TerradsTerrads
← Glossary

Predicted lifetime value

  • Also called pLTV
  • Updated November 11, 2026

Predicted lifetime value (pLTV) is a forecast, from your own model, of how much a newly acquired customer will be worth over time. You send it to Meta through the Conversions API within 7 days of the conversion, and the Maximize value of conversions performance goal then bids toward the people you expect to be worth the most, instead of judging them by their first purchase.

How does Meta define pLTV?

Meta's Set up maximize value of conversions for predicted lifetime value page:

pLTV value optimization lets you acquire high-value customers based on your own lifetime value predictions. You make these predictions early in the acquisition funnel and send them to Meta, which optimizes delivery toward the users you expect to be most valuable over time.

It is one of the conversion value types on About maximizing the value of conversions, described as "a customer's expected long-term worth, sent up to 7 days after the conversion event." The value hangs on one conversion event, the moment you acquire the user, which "happens only once per customer". And Meta notes: "You can only associate one conversion event with pLTV per dataset (pixel)."

When does pLTV make sense?

When early behavior is a poor guide to later value. Meta: "pLTV is relevant when the correlation between short-term and longer-term purchase behaviors is low". Its two examples are a subscription with a free trial, where a user who hasn't paid by day 7 may become a high-value customer, and ecommerce, where "a low-value purchaser from D0–D7 who becomes a high-value purchaser by D365" would look weak to ordinary value optimization.

What are the requirements for pLTV?

From the setup page:

RequirementWhat Meta says
Data connection"an existing web Conversions API integration"; "pLTV does not support app events at this time"
Model"validated in some way, such as one already used for optimization, measurement, or budget decisions"
Values"at least 5 distinct pLTV values. All values must be positive (not binary or negative), and the highest value should be at least 3x greater than the lowest"
Volume"a minimum of 100 conversion events per week attributed to Meta for each of the last 4 weeks"
Event"Your conversion event must send a value and currency."

The about page is looser on values, asking for "at least 5 unique values, all of which must be non-negative", which would allow a zero. Follow the setup page's stricter wording. Access can lag: "It may take a few weeks after meeting the requirements before pLTV becomes available for your account", and after setup Meta says to "allow roughly 2 weeks for Meta's models to train".

Real-time or delayed pLTV?

Meta offers two ways to send the value, in its developer guide and on the setup page:

  • Real-time: the predicted_ltv value goes with the conversion event itself.
  • Delayed: you send the conversion as usual, then an AppendValue event "within 7 days". The clock starts when you send the conversion, so "if you calculate pLTV on day 5, you have at most 2 days to send the data back to Meta."

Faster isn't better here. The developer guide says "sending pLTV values faster does not currently improve performance", and both pages warn not to hold back the conversion event while you wait for the prediction.

How do you read pLTV results?

Add the pLTV column for your event, then read Results value and Results ROAS. Meta adds a warning: "Return on ad spend (ROAS) is comparable only across pLTV campaigns." A pLTV campaign's ROAS can't be set against a purchase-value campaign's. With delayed sending, wait for the campaign to end plus your delay before reading final numbers.

What it means for your ads

  • Plan a test of at least 3 weeks with no ROAS goal. Meta says peak performance "is typically reached after 2 weeks" and recommends "excluding the first week of the campaign from reporting."
  • Budget for the volume. Meta asks for enough to cover "at least 100 conversion events attributed to Meta per week for 3 weeks."
  • Start with 1-day or 7-day click attribution, as Meta recommends for a first test.
  • With no model yet, a customer value column on a customer list can still steer a value-based lookalike toward your best customers.

Sources

Meta revises these pages without notice, so check the source before changing an integration.

Terrads makes a batch of different ads in one go, so a three-week value test has fresh creative ready when the first ads tire.