Build or buy pLTV

Cohort pLTV belongs in-house. User-level pLTV, fast enough to bid on, is what Churney does.

Build the cohort model yourself and buy the bidding signal. A cohort model predicts what a group of customers, such as one week of sign-ups from one campaign, will be worth. It drives budgets, payback and forecasts, and your data team knows the business best, so it belongs in-house and Churney encourages it. A user-level model predicts each new customer’s value within hours, ranked and calibrated well enough for Meta, Google and TikTok to bid on. That is what Churney builds and runs.

Written by Churney, so read it with that in mind. Meta facts come from Meta’s developer docs, checked on 15 November 2026.

Side by side

Cohort pLTV, in-house
User-level pLTV, Churney
What it predicts
Cohort pLTV, in-houseThe value of a group of customers (a campaign, a week, a country) at day 90, 180 or 365.
User-level pLTV, ChurneyThe value of each new customer, within hours of the first session or conversion.
What it is for
Cohort pLTV, in-houseBudgets, payback, forecasts and the finance view of acquisition.
User-level pLTV, ChurneyThe value Meta, Google and TikTok bid on.
Graded on
Cohort pLTV, in-houseError on the cohort total.
User-level pLTV, ChurneyHow well it ranks customers early, and how well the value sent is calibrated.
How fast it must be
Cohort pLTV, in-houseA daily or weekly run is fine.
User-level pLTV, ChurneyInside the platform’s window. Meta takes a delayed value up to 7 days after the conversion (Meta’s docs).
Where it goes
Cohort pLTV, in-houseDashboards, sheets and planning tools.
User-level pLTV, ChurneyMeta Conversions API (AppendValue), Google Ads and TikTok, directly or through AppsFlyer or Adjust.
Who runs it
Cohort pLTV, in-houseYour data team.
User-level pLTV, ChurneyChurney: monitoring, drift, retraining and match rates.
How it is proven
Cohort pLTV, in-houseBacktests against cohorts that have matured.
User-level pLTV, ChurneyA split test against your current campaigns, success agreed before launch, read on your own attribution.
What it needs
Cohort pLTV, in-houseCohort revenue history.
User-level pLTV, ChurneyUser-level revenue and events in a warehouse, read-only. No SDK.

Which fits

Build in-house if

  • You need forecasts, payback and budget allocation. Always build this one yourself.
  • You spend less than $300k a month on paid acquisition, where a bidding signal rarely pays for itself.
  • You have engineers to own a production bidding signal: per-platform calibration, AppendValue integration, match-rate and drift monitoring, and a test to prove it.

Buy Churney for bidding if

  • You spend $300k or more a month and customer value shows up after the first conversion: renewals, trial-to-paid, repeat orders.
  • You want a signal that meets Meta’s pLTV rules: at least 100 conversions a week for 4 weeks, 5 or more distinct values, the highest at least 3x the lowest.
  • The result has to hold up with finance: a split test against business as usual, read on your own numbers.

What building the bidding model takes

  1. Features available within hours of a customer’s first session, not after the first week.
  2. Calibration per platform, so the value Meta sees and the value Google sees mean what they should.
  3. The delivery: Conversions API with AppendValue, Google Ads conversion values, TikTok events, or the MMP for apps.
  4. Monitoring for match rates, drift and changes in your product or pricing, with retraining.
  5. A test design that proves the uplift before you roll it out.

For the cohort model, start from the cohort-pLTV model recommendations in the pLTV Audit.

FAQ

Should we build pLTV in-house?

Build the cohort model in-house: it is the planning model and your team knows the business best. Buy or build the user-level bidding signal depending on your spend and engineering capacity.

Can one pLTV model do budgets and bidding?

No. A budget forecast is graded on the cohort total; a bidding signal is graded on ranking customers within hours. They need different models. See two models, never one.

Does Churney replace our data team’s model?

No. Churney runs the user-level bidding signal. Your cohort model keeps doing budgets and forecasts, and the two can be checked against each other.

What does the pLTV Audit give our data team?

After the three checks, the pLTV Audit gives cohort-pLTV model recommendations: the cohort model to try first, with settings filled in from what the checks measured. Free at tools.churney.io.

How long until a Churney signal is live?

Week 8 of a four-month pilot, after data checks and a written test design. The result is read out in week 17.

See whether your data predicts long-term value

A 30-minute call. We look at your setup and tell you plainly whether pLTV is worth testing.

Book a call

Sources: Meta, pLTV integration guidance (updated 9 July 2026); how the pLTV Audit is scored; cohort vs user-level pLTV. Comparing vendors? See pLTV vendors compared.

Is Churney right for you?

Three free, read-only checks show whether your campaigns, revenue and warehouse are ready. About ten minutes.

Run the free pLTV Audit