ‘purchase frequency (F) = 1’. If the F score is divided by percentile, these 80% of customers will all fall into the '1' group, making it impossible to distinguish between 'first-time purchasers' and 'customers who left after making their first purchase a year ago'.
Therefore, this engine has the following ‘Hybrid scoring model’We strongly recommend that you adopt:
- R (recency) and M (purchase amount) scoring: Scores are awarded based on percentile rank. This accurately identifies relative value. For example, the top 33.3% are allocated 5 points, the next lowest 26.7% are allocated 4 points, etc.
- F (Purchase Frequency) Scoring: Scores are awarded based on absolute count rather than percentile.
- One-time purchase: F=1 point
- 2nd purchase: F=2 points
- 3 purchases: F=3 points
- 4 purchases: F=4 points
- Purchase more than 5 times: F=5 points
This hybrid model precisely captures the relative superiority of R and M, while clearly separating the absolute behavioral differences in F (one-time and two-time buyers), enabling practical segmentation.
2.3 RFM segment definition and execution schedule
Based on the RFM score, customers are defined into 10 to 11 core segments and the automation engine's response strategy is established for each segment. This table is the basis for all ‘If-Then’ logic in the retention engine.
| RFM segment | RFM score | characteristic | Specific gravity (estimated) | Core Execution Strategy |
|---|---|---|---|---|
| core customer (Core VIP) |
R=5, F=5, M=5 | Customers who bought most recently, most often and spent the most: your highest-value customers. | 1-2% | High-Touch (1:1) management. Assignment of a dedicated manager. Excluding automatic discount messages. Exclusive benefits available |
| loyal customer (Loyal) |
R=4-5, F=4, M=4-5 | Active, high-value customers. The basis of the retention engine | 5-10% | Automation Nurture. Status awareness, new product pre-release. Engine trigger: anti-dormant scenario |
| potential loyal customers (Potential) |
R=4-5, F=2-3, M=3-4 | Purchased 2-3 times recently. Very high likelihood of becoming a loyal customer | 10-15% | Automated Upsell. The goal is to increase F (frequency). Engine Trigger: Cross-selling Scenario |
| new customer (New / FTB) |
R=5, F=1, M=1-5 | First purchase completed. Core management group with both risk of departure and potential | 20-30% | Automated Onboarding. The goal is to secure a second purchase. Engine Trigger: FTB Flow |
| Dormant Risk Customers (At-Risk) |
R=2-3, F=3-5, M=3-5 | Customers who had high value in the past, but whose recency (R) is rapidly decreasing | 15-20% | Automated dormancy prevention (Prevention). Key targets for ‘golden moment’ triggers based on purchasing cycle |
| Churning Customers (High Value) (Churned) |
R=1, F=4-5, M=4-5 | Previously best customers who have dropped out | 5% | Automation Reactivation (Benefit). Engine Trigger: Powerful Benefits Scenario |
| Churning Customers (Low Value) (Lost) |
R=1, F=1, M=1 | I left after purchasing once a long time ago. low value | 20-30% | Campaign Exclude. Avoid wasting your marketing budget |
Part 3: Identifying ‘Golden Moments’: Buying Cycle Analysis
If RFM defines ‘to whom’, the customer’s ‘Purchase Cycle’ Analysisis the engine's core trigger that determines 'when' to send a message. One-size-fits-all rules like “remind me in 30 days” fail. Predictions based on customers’ individual behavioral patterns are needed.
3.1 Strategic meaning of ‘cycle of use’
A 'cycle of use' based approach transforms CRM from a 'sales' to a 'service':
- ❌ Reactive Sales: “It’s been a while. Get a discount coupon.”
- ✅ Proactive Service: “The time has come when you need it. We will give you tips on how to use the product.”
3.2 Four statistical models for purchasing cycle calculation
To increase the precision of your engine, you need to calculate and utilize four types of purchasing cycle metrics:
1. User Average Purchase Cycle (UAPC)
use: It is a personalized indicator that predicts each user’s next purchase point.
2. Brand average purchase cycle (BPC: Brand Purchase Cycle)
use: This is a standard indicator that measures the average repurchase speed across brands.
3. User-Product Average Purchase Cycle (UPAPC)
use: Predict when a specific user will repurchase a specific product (e.g. consumables).
4. Average purchase cycle for each product (PAPC: Product Purchase Cycle)
use: It is a standard indicator that measures the average repurchase cycle of a specific product.
📊 Statistical Precision: Individual users (UAPC, UPAPC) are likely to be consistent with their past behavior, so the average value has predictive power. However, brand (BPC) and product (PAPC) indicators are greatly influenced by extreme outliers such as bulk buyers or futures buyers. Using average values in this case can seriously distort the period. Therefore, in order to exclude the influence of outliers and find a stable reference point, Be sure to use the medianYou must do it. This statistical rigor ensures the engine's accuracy.
3.3 Utilizing the purchasing cycle as a ‘proactive trigger’
Many traditional CRMs only send messages after the purchase cycle has elapsed. This means the customer is already on the churn path, and the message becomes “We lost you.”
Because the purpose of the retention engine is to ‘prevent dormancy’, Proactive triggerYou must use . In other words, we target the ‘golden moment’ right before the purchase cycle expires.
- Engine Trigger 1 (Consumable): PAPC - 5 days (e.g. for 30-day cycle consumables, ships on the 25th day)
- Engine Trigger 2 (Normal Sleep): UAPC
This 'golden moment' trigger is the core logic that induces natural repurchase by delivering a message based on 'use value' rather than discount pressure, right before the customer recognizes the need.
Part 4: Automated Retention Scenarios (Engine 'Behavior')
Now, by combining RFM ('Who') and purchasing cycle ('When'), we design the specific 'What' the engine will perform, i.e., the automation scenario.
4.1 Retention Loop Framework: Onboarding → Nurturing → Reactivation
All automation scenarios ‘Onboarding new customers’, ‘Nurturing existing customers (preventing dormancy)’, ‘Activating lost customers’It is structured around a three-step loop called .
4.2 Scenario 1: First-time buyer (FTB) onboarding sequence
Who: RFM Segment “New / FTB” (R=5, F=1)
Goal: Secure second purchase (move to F=2) and collect customer data (reviews)
Automation Sequence:
- trigger: Complete your first purchase
- D+1 Day (Value Offered):
- Channel: Email
- Content: “First Purchase Appreciation Guide: 1-Minute Tips for 200% Utilization of [Purchased Product Name]”
- Day D+7 (data collection):
- Channel: Notification Talk
- Content: “{{Customer Name}}, you will receive 3,000 P immediately when you write your first purchase review.”
- Day D+21 (Induce repurchase):
- Channel: Email / App Push
- Content: “I recommend a coordinating item that perfectly matches the [product purchased] you purchased last time.”
4.3 Scenario 2: Proactive anti-dormant sequence
Who: RFM Segment “At-Risk” (R=2-3)
Goal: Proactive notification based on purchase cycle (UAPC) to naturally encourage repurchase
Automation Sequence (2-Step Escalation Strategy):
- Trigger 1 (value first): (today - last purchase date) > (UAPC × 0.9) or PAPC - 5 days
- Message 1 (Provide Value):
- Channel: App Push / Email
- Content: “Pro-Tip: The Professional Guide for [Purchased Item]” (No Discount)
- Trigger 2 (Incentive): 10 days have passed since message 1 was sent + no purchases made
- Message 2 (light benefits):
- Channel: Notification Talk
- Content: "Are you sure you need [Product Name]? We have prepared a 10% repurchase coupon for {{Customer Name}}."
Strategy Commentary: This two-stage escalation flow respects the principle of 'avoiding discounting' and sends value-based messages first, but applies a realistic 'safety net' (light discount) to prevent churn if the value proposition fails.
4.4 Scenario 3: Benefit-based reactivation flow
Who: RFM Segment “Current Customers (High Value)” (R=1, F=4-5)
Goal: Bring back high-value core customers with powerful benefits
Automation Sequence:
- trigger: RFM = "churn (high value)" AND last purchase date > 90 days
- Message 1 (Powerful Benefits):
- Channel: Email
- Content: ‘1 month free trial’ of subscription service or “25% discount coupon for returning {{customer name}}”
- Message 2 (Urgency): 3 days have passed since message 1 was sent + no purchases made
- Channel: Notification Talk
- Content: "{{Customer Name}}, your special 'Welcome Back' 25% offer expires in 48 hours."
Strategy Commentary: Customers in this segment do not respond to a typical 10% discount. In order to break their indifference, you need to create a justification for their return by providing exceptional and exclusive benefits that regular customers cannot receive.
Part 5: A ‘Value in Use’ Based Copywriting Set
5.1 Strategy: Why ‘Value’ Beats ‘Discount’
The key to successful CRM messaging is ‘Providing continuous value’It is in Most failed CRM copies do not meet the ‘6 Ws’. The message “Special discount offer!” is unclear as to what or why to act.
the other side, Copy based on ‘value in use’clarifies the rule: "(Customer) purchased Product A 30 days ago (When). Use these tips to extend the life of the product (Why)."
5.2 ‘Value in Use’ Copywriting A/B Testing Matrix
Below is a set of copy for A/B testing that you can put into practice right away. The optimal plan must be derived by measuring the performance (CTR, CVR) of 'Variation (value-based plan)' compared to 'Control (existing discount plan)'.
| Scenario/Goal | Control (discount-based copy) | Variation (value-based copy) | copy basis |
|---|---|---|---|
| Goal 1: Purchase cycle reminder |
"[10% OFF] Customer, you didn't forget to repurchase, right? Repurchase right now!" |
title: Tips for using [Product Name] from {{Customer Name}}! "It's been about 30 days since you purchased it, did you know the [hidden function/tip]? Use it smarter with this 1-minute guide." |
Provides specific ‘how-to’ information based on the six-down principle. Approach ‘service’ rather than sales |
| Goal 2: cross selling |
"New products in stock! Buy now and get 15% off. Check out our popular new products." |
title: Perfect styling suggestion for [Product Name] "Match the [product name] purchased by {{customer name}} with 'this item' selected by three stylists this season. Check the lookbook." |
Provides informational value of coordination/styling rather than discounts |
| Goal 3: Reactivation |
“I missed you! Special 20% discount coupon only for {{customer name}}!” |
title: Best sellers that {{customer name}} is missing "More than 1,200 customers recently left 5-star reviews on [our new bestseller]. Check out our most loved products now." |
Stimulate curiosity and trust by using ‘social proof’ rather than ‘discount’ as the core motivation |
| Goal 4: Shopping Cart Abandonment |
“You forgot an item! Complete your order now with a 10% coupon.” |
title: [Product name] is waiting in your shopping cart. "P.S. This product ranks #1 in customer satisfaction [e.g. dry skin]. Don't forget to experience the benefits." |
Strengthen purchase decisions by reminding customers of their purchase history and the core value of the product (e.g., No. 1), rather than price discounts. |
Part 6: Strategic Balance: 1:1 CRM (High-Touch) vs. Automated CRM (Low-Touch)
‘Adjusting the proportion of 1:1 CRM and automated CRM’ is the core of retention strategy. These two models are not mutually exclusive; Strategically distribute according to customer segmentsIt has to be.
6.1 High-touch vs. low-touch model definition
- Low-Touch / Automation CRM: It is also called ‘Tech-touch’. It is 1:N (one to many) communication using technology (automation, segmentation). Highly scalable, cost-effective, and data-driven. The retention engine designed in this report itself is a low-touch model.
- High-Touch / 1:1 CRM: It is a 1:1 (one-to-one) human-based approach. It's resource-intensive and expensive, but it builds customer relationships at the deepest level. This should be reserved for your top customers with high average order value (AOV) or LTV.
6.2 Weight adjustment strategy using RFM segments
‘1:1 to automation’ is not determined by a single ratio such as 80:20. The mechanism by which the RFM model itself determines this proportionno see.
Low-Touch Applicable (approximately 95% of customers)
Target: All RFM segments except “Voice Customers (VIPs)” (“New Customers,” “Potential Loyal Customers,” “Dormantic Customers at Risk,” “churners,” etc.)
Strategy: These 95% of customers are 100% managed through the automated retention engine designed in Part 4. This is a scalable strategy that delivers maximum efficiency with minimal resources.
High-Touch application target (approximately 5% of customers)
Target: Part of the RFM segments “Core VIP” (R=5, F=5, M=5) and “Loyal Customers” (R=4-5, F=4)
Strategy: This top 5% of customers should be excluded from standard automation scenarios (especially discount messaging). Instead, they should be assigned to a ‘VIP Manager’ or ‘CX Specialist’ and receive one-on-one human management.
Examples of 1:1 activities:
- Providing pre-purchase opportunities through personal contact before new product launches
- Send a personalized thank you card or small gift
- 1:1 styling consulting or product usage consultation
- Special invitation to brand offline events
⚖️ Optimal balance point: This hybrid model is the most delicate balance that maximizes the profitability of the entire business by providing ‘scalable efficiency’ for 95% of customers and ‘irreplaceable loyalty’ for 5% of core customers.
Part 7: Comprehensive and strategic implementation roadmap
7.1 End-to-End Retention Engine Flow Chart
The overall operational flow of the retention engine incorporating all components designed in this report is as follows:
- [Customer completes first purchase]
- → RFM Segment Identification: “New Customers (FTB)” (Part 2)
- → Scenario 1 Autorun: FTB Onboarding (Part 4)
- → (‘Review request’ sent on D+7, ‘Cross-sell’ message sent on D+21)
- → [Customer, 2nd purchase completed]
- → RFM Reclassification: “Potential Loyal Customers” (Part 2)
- → [System: Start calculating the customer’s UAPC (Average Purchase Cycle per User)] (Part 3)
- ... [Time lapse, no additional purchases]...
- → [Trigger Occurs: (Today - Last Purchase Date) > (UAPC × 0.9)] (Part 3)
- → [Engine: Automatic execution of ‘Scenario 2: Dormant Prevention (1st: Value Message)’] (Parts 4 and 5)
- → [Customer completes 3rd purchase] (Loop successful, returns to number 7)
- [Alternative route: No purchase after number 10]
- → [Trigger Occurrence: (10 days after 1st message)]
- → [Engine: Automatic execution of ‘Scenario 2: Dormant Prevention (2nd: Light Discount)’] (Part 4)
- → RFM Reclassification: “Current Customers (High Value)” (Part 2)
- → [Engine: Automatic execution of ‘Scenario 3: Reactivation (Powerful Benefits)’] (Part 4)
- → [Customer return or permanent departure]
7.2 Final recommendations and implementation roadmap
We propose a five-step implementation roadmap to successfully build this retention engine:
5-step implementation roadmap
- Step 1 (Data): Develop and implement data logic to calculate RFM scoring (Part 2) and the four purchasing cycles (Part 3).
- Stage 2 (Technology): Build three core automation scenarios (FTB, Dormant Prevention, Reactivation) within your CRM solution (e.g. Braze, Amplitude, in-house developed, etc.).
- Step 3 (Content): A/B test ‘value in use’ copy to verify and optimize performance compared to existing discount messaging.
- Step 4 (Strategy): We introduce the high-touch/low-touch model in Part 6. Define ‘key customer (VIP)’ segments, exclude them from automated discount campaigns, and transfer them to dedicated personnel.
- Step 5 (Measurement): Build a KPI dashboard and start monitoring daily/weekly. ‘Repurchase rate within 90 days’ and ‘LTV’ as North Star indicatorsWe continuously upgrade our engines.
Are you ready to build your CRM retention engine?
Design an automation loop from first purchase to repeat purchase today.
CRM retention engine construction consultation