How many postbacks does SKAN 4 send per install?

Up to three for the winning ad network, one for each conversion window: days 0 to 2, days 3 to 7 and days 8 to 35 after install. The first can carry a fine conversion value, the second and third carry coarse values only. Networks whose ads were eligible but did not win can receive a non-winning postback without conversion data.

How long does a SKAN postback take to arrive?

Each window has to close first, then Apple adds a random delay. The first postback arrives roughly 24 to 48 hours after its window ends, and the second and third arrive 24 to 144 hours after theirs. Locking a conversion value early ends the window sooner and so speeds up the postback.

Can the advertiser receive SKAN postbacks directly?

Yes. Apple sends the winning postback to the ad network, and since iOS 15 the advertiser can receive a copy by declaring an endpoint in the app's Info.plist. MMPs use these copies to decode conversion values and report SKAN results across networks.

How do I know a SKAN postback is real?

Every postback includes an attribution signature created with Apple's private key. The receiver verifies it with Apple's published public key and checks that the signed fields match the payload. A postback that fails verification should be discarded, because the fields can otherwise be forged.

What is a SKAdNetwork Postback? Complete Guide for 2026

Q: What is a SKAdNetwork postback?

A SKAdNetwork postback is an HTTP POST that Apple's servers send to an ad network, and optionally a copy to the advertiser, when an install driven by that network's ad is confirmed on iOS. It contains a campaign source identifier, conversion data and an Apple signature, but no user or device identifiers.

What a SKAdNetwork Postback Is

A SKAdNetwork postback is the report at the end of Apple's privacy-preserving attribution flow. When an ad network's ad leads to an install on iOS and the user opens the app, Apple waits for the measurement window to close, applies a random delay, and sends a signed HTTP POST describing the conversion. The postback is how anyone learns that a SKAN install happened at all, because nothing is reported in real time and no device ID is ever attached.

Unlike a standard postback, which an MMP fires to a network after it attributes an install, a SKAN postback flows the other way: it comes from Apple, and the MMP and network are both receivers. Apple, not the MMP, decides which network wins.

The Fields That Matter

The exact fields depend on the SKAdNetwork version and the privacy tier Apple assigns. In SKAN 4 the payload includes:

version, the SKAdNetwork version the impression was registered with.
ad-network-id, the identifier of the network that registered the ad.
source-identifier, a hierarchical number of two to four digits that the network uses to encode campaign, ad group or creative. Low-volume campaigns only receive the first two digits.
conversion-value and coarse-conversion-value, the fine value from 0 to 63 and the coarse value of low, medium or high. Which one appears depends on the window and the crowd anonymity tier.
postback-sequence-index, 0, 1 or 2 for the first, second and third window.
did-win, true for the winning network and false for networks that receive a non-winning postback.
fidelity-type, 1 for a click or StoreKit-rendered ad and 0 for a view-through impression.
redownload, which flags a reinstall by the same Apple account.
source-app-id or source-domain, the publisher app or website, only at the highest tier.
attribution-signature, Apple's signature over the key fields.

Timers and Windows

SKAN 4 measures three conversion windows after the first app open: days 0 to 2, days 3 to 7 and days 8 to 35. The app can update its conversion values inside each window, and can lock a window early once the value is final.

When a window ends, Apple holds the postback for a random period before sending it: 24 to 48 hours for the first postback and 24 to 144 hours for the second and third. The delay prevents anyone from matching a postback to a specific install by timing. In practice, a campaign's first-window results settle within a few days, while its full 35-day picture takes more than five weeks.

This delay shapes how a team reviews iOS campaigns. Budget decisions made on the last 48 hours of SKAN data are made on incomplete data. A dashboard that shows when each window closes, and how many postbacks are still expected, is more useful than one that shows a single install count.

Winning and Non-winning Postbacks

SKAN 4 introduced non-winning postbacks. When several networks registered eligible impressions for the same install, the winner receives a full postback and up to five other networks with eligible impressions receive a postback with did-win set to false and no conversion values. That tells a network its ad contributed, without giving it credit for the install.

Advertisers receive copies of the winning postback only. That keeps MMP reporting to one credited source per install, consistent with how device-level last-touch attribution works.

Verifying and Decoding Postbacks

Two jobs turn raw postbacks into usable reporting. The first is verification: check the attribution signature with Apple's public key for that SKAdNetwork version, and reject any postback that fails. The second is decoding: translate the source identifier back into campaign names using each network's mapping, and translate fine and coarse values back into events and revenue using your conversion value mapping.

In Linkrunner, postback copies go to linkrunner-skan.com, and the SKAN dashboard shows installs by privacy tier, conversion value distribution, trends over time and a table you can break down by campaign. AdAttributionKit postbacks are received too and verified against Apple's keys, so SKAN and AdAttributionKit installs sit in one view.

Common Reasons Postbacks Look Wrong

Conversion values are null. The campaign did not meet the crowd anonymity threshold, or no event in the mapping fired before the window closed.
Install counts are lower than the network reports. Networks often model SKAN results or include view-through installs. Compare like for like, by window and fidelity type.
Copies are missing. The copy endpoint was not declared in the Info.plist of the shipped build, or the network registered impressions with an older SKAN version that sends fewer fields.
Values never rise above zero. The app does not call the update API on the events you mapped, which is worth testing on a real device before launch.

What is SKAdNetwork Postback? Complete Guide for 2026