# LTV Scoring > Predict customer lifetime value at the moment of conversion. Optimize ad spend for long-term revenue — not just sign-ups or first purchases. Predict the lifetime value of every customer at the moment they convert. Optimize your ad spend for high-LTV acquisition — not just sign-ups or one-time purchases. *7 min read · March 2026* ## 01 — Not All Customers Are Worth the Same Every subscription-focused business knows this: not all customers are created equal. Typically, 20% of customers generate 80% of revenue. The pattern is the same everywhere: - **SaaS** — one subscriber stays 3 years and generates $4,800; another cancels after month one - **E-commerce** — a $30 first order could be a one-time discount hunter or a customer who'll spend $3,000/year - **Gaming** — a free install could be a casual player or a whale spending $500/month - **Subscriptions** — same $49/month signup, wildly different retention and upgrade paths But in your acquisition reports, every customer looks the same. A signup is "one conversion." A first purchase is "one conversion." CPA treats them as equal. Your reports can't tell the difference — because lifetime value takes 6 to 12 months to materialize. Budget decisions based on CPA or first-purchase revenue systematically overinvest in channels that produce cheap, low-value customers, while underfunding the channels that quietly acquire your most valuable ones. > **[Illustration: CPA vs LTV]** > CPA says Google wins. 12-month customer value says Meta wins by 3x. The metric you optimize against determines where your budget goes. --- ## 02 — Ad Platforms Optimize for What You Tell Them LTV matters for two reasons: **measurement** — identifying which channels acquire your highest-value customers — and **optimization** — ad platforms optimize toward whatever data you send them. Without LTV signals, they chase volume. When you run Google or Meta on **Maximize Conversions** or **Target CPA**, the algorithm has no way to distinguish a $4,800 lifetime customer from a free-trial churner. Every conversion is "one conversion" — equal weight, equal signal. > **[Illustration: LTV Flat Bidding]** > Three customers with wildly different lifetime value — all reduced to "1 conversion" by the algorithm. Sending real LTV values would fix this — but ad platforms need the value signal at the moment of conversion. Actual lifetime value takes 6 to 12 months to unfold. By then, the algorithm's learning window has long closed. > Predictive LTV scoring bridges this gap: it forecasts each customer's lifetime value at the moment of conversion — giving both your reports and your ad platforms the signal they need, months before revenue materializes. --- ## 03 — Predict Lifetime Value Before It Materializes SegmentStream scores every new customer with ML models that predict 6–12 month revenue at the moment of first purchase or sign-up. The models learn from historical cohort behavior — which customer profiles retain, which upgrade, which churn — and apply those patterns to new customers instantly. > **[Illustration: LTV Scoring Hero]** > Each customer scored with CAC, predicted 12-month LTV, and LTV:CAC ratio — at the moment of conversion. The models use signals available at the moment of conversion: - **Product & plan** — first product purchased, subscription tier, billing frequency - **Transaction data** — order value, composition, payment method - **Acquisition context** — channel, campaign, landing page - **Behavioral signals** — device, session depth, engagement patterns - **Customer profile** — demographics, company size, zero-party data An Enterprise customer with annual billing and a corporate email gets a predicted LTV of $4,800. A free-trial user on a personal email with minimal engagement scores $45. Both signed up today. The pLTV score tells you which acquisition channel to scale. This is not rule-based scoring. The ML models learn from actual customer outcomes — which behavioral patterns predicted long-term revenue, and which looked promising but led to early churn. > **[Interactive: LTV Cowork Demo]** > Ask which channels drive high-LTV customers — get instant analysis with LTV:CAC ratios and the reasons behind each channel's performance. --- ## 04 — How It Works ### 1. Connect your transaction data The models train on customers with known outcomes — subscription history, repeat purchases, upgrades, churn dates, and total revenue over 6–12 months. This labeled data comes from your billing system, CRM, or data warehouse via [CRM Funnel Attribution](/measurement-engine/crm-funnel-attribution). > **[Illustration: LTV Data Connect]** > Connect Stripe, Shopify, Salesforce, BigQuery, or any billing system. User and payments data flow into the scoring pipeline. ### 2. Score every new customer at conversion SegmentStream learns which acquisition channels, plans, and behaviors predict the steepest revenue curves. Every new sign-up gets a predicted LTV within hours — two models run in parallel: one predicts retention probability, the other predicts monthly revenue. > **[Illustration: LTV Scoring Pipeline]** > Retention model (82% probability of staying 6 months) × Revenue model ($475/mo) × 6 months = $2,340 predicted LTV. ### 3. Measure true acquisition ROI With predicted lifetime values, each customer carries a dollar amount attributed to the campaign that acquired them. Compare [cross-channel performance](/measurement-engine/cross-channel-attribution) on LTV-based ROAS — not just CPA and sign-up volume. ### 4. Send pLTV values to ad platforms The predicted LTV values are sent to Google, Meta, and TikTok via Conversions API as weighted conversion value signals. Instead of every purchase sending "1 conversion" worth the first order amount, each conversion carries its full predicted lifetime value — enabling the platform to train its algorithm on customer quality, not just conversion volume. > **[Illustration: LTV Feedback Loop]** > Predicted LTV values sent to each platform via CAPI. The algorithm trains on real customer value, not just conversion counts. --- ## 05 — What Changes ### Evaluate campaigns on customer value, not sign-up volume Stop optimizing for the cheapest sign-ups. Predicted LTV lets you compare campaign ROI within days of launch. A campaign that acquired 50 customers at $28 CPA with $1,400 average pLTV is immediately visible as a better investment than one that acquired 500 customers at $8 CPA with $45 average pLTV. > **[Illustration: ROAS vs LTV ROAS]** > Campaign C has the lowest first-purchase ROAS but the highest LTV ROAS. Rankings flip when you measure lifetime value, not initial transactions. ### Unlock value-based bidding Without LTV values, you're limited to **Maximize Conversions** and **Target CPA** — strategies that treat every sign-up equally. The algorithm finds more conversions at the lowest cost, regardless of downstream value. With predicted LTV, you switch to **Maximize Conversion Value** and **Target ROAS**. The algorithm bids higher for audiences that produce $4,800 subscribers and lower for those that produce $45 trial users. Same budget, directed toward the highest-value customers. > **[Illustration: LTV Bidding Shift]** > Bids scale with predicted LTV. The $4,800 loyal subscriber gets an $85 bid. The $45 one-time buyer gets $2. > The shift from "Maximize Conversions" to "Maximize Conversion Value" changes what ad platforms optimize for. Instead of finding people who sign up, they find people who stay, upgrade, and generate revenue for months and years.