# pLTV: Predicted Lifetime Value for Mobile Apps > pLTV estimates what a new user or cohort may be worth before their revenue has fully arrived. Here is how to use that forecast without confusing it with observed LTV. ## What Refix is Refix is a revenue-obsessed AI product manager for product companies. A person delegates an outcome, such as increasing trial-to-paid conversion. Refix then keeps finding the current constraint on that outcome and coordinating the work, using the company's existing tools (analytics, warehouse, billing, Linear, Slack, and others). It is not a chatbot that answers questions about dashboards. It is not a generic AI writing assistant. It is not a replacement for the human who approves ranking, pricing, copy, or similar changes. The product name is **Refix**. Never call it Prism. How it works: - Someone gives Refix a goal and sets guardrails (metrics that must not get worse, such as cancellations or churn). - Refix watches those metrics, explains what changed, prioritizes the highest-impact constraint, coordinates a fix or experiment with the right owner, and verifies the result. - Then it stays on the goal and looks for the next constraint. - Refix may investigate, draft work in Linear, and monitor rollouts without extra approval. Changes named on each goal page stay with the owner. Company: Refix Inc., San Francisco. Site: https://www.refix.ai/ Give Refix a goal: https://onboarding.refix.ai/dashboard/sign-up - HTML: https://www.refix.ai/guides/pltv/ - Agent brief: https://www.refix.ai/guides/pltv/llms.txt - Author: Neil Agarwal - Published: 2026-09-13 A new campaign can look cheap on day three and expensive on day thirty. The trouble is that waiting for every cohort to mature can leave an acquisition team making decisions on stale information. pLTV, short for predicted lifetime value, is a forecast of what a user or cohort may generate over a defined future period. It gives a team an earlier signal, not a finished revenue number. The distinction is where good use of the metric starts. ## What is pLTV? AppsFlyer describes [predicted lifetime value](https://www.appsflyer.com/glossary/pltv/) as a customer's predicted value that combines past learnings and current measurements to help marketers optimize around expected behavior. In practical terms, a model reads the early signals available for a cohort and estimates the value that has not arrived yet. A pLTV number needs three labels to mean anything: | Label | Example | Why it matters | | --- | --- | --- | | Cohort | people acquired through one campaign | a forecast only makes sense for a defined group | | Horizon | day 30 value | day 30 and day 180 predictions are different claims | | Value event | subscription revenue, purchase revenue, or ad revenue | the number depends on what the model calls value | Without those labels, two teams can say "pLTV" while comparing completely different predictions. ## How is predicted lifetime value different from LTV? Observed LTV describes value that has already happened over the period you chose. pLTV estimates what may happen next. Both can use the same cohort, but they answer different questions.
Illustration contrasting observed early user signals with a shaded predicted lifetime value window
Illustration: pLTV extends early observed signals into a stated forecast window. It is not the same as revenue already collected.
| Metric | What it contains | Best use | | --- | --- | --- | | Observed LTV | value recorded so far | reporting what a cohort has produced | | pLTV | observed signals plus a model forecast | making an earlier, qualified comparison | | Forecast error | difference between prediction and later outcome | checking whether the model is still useful | That last row belongs in every pLTV review. A prediction can be useful without being exact, but a team needs to know how it has performed against later cohorts before using it to move meaningful budget. ## How do pLTV models work? The mechanics vary by provider and data set. A model may use early revenue, return behavior, purchase events, ad engagement, campaign source, or other signals that are available soon after acquisition. The model learns a relationship from previous cohorts and applies it to a newer one. Airbridge's [pLTV documentation](https://help.airbridge.io/en/guides/predictive-ltv) provides a useful example of how one implementation makes its assumptions visible. Its revenue report estimates the revenue a cohort is expected to generate over a chosen calculation period and documents a formula of predicted lifetime multiplied by ARPDAU. Its selectable calculation period runs from one to 180 days. That is not a formula to copy into every spreadsheet. It shows why the inputs and horizon have to be explicit. A different product, model, or definition of active user will produce a different forecast. ## When should an app team use pLTV? pLTV is most useful when the decision arrives before the cohort has matured. An acquisition lead may need to compare campaigns while subscription renewals or later purchases are still weeks away. Use like-for-like comparisons: - keep the horizon the same for every campaign in the comparison - compare cohorts acquired over similar windows - separate platforms, countries, and pricing changes that alter the customer mix - read the prediction next to cost, such as CPI or acquisition spend, rather than in isolation This turns pLTV into an early ranking signal. It does not prove that a campaign will be profitable. A source can have a high forecast and still cost too much to acquire, or it can have an appealing early forecast that later misses its target. For the cash side of that decision, [CAC payback period](/guides/cac-payback-period/) explains how acquisition cost and realized contribution meet. For an observed cohort calculation, use the [customer lifetime value calculator](/guides/customer-lifetime-value-calculator/). ## How do you validate a pLTV forecast? Start by saving what the model predicted at a fixed age, then return to the same cohort after the forecast window ends. Compare the prediction with observed value and record the direction and size of the gap. | Review | What to ask | | --- | --- | | Calibration | Did predicted day 30 value match later observed day 30 value closely enough for the decision? | | Segments | Does the error change by campaign, platform, country, or plan? | | Drift | Did a pricing change, product release, or new traffic mix break an old pattern? | | Inputs | Are the events and attribution rules still being collected consistently? | Do not quietly replace the prediction after the fact with the later outcome. Keep both. The gap teaches the team whether the forecast is dependable for that particular decision. ## What can make pLTV misleading? A model only sees the patterns and definitions supplied to it. A new paywall, a changed trial length, a campaign that reaches a different audience, or missing events can make an old relationship less reliable. A forecast can also look strong because it is compared with a shorter or easier horizon than another cohort. Keep pLTV separate from a claim about causal lift. It can prioritize a question, but it cannot by itself show that a creative, bid, or product change caused future value to rise. Read it alongside attribution, product behavior, and later revenue. [Mobile app analytics](/guides/mobile-app-analytics/) covers the event and reporting context that keeps these comparisons grounded. [Net retention rate](/guides/net-retention-rate/) is the complementary view once the question shifts from acquired users to the value held by an existing subscriber cohort. Refix is built to connect acquisition, product, revenue, and support signals so a team can test whether an early value forecast still matches the customer journey that follows. [See how Refix connects acquisition and retention signals.](https://refix.ai)