Video summary

SKAdNetwork 4.0: Understand Apple's SKAN Update For Advertisers

Main summary

Key takeaways

Technology

Webinar summary: Apple SKAdNetwork 4.0 updates (SKAN 4.0)

Speakers / panel goals

A panel of measurement and marketing-technology experts discusses Apple’s SKAdNetwork 4.0 changes introduced at WWDC. The discussion focuses on remaining challenges after iOS 14.5/ATT, and how SKAN 4.0 features impact:

  • LTV measurement
  • source breakdown
  • web-to-app attribution

The session also includes Q&A covering implementation details and practical workarounds.

1) Ongoing SKAN challenges (about 1 year after iOS 14.5)

Key recurring difficulties across advertisers:

  • Limited “source of truth” for lower-funnel outcomes (LTV / revenue / purchases)

    • Teams have had to rethink measurement because user-level attribution is not available.
    • As a result, teams rely on aggregated SKAN signals plus internal data.

  • Measurement scalability across apps/portfolios

    • Even after extensive testing, advertisers still struggle to build an approach that works consistently across multiple apps and aligns well with SKAN data.

  • ATT and data latency (laundering windows)

    • Waiting for postbacks after the laundering window slows optimization cycles and limits how quickly teams can learn.

  • Conversion value strategy and the “short windows” problem

    • Platform guidance (e.g., Facebook/Google) often uses ~24–48 hour windows.
    • Panelists argue these windows are too short to reliably predict longer-term outcomes such as day-360 ROI.

  • Null / missing conversion values

    • “Null value” behavior is a major pain point.
    • It reduces the usefulness of SKAN signals and increases uncertainty in optimization.

2) What SKAN 4.0 changes (main technical updates)

A) Improved LTV measurement via multiple postbacks (postback logic)

SKAN 4.0 introduces three postbacks (instead of fewer in earlier versions), improving the ability to infer LTV over time. Implementation/behavior notes emphasized:

  • Only the first postback is guaranteed.
  • Subsequent postbacks may be sent depending on conditions, including:
    • whether a user is active in certain windows
    • whether conversion value updates occur
  • Conversion value granularity differs by postback
    • The 0–63 conversion value (fine-grain) appears only in the first postback.
    • Later postbacks use coarse-grain buckets (low / medium / high) and/or may involve null values, depending on thresholds/eligibility.

Practical example provided: revenue can be bucketed into low/medium/high for day 2–7 and 7–35 measurement windows. This is less granular than fine-grain, but considered more useful than prior options.

B) Source ID: extending campaign breakdown from 2 digits to 4 digits

SKAN 4.0 changes the identifier used for campaign structure:

  • “Campaign ID” becomes “Source ID”
  • Expanded from 2 digits to 4 digits

Intended benefit:

  • more campaign granularity, enabling better slicing (e.g., geo/creative/campaign-like cohorts), subject to privacy-related guarantees.

Panelist note:

  • not all digits are guaranteed under privacy thresholds, but advertisers expect improved cohort visibility overall.

C) Web-to-App attribution support (web traffic now measurable)

A new SKAN 4.0 feature:

  • Web-to-app becomes supported (previously not supported in SKAN)

Why it matters:

  • particularly relevant for Google search-driven web traffic

What it enables:

  • measuring outcomes like click from web → install using the same “one link” mechanism.

SDK considerations:

  • while the flow is intended to be mostly “out of the box” for web-to-app, an SDK update for SKAN 4 is likely required.

3) How these updates affect advertisers (use cases)

  • LTV/subscription apps benefit from longer measurement windows

    • Day-0/1 postbacks or 24-hour measurement can miss subscriptions that happen later (e.g., around day 30).
    • SKAN 4.0’s LTV windows are positioned as more suitable for delayed monetization.

  • Reducing null-rate concerns

    • Panelists expect more data usefulness because values that previously became null may now map into coarse buckets (low/medium/high), depending on Apple’s logic.

  • Web-to-app helps bridge attribution gaps

    • Match Group notes web-to-app helps retain audience/creative learning from web flows and can feed into scan campaigns.
    • It’s characterized as a lower-funnel signal for ad networks compared to install-only signals.

4) Q&A highlights (implementation / interpretation questions)

  • Null rate and new conversion buckets

    • Previously “not eligible” conversion values could become null.
    • SKAN 4.0 introduces fine vs coarse conversion value types, potentially reducing null outcomes by mapping more results into low/medium/high (exact behavior depends on Apple thresholds).

  • Threshold setting for low/medium/high

    • Advertisers don’t fully control Apple’s privacy thresholds.
    • However, they can configure their own bucketing via conversion studio/modeling logic.

  • Web-to-app applicability by browser

    • Uncertainty acknowledged; documentation/testing was unclear.
    • Expectation: support is not limited to Safari.

  • Source ID assignment responsibility

    • Panelists generally prefer ad networks assign the 4 digits,
    • but only if it can map back to creative/asset-level so advertisers can optimize meaningfully.

  • Mixed media modeling discussion

    • Mixed media measurement is challenging because some key signals (e.g., offline influence such as YouTube or influencer push) aren’t always available the same way.

  • Opt-out users / revenue estimation

    • SKAN measures differently for opt-out/consented users.
    • Panelists describe enriching revenue/LTV estimates using consented data, probabilistic approaches, and additional signals.

  • Geography measurement without Geo in SKAN

    • Workarounds discussed:
      • use impression data by country from cost data combined with install data to approximate geographic contribution
      • assign campaigns/slots to geos for testing when targeting is region-limited

  • Timing of release

    • SKAN 4 is expected later in the year, but no exact Apple date was provided in the discussion.

5) “Wish list” for future SKAN versions (5/6)

Recurring desires:

  • More advertiser endpoint / better visibility

    • Send organic + attributed postbacks to the advertiser endpoint for better benchmarking and behavior comparison.

  • Improved guidance and partnership / more advertiser solutions

    • More clarity on how to use SKAN features correctly and effectively.

  • More publisher environment transparency (privacy-safe)

    • A less blunt way to identify “source app ID / publisher environment” while preserving privacy.

  • Revenue visibility improvements at useful granularity

    • One panelist wishes for revenue at campaign level (or richer KPIs), rather than only at higher aggregation.

  • More guidance on cohorting and modeling changes

    • Specifically, how to use the extra Source ID digits for meaningful cohorts and optimization.

Main speakers / sources (as mentioned in the subtitles)

  • Moderator: Minnie Katzenmaier (AppsFlyer)
  • Roy — Director of Product, AppsFlyer (iOS solutions / SKAN innovation)
  • Margaret Ochoa — Digital Marketing & Measurement Lead, Match Group (portfolio including Match, Tinder, Hinge)
  • Caroline Austin — Paid Acquisition Lead, Match Group (Affinity app segment)
  • Emily Tierney — Marketing Technology Lead, Supercell
  • Matt Allenwood — Director of Product Management, Liftoff (Vungle direct ads, data products, privacy initiatives)

Original video