The mistake is not using new retargeting measurement. The mistake is treating every reported return as if it were the same business event.

AppsFlyer made re-engagement and re-attribution measurement available for extended platforms—CTV, PC and console—effective August 1, 2026. That opens a useful lifecycle view for teams whose customers do not live on one mobile device. It does not make the resulting numbers interchangeable with new installs, or automatically comparable with an existing LTV model.

A classified return is a measurement input. It becomes a budget signal only after its cohort and decision boundary are explicit.

Three conversion types, three questions

Re-engagement credits activity that brings an already-installed user back. Re-attribution credits a lapsed user who reinstalls and reopens under the defined eligibility rules. A new install is a separate event again. Those distinctions matter because each can carry different windows, paths and economic meaning.

CLASSIFICATIONWHAT IT OBSERVESWHAT IT DOES NOT YET PROVE
RE-ENGAGEMENTAn existing user returned during the configured window.That the campaign caused the return or created incremental future value.
RE-ATTRIBUTIONA lapsed user reinstalled and reopened under the stated rule.That the reinstall should be valued as new acquisition.
NEW INSTALLAn install met the platform’s attribution rule.That it is comparable with lifecycle or cross-device outcomes.

The useful question is therefore not “Did retargeting work?” It is “Which event happened, how was it classified, and which decision is safe to make from that classification?”

Why configuration belongs in the measurement contract

AppsFlyer’s release describes three configuration layers: the app, the media-source integration and the attribution link for non-self-reporting networks. Treat these as part of the analytical evidence. If a configuration is not documented, a difference in counts could be a marketing outcome, a reporting-rule outcome or a setup outcome.

Support is also not uniform. AppsFlyer notes that Roku supports re-engagement but not re-attribution, and that cross-platform retargeting into mobile is out of scope. That is not a defect to smooth over in a total. It is a boundary to retain in the decision record.

A six-gate release for extended-platform results

  1. 01

    Name the conversion

    Classify the result as re-engagement, re-attribution or a new install before it appears in a performance summary. These are different events, not alternative labels for the same outcome.

  2. 02

    Prove the setup

    Record the app-level setting, the media-source setting and, where relevant, the attribution-link configuration. A campaign cannot be reconciled from the dashboard alone when the configuration path is unknown.

  3. 03

    Keep the windows visible

    Write the re-engagement and inactivity windows alongside the result. A return inside a 30-day window and a lapse that qualifies for re-attribution should not silently inherit an acquisition cohort definition.

  4. 04

    Record network limits

    Ask what the network can actually report. For example, AppsFlyer documents that Roku supports re-engagement but not re-attribution. A combined total must retain that boundary.

  5. 05

    Choose the reporting surface

    Confirm where the classification is available and which export can preserve it. A metric visible in a dashboard is not automatically available in the raw-data view a finance or data team uses to reconcile LTV.

  6. 06

    Release one bounded decision

    Only after the prior gates agree should the team decide whether to optimise the campaign, run a reconciliation, or design an incrementality test. Attribution classification itself does not prove causal value.

Use a separate release state before merging LTV

STATEWHAT IS KNOWNALLOWED ACTION
CONFIGURATION HOLDThe event class, windows or configuration evidence is incomplete.Do not merge into combined LTV or payback; repair the measurement record.
RECONCILIATION READYThe classification, network limitations and reporting surface are known.Reconcile cohort treatment with product and finance records before a spend conclusion.
DECISION READYThe intended KPI boundary is explicit and the required reconciliation passed.Run a scoped optimisation or incrementality test with the unresolved causal claim still visible.

Source and limits

  • AppsFlyer: retargeting support for CTV, PC and console platforms, updated July 28, 2026. Used for feature scope, effective date, configuration layers, platform constraints and reporting locations.
  • The source documents attribution and reporting behaviour for one platform. It does not establish campaign incrementality, complete cross-channel reconciliation or the correct LTV policy for every business.