1
Observed customer behaviour
The model observes how recently customers interacted, how often they buy or engage, and how much economic value they generate. This behavioural base is commonly summarised as RFM.
DOC ROI · Portfolio Health Supervision · Γ
Customer portfolio health: stronger, stable or under pressure?
A continuous supervision model for customer behaviour, future value and customer-loss risk. Work with portfolio data, the Don Espadín applied example, or a 1,000-customer simulation controlled through simple sliders.
01 · Understand before calculation
Portfolio health without the technical fog
The purpose of this pill is not to memorise formulas. It is to learn how to watch a customer portfolio continuously, detect deterioration early and decide which commercial or experience lever deserves attention.
2
Expected future customer value
Future customer value is estimated from the expected strength and duration of the relationship. Longer relationships, greater frequency or higher-value purchases normally increase expected value.
3
Expected customer-loss risk
Churn represents the proportion of customers expected to leave. Lower churn means stronger retention, and stronger retention protects future value.
4
Comparison with a reference
Current value and retention are compared with a reference: a previous period, a target, a benchmark or an authorised business baseline.
5
Evolution over time
A healthy portfolio can deteriorate slowly. The time dimension matters because one snapshot may hide a trend that becomes expensive later.
6
Corrective action
The result is useful only if it points to a decision: protect retention, rebuild value, improve experience, grow frequency or preserve a healthy pattern.
Recency · How long since the last interaction?
Lower is usually better. A customer who bought 10 days ago is normally more active than one who last bought 180 days ago.
Example: 25 days since the last purchase.
Frequency · How often does the customer return?
Higher frequency usually signals habit, usefulness or loyalty.
Example: 6 purchases during the last 12 months.
Monetary value · How much value did the customer generate?
In this lab it is the economic amount generated during the chosen observation period.
Example: €420 during the last 12 months.
NPS · Would the customer recommend the brand?
A customer-experience signal from −100 to +100. It supports the reading of relationship quality; it does not replace future-value or churn evidence.
Example: +55 is a positive recommendation signal.
LTV · Lifetime Value · What might this customer relationship be worth while it remains active?
It represents the future economic value expected from the relationship over a defined horizon.
Example: expected future value of €380.
Churn · What share of customers may leave?
If churn is 15%, roughly 15 out of every 100 customers are expected to be lost during that period, subject to the model and evidence used.
15% churn means approximately 85% retained.
Retention · What share remains?
Retention is the opposite view of churn. If churn is 15%, retention is 85%.
Lower churn → higher retention → healthier future value, all else equal.
Reference · Compared with what?
The health signal needs a sensible baseline. A reference can be last year, an approved target or another valid benchmark using the same definitions.
Current LTV €380 vs reference LTV €350.
Healthy portfolio
Future value and retention are at or above the chosen reference. Management focus: protect what works and scale it without damaging experience.
Customers stay, but value is weak
Retention is holding, yet future value is below reference. Management focus: frequency, cross-sell, premium mix, proposition and next-best action.
Value is strong, but customers may leave
The portfolio is economically attractive but fragile. Management focus: friction, complaints, loyalty, service recovery and early-warning outreach.
Both value and retention are weak
This is the most demanding situation. Management focus: diagnose causes before pushing sales, recover the relationship and rebuild expected value.
Important: a company-wide average can look acceptable while one customer segment is deteriorating quickly. That is why the laboratory shows both the total portfolio and the segment view.
Formal KAI·ROI layer
One short technical note — and then back to management
Within KAI·ROI v1, Γg(i),t is the portfolio-health factor for group g(i) at time t. It combines expected customer value and retention relative to a reference. RFM can support CC inside SPO and can also support LTV/churn prediction, but it does not replace Γ. Team and resource productivity remain in Pi.
Γ = weight of future value × relative LTV + weight of retention × relative retention
Evidence discipline: Unknown is not zero. Missing data remains missing; simulated or assumed values are labelled as such.
02 · Interactive laboratory
Build, load or explore a portfolio
Three operating routes are available: create a synthetic portfolio from selected averages, upload customer data, or load the Don Espadín applied example. Results update immediately when the controls change.
Upload portfolio CSV
Use the provided template. Keep the same definitions and time horizon for current and reference values. The browser performs the analysis locally.
Don Espadín · premium mezcal applied case
The case contains four illustrative customer groups across two time periods. It is designed to show how one portfolio can contain healthy segments and deteriorating segments at the same time.
Illustrative evidence: Don Espadín values are simulated case assumptions, not observed commercial data.
LIVE PORTFOLIO VIEW · updates with every adjustment
What the simulator is doing
From one set of averages to 1,000 illustrative customers
The model creates a synthetic portfolio around the selected assumptions. Customer values vary around the averages, so the result behaves like a portfolio rather than a single representative customer. Four behavioural groups support supervision: Core, Growth, Watch and At Risk.
Simulated customers1,000
Enough cases to make the charts stable while keeping the exercise fast in the browser.
Time snapshots4
The portfolio moves gradually from its reference towards the selected current values.
Customer variation20%
No two customers are identical. This shows how an average can hide different risk pockets.
Evidence statusSimulated
The result is illustrative. It must never be presented as observed company evidence.
Sensitivity check: Generate the portfolio, then reduce churn by 3–5 percentage points. The health score, trend, segment table and corrective measures update immediately.
No portfolio loaded yet. Simulate a company, upload a CSV or load Don Espadín.
Portfolio health—Latest period
Future value vs reference—Expected customer value
Retention vs reference—Customers staying
Customers in weak groups—Share needing attention
Management reading
Load a portfolio to begin.
The management reading identifies the main driver of the result and the first area requiring investigation.
WeakerReferenceStronger
Health over time
Is the portfolio improving or deteriorating?
Value × retention map
Which customer groups combine future value and staying power?
Main result drivers
Plain-language sensitivity view.
Corrective measures
Actions will appear after a portfolio is loaded.
| Customer group | Customers | Future value vs ref. | Retention vs ref. | Health Γ | Scenario Γ | Status | Behaviour signal |
|---|---|---|---|---|---|---|---|
| No data yet. | |||||||
Evidence & supervision
Keep the evidence status visible. Simulated, estimated and observed values are not the same thing.
Evidence status will appear after a portfolio is loaded.
03 · Minimum data required
Eight business ideas, translated into simple fields
A large data warehouse is not required to start. The minimum requirement is a consistent definition set, a customer identifier, a group, a time stamp and variables describing behaviour, future value and customer-loss risk.
R
Recency
Days since the last meaningful interaction. A lower number usually indicates a more active customer.
F
Frequency
How many times the customer bought or interacted during the chosen period.
M
Monetary
The economic value generated in the observation period.
N
NPS
A recommendation and experience signal. Useful for diagnosis, but not a substitute for future-value or churn evidence.
V
Future value · LTV
The value expected from the customer relationship over the selected future horizon.
C
Churn
The proportion of customers expected to leave during the chosen period. Lower churn means higher retention.
customer_idUnique customer identifier.segmentCustomer group used for supervision.snapshot_dateObservation date, format YYYY-MM-DD.recency_daysDays since the last meaningful interaction.frequency_12mInteractions or purchases in the last 12 months.monetary_12mEconomic value generated in the last 12 months.npsRecommendation signal from −100 to +100, when available.ltv_currentCurrent or expected future customer value.ltv_referenceFuture-value baseline using the same horizon.churn_currentCurrent or expected churn as a decimal: 0.15 = 15%.churn_referenceReference churn using the same time period.evidence_typereal, estimated, proxy, assumed or simulated.1. Enter averages
The sliders describe the centre of the simulated portfolio: typical recency, frequency, value, NPS, expected future value and churn.
2. The lab creates variation
It generates 1,000 synthetic customer profiles around those values. The customer-diversity slider controls how dispersed they are.
3. Four snapshots are created
The simulation moves gradually from the reference towards the current assumptions, producing a simple time trend.
4. Test sensitivity
Move future value or churn and observe how the health score, chart, groups and suggested actions change.
Simulation rule: the generated 1,000 customers are illustrative data. They illustrate sensitivity and portfolio mechanics; they are not a prediction about a real company.
04 · LEGO-style operating guide
Use the same routine every month
Choose a route
Start with the simulator, upload a portfolio CSV, or load Don Espadín to inspect the mechanics first.
Check definitions
Make sure churn, LTV and references use the same time horizon. Do not compare monthly churn with annual churn.
Read the total
Look at portfolio health, future value, retention and the share of customers sitting in weak groups.
Open the groups
Find out whether the problem comes from low expected value, weak retention or both.
Move one lever
Reduce churn or change future value in the sensitivity bars. Observe how quickly the portfolio responds.
Repeat over time
Refresh the data weekly or monthly. Portfolio supervision is valuable because it detects direction, not just a single score.
Common troubleshooting: if the result looks extreme, check percentages vs decimals, duplicate customers, inconsistent segment definitions, different LTV horizons, or a reference period that is not comparable with the current period.