Customer Lifetime Value Modeling

Customer Analytics Predictive Retention + Monetization ← Simpler Version

Estimate forward-looking customer lifetime value using two jointly estimated behavioral models: a retention model that predicts the probability a customer stays active each period, and a conditional spend model that forecasts revenue when active. Adjust marketing levers to simulate CLV impact and evaluate ROI before you spend.

QUICK START: Choose Your Path

OVERVIEW & WHAT THIS TOOL SOLVES

Two customers, identical histories. CLV modeling reveals divergent futures. A targeted intervention (⚡) at the right moment shifts Customer B's trajectory — and delivers measurable return.

Not all customers are worth the same, yet organizations routinely allocate budgets using backward-looking averages like past spend or simple RFM tiers. Two customers who each bought twice last year might have wildly different futures—one is a deal-hunter about to churn, while the other is beginning a multi-year relationship. Customer Lifetime Value (CLV) Modeling fits a dual regression system—a logistic retention & reactivation model paired with an OLS conditional spend model—to forecast each customer's discounted future cash flows and simulate the ROI of targeted marketing interventions.

🎯 Questions It Answers

What is the forward-looking dollar value (CLV) of each individual customer and segment over our planning horizon?
Which marketing levers primarily keep customers active (retention) versus increasing basket size when they buy (spend)?
If we invest in more samples, credits, or CSM hours for a specific segment, does the incremental CLV lift beat the intervention cost?

📂 What Data You Need

Customer-Period Panel Data: Customer ID, time period index (1, 2, ...), binary activity flag (0/1), and period spend
Marketing Lever Columns: Period-level touchpoints (e.g., samples sent, discount dollars, CSM hours)
Segment & Attribute Columns: Optional customer segments, acquisition channels, or cohorts, plus a periodic discount rate

💡 What You'll Discover

Portfolio & Segment KPIs: Mean CLV, Median CLV, Total Portfolio Value, Expected Active Periods, and CLV/CAC ratios
Dual-Model Lever Matrix: Side-by-side linear and log-saturation coefficients showing diminishing returns on retention and spend
What-If ROI Simulator: Gross CLV lift, intervention cost, net lift, and ROI from repricing or reallocating marketing levers

🔍 Key Nuances & Common Traps to Watch For

Why Two Separate Models? Combining retention and spend into one regression confuses loyal low-spenders with sporadic big-spenders. Modeling whether they buy (logistic) separately from how much they spend when active (OLS) lets you target retention vs. upsell levers independently.
Reactivation vs. Contractual Survival: Because prior-period activity is a predictor, the model estimates both a stay probability (when active last period) and a reactivation probability (when lapsed last period). In non-contractual retail, reactivation keeps future activity from decaying to zero—so expected active periods is NOT simply 1 / (1 − retention).
Automatic Diminishing Returns (Linear + Log Pairs): For every marketing lever you select, the tool enters both the raw value and its ln(m + 1) transform into both regressions so the data can reveal concave saturation without manual feature engineering.

📐 Under the Hood: Model Equations & Variable Definitions

1. Discounted Customer Lifetime Value (CLV)

\[ \text{CLV}_k = \sum_{t=1}^{T} \frac{\hat{P}(\text{active}_{k,t}) \cdot \hat{E}[\text{spend}_{k,t} \mid \text{active}] \cdot g - c_k}{(1 + r)^t} \]

💬 Read it in plain English: A customer's lifetime value is the sum across all future planning periods of their probability of being active in that period times their expected spend when active, times the gross margin, minus the marketing cost declared for that customer, discounted back to today's dollars.

CLV_k

Customer Lifetime ValueOUTPUT ⓘ

Discounted present value of expected future margin, net of declared marketing cost, for customer k across the horizon (revenue at the 100% margin default with no costs declared).

Forecast HorizonYOUR DATA ⓘ

Number of future periods (e.g., 24 months) over which CLV is projected.

Periodic Discount RateYOUR DATA ⓘ

Firm cost of capital per period (e.g., 0.01 = 1% per month) used to discount future cash flows.

(1 + r)^t

Present-Value Discount FactorINFERENCE ⓘ

Compound discounting denominator that reduces the weight of cash flows further in the future.

P̂(active_k,t)

Predicted Activity ProbabilityOUTPUT ⓘ

Estimated probability that customer k is active in future period t (from the retention state-transition recursion).

Ê[spend_k,t | active]

Predicted Conditional SpendOUTPUT ⓘ

Estimated dollar spend for customer k in period t given that they are active (from the OLS spend model).

Gross MarginYOUR DATA ⓘ

Share of revenue kept as gross profit, from Gross Margin (%); the 100% default leaves spend as revenue.

c_k

Per-Period Marketing CostYOUR DATA ⓘ

Declared unit cost times customer k's usual lever volume, charged every period whether or not the customer is active; 0 when no costs are declared.

2. Retention & Reactivation State-Transition Recursion

\[ P(\text{active}_t) = P(\text{active}_{t-1}) \cdot p^{\text{stay}} + \bigl(1 - P(\text{active}_{t-1})\bigr) \cdot p^{\text{react}} \]

💬 Read it in plain English: The chance a customer is active in period t equals the chance they were active last period and stayed, plus the chance they were inactive last period and reactivated.

p^stay

Retention Probability (Active → Active)OUTPUT ⓘ

Predicted probability of remaining active when active_t−1 = 1: σ(α + β₁ + xβ).

p^react

Reactivation Probability (Lapsed → Active)OUTPUT ⓘ

Predicted probability of returning after an inactive period when active_t−1 = 0: σ(α + xβ).

P(active_t−1)

Prior-Period Activity ProbabilityOUTPUT ⓘ

Probability that the customer was in the active state in period t − 1.

S_t = ∏ p^stay

Unbroken Survival BenchmarkINFERENCE ⓘ

Probability of never having a single quiet period through period t (reported for comparison; CLV uses P(active_t)).

3. Logistic Retention Model with Log-Saturation

\[ \text{logit}\bigl[P(\text{active}_t)\bigr] = \alpha + \beta_1 \cdot \text{active}_{t-1} + \sum_{j=1}^{J} \left[ \beta_j^{\text{lin}} m_j + \beta_j^{\text{sat}} \ln(m_j + 1) \right] + \mathbf{X}\boldsymbol{\gamma} \]

💬 Read it in plain English: The log-odds of a customer being active this period depend on their baseline propensity, whether they were active last period, both linear and diminishing-return effects of each marketing lever, and their segment attributes.

Retention InterceptESTIMATED ⓘ

Baseline log-odds of activity when lagged activity and marketing levers are zero.

active_t−1

Lagged Activity IndicatorYOUR DATA ⓘ

Binary indicator (1 if active in period t − 1, 0 if inactive) capturing habit/inertia.

β₁

Habit / State-Dependence CoefficientESTIMATED ⓘ

Increase in log-odds of activity from having been active in the immediately preceding period.

m_j, ln(m_j + 1)

Raw & Log-Saturated Marketing Lever jYOUR DATA ⓘ

Level of marketing intervention j in raw units and its concave log transformation.

β_j^lin, β_j^sat

Retention Linear & Saturation SlopesESTIMATED ⓘ

Estimated linear and diminishing-return effects of marketing lever j on retention log-odds.

X, γ

Customer Attributes & WeightsESTIMATED ⓘ

Row vector of customer attribute/segment indicators X and their estimated retention effects γ.

4. Conditional Spend Model (Active-Only OLS)

\[ E[\text{spend}_t \mid \text{active}_t = 1] = \alpha_s + \sum_{j=1}^{J} \left[ \delta_j^{\text{lin}} m_j + \delta_j^{\text{sat}} \ln(m_j + 1) \right] + \mathbf{X}\boldsymbol{\phi} \]

💬 Read it in plain English: Among customers who are active in a period, expected dollar spend is a linear combination of baseline spend, linear and saturating marketing lever effects, and customer segment characteristics.

α_s

Baseline Active Spend InterceptESTIMATED ⓘ

Expected dollar spend per active period when all marketing levers and attribute dummies are at baseline.

δ_j^lin, δ_j^sat

Spend Linear & Saturation SlopesESTIMATED ⓘ

Estimated dollar impact of marketing lever j (raw and log-transformed) on basket size when active.

X, ϕ

Customer Attributes & Spend WeightsESTIMATED ⓘ

Customer attribute/segment indicators X and their dollar effects ϕ on conditional spend.

Number of Marketing LeversYOUR DATA ⓘ

Count of marketing intervention variables selected in the column mapper.

DATA SOURCE

Load a marketing use case:

📊 CLV Case Studies

Load a preset scenario with pre-configured customer data to explore CLV modeling approaches across different business types.

CLV projects forward from one retention rate. See the retention curve for when customers actually leave.

Upload a period-level CSV: one row per customer per time period. Required columns: customer_id, period, is_active, spend. Optional: segment and any marketing / attribute columns.

Drag & Drop raw data file (.csv, .tsv, .txt, .xls, .xlsx)

Period-level longitudinal format — one row per customer per period

📊 CLV Analysis Results

—

Median CLV

—

Mean CLV

—

Portfolio CLV

—

Retention P(active | was active)

—

Reactivation P(active | had lapsed)

—

Expected Active Periods over the horizon

—

Avg Spend (Active)

APA-Style Report

Fit the model to see the statistical summary.

Managerial Interpretation

Fit the model to see what the projection does and does not support.

View:

📊 How to Read This Chart

Each period on the x-axis represents a future time step. The y-axis shows the expected revenue contribution from the average customer in that period — this is the product of the estimated retention probability and expected conditional spend, discounted to present value. The curve naturally declines over the horizon as churn compounds: even customers who are active today face some probability of churning in each subsequent period.

Key insight: The steepness of the decay reflects your retention rate. A 90% monthly retain rate gives a much flatter curve than 70% — and that difference translates directly into dramatically different CLV totals.

Segment models: When your data includes a segment column with sufficient observations per group, the tool fits a separate retention and spend model for each segment. Each customer's CLV is then computed using their own segment's model parameters rather than a pooled average — so a high-spend segment's CLV isn't diluted by lower-spend segments, and vice versa. You can inspect per-segment coefficients on the Model Coefficients tab using the segment dropdown.

Average CLV for customers grouped by value decile. D1 = lowest 10% of customers by CLV, D10 = highest 10%.

💡 How to Use CLV Deciles

The decile chart reveals concentration of value in your customer base. If D10 is many times larger than D1–D5, a small share of customers drives the majority of lifetime revenue — a common pattern in subscription and B2B contexts.

D9–D10: Your highest-value customers. Prioritize retention programs, dedicated account management, and early churn detection for this group. Losing even one D10 customer can outweigh gaining several D1–D5 customers.
D5–D8: Growth candidates. These customers are already showing positive engagement; upsell and cross-sell programs are likely to find receptive audiences here.
D1–D3: Review acquisition economics. If CAC for these customers approaches or exceeds their CLV, the acquisition investment is not recovering. Consider whether onboarding interventions could shift new customers toward higher-decile trajectories.

💡 Reading the Coefficient Tables

Retention model coefficients are on the log-odds scale. Positive = increases probability of staying active.
Every marketing variable appears twice — once as the raw value (free_samples) and once log-transformed (log_free_samples). This applies to all marketing variables in both models. The model estimates both forms simultaneously and lets the data determine how much of each variable's effect is linear vs. saturating.
Saturation signature: look for a positive log coefficient paired with a smaller or negative raw coefficient — this means diminishing returns. A significant raw coefficient and near-zero log coefficient means the effect is roughly linear over the observed range.
Spend model coefficients are in raw dollar (or revenue unit) terms per unit change in the predictor. The same raw + log pair structure applies here too.
Cross-effects: A marketing variable appearing significant in both models is influencing both retention and spend — relevant for budget allocation decisions.
p < 0.05 highlighted in green; non-significant in grey.
Per-segment models: When your data includes a segment column and each segment has enough observations, the tool fits separate retention and spend models per segment — capturing the fact that one marketing lever may drive retention among high-value customers but be irrelevant to budget-sensitive ones. Use the View coefficients for dropdown above to switch between the pooled model (which averages across everyone) and any individual segment model. The pooled model is shown by default as a reference; the CLV engine itself uses each customer's own segment model when one is available.
Export Coefficients CSV downloads all model estimates (pooled + per-segment when applicable) in a tidy CSV suitable for reporting or further analysis.

What is each lever associated with?

One lever varied across low, typical and high, with every other lever held at its norm and customer attributes left as they are. The model was fitted once — choosing a lever here changes what you are looking at, not the specification, so the rows stay comparable.

Marginal effects at a glance

Each dot is a marketing variable, positioned by its estimated effect on retention (x-axis) and conditional spend (y-axis). Variables in the upper-right are dual drivers — they improve both whether customers stay and how much they spend when active. The elasticity table below translates raw coefficients into business-interpretable marginal effects.

💡 How to Use This Matrix

Upper-right (Dual Driver) — increases both retention and spend. Highest ROI potential; cutting this variable likely harms CLV from two directions simultaneously.
Lower-right (Retention Driver) — keeps customers active but doesn't increase spend per period. Prioritize when churn is your primary problem.
Upper-left (Spend Driver) — increases revenue when active but doesn't improve retention. Useful for monetization campaigns targeting already-loyal customers.
Lower-left (Neither significant) — no detected effect in the data. These variables either don't matter, or lack sufficient variation to estimate effects reliably. Reconsider whether they belong in the model.
Log-term interactions: Note that every continuous marketing variable also has a log-transformed version estimated simultaneously (capturing saturation). Distance from zero is the linear component only; the full marginal effect at a given value is shown in the elasticity table.

Retention & Spend Elasticity at Mean Values

Marginal effects computed at each variable's mean — accounting for both the linear and log-saturation terms simultaneously. Percentage-point change in retention probability; dollar change in spend per active period.

Marketing Variable	Mean Value	↑ Retention Prob. (pp per +1 unit)	↑ Active-Period Spend ($ per +1 unit)	Strategic Role

How CLV is distributed across your customer base — and which customers represent the most at-risk value.

🚨 Churn Risk Triage

💡 Using Segment CLV for Strategy

CLV by segment answers: where is future customer value concentrated, and what should you do about it? Use the table's Avg Activity Rate and Avg Spend columns together to diagnose why a segment ranks high or low, then choose the right lever. Avg Activity Rate is the share of observed periods a segment's customers were active: the table's read on staying power (“retention” below), not the model's period-to-period retention probability.

High CLV, high retention, high spend — your most valuable customers. Prioritize retention: even a 5-point drop in retention compounds heavily over the forecast horizon. Invest in loyalty programs, early churn detection, and dedicated account management.
High CLV, high retention, moderate spend — loyal but under-monetized. Strong candidates for upsell and cross-sell campaigns. Retention is already healthy; incremental spend has outsized effect on CLV here because it applies across many future periods.
Moderate CLV, low retention, high spend — high spend per active period but poor staying power. Focus on re-engagement and churn prevention rather than spend stimulation. Fixing retention will increase CLV more than marketing spend increases would.
Low CLV, low retention, low spend — least efficient segment. Re-evaluate acquisition cost vs. lifetime return. If CAC is high for this segment, consider re-allocating budget to higher-CLV segments or redesigning onboarding to shift new customers toward higher-value behavioral patterns early.

Compounding insight: CLV differences between segments reflect both factors simultaneously — a segment that retains 10% better and spends 10% more per active period compounds to a far larger lifetime value gap over a multi-year horizon than either difference alone would suggest.

Customer Lifetime Value Modeling

👨‍🏫 Professor Mode: Guided Learning Experience

QUICK START: Choose Your Path

Try an Example

Upload My Data

Learn About CLV

OVERVIEW & WHAT THIS TOOL SOLVES

🎯 Questions It Answers

📂 What Data You Need

💡 What You'll Discover

1. Discounted Customer Lifetime Value (CLV)

2. Retention & Reactivation State-Transition Recursion

3. Logistic Retention Model with Log-Saturation

4. Conditional Spend Model (Active-Only OLS)

DATA SOURCE

Use a Case Study

📊 CLV Case Studies

Upload Customer Data

🗂️ Map Your Columns

Marketing Intervention Columns

Customer Attribute Columns (optional)

⚙️ Model Configuration

💰 Declare Marketing Action Costs

📊 CLV Analysis Results

Strategic Snapshot — What These Results Mean for Marketing Decisions

APA-Style Report

Managerial Interpretation

What is each lever associated with?

Marginal effects at a glance

Retention & Spend Elasticity at Mean Values

🚨 Churn Risk Triage

🎯 Customer Archetype Simulator

🎛️ What-If Scenario Analysis

Marketing Costs (optional)