Teardown: A SKAdNetwork Postback Log for a Mobile App Launch
Objective: Given a synthetic-realistic SKAdNetwork postback summary across 12 iOS ad sets, identify which campaigns are producing unusable data and which conversion-value mapping decisions are actively hiding the app's most valuable users.
You're auditing Robinhood's iOS UA campaign structure after the mobile team flags that 'SKAN data looks useless this month, we can't tell which ad sets are working.'
Read the postback summary, separate genuine SKAN limitations from fixable schema and structure mistakes, and write up what to change before next month's spend.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, sufficient for a monthly-sized postback export
Paid upgrades (optional, faster/deeper)
No paid MMP dashboard access is required to complete this teardown
The process
Specimens to review
Which ad sets in this postback summary are structurally incapable of returning usable conversion-value data, and what single change fixes most of them at once?
Robinhood iOS UA, SKAN 4 postback summary, last 30 days Ad Set Daily Installs Postbacks Received Null Conversion Value Rate Brokerage-US-Broad 310 298 6% Brokerage-US-Lookalike 22 19 84% Crypto-US-Interest 14 11 91% Crypto-UK-Broad 95 88 9% Options-US-Retarget 8 6 100% Options-US-Broad 180 171 11% Savings-US-Lookalike 17 14 88% Savings-US-Broad 240 229 7% Card-US-Interest 19 16 90% Card-US-Broad 150 143 10% Brokerage-CA-Broad 12 10 93% Crypto-CA-Broad 9 7 100%
Specimen: synthetic, realistic
Review this conversion value schema against the lesson's worked example. What is wrong with how the top value band is defined?
Robinhood SKAN conversion value schema (as configured in AppsFlyer) Value 0-15: App opened, no action Value 16-31: Completed KYC signup Value 32-47: Funded account (any amount) Value 48-55: Placed first trade Value 56-63: Funded account OR placed first trade (same range, either action)
Specimen: synthetic, realistic
Final deliverable
A written teardown flagging every low-volume ad set, a consolidation recommendation, and a corrected conversion value schema with one action per value band.
See a reference example
Nubank iOS UA, SKAN teardown findings (excerpt) STRUCTURAL: 5 of 9 ad sets under 20 daily installs, consolidate Interest + Lookalike splits per product into one Broad ad set each SCHEMA: Values 50-63 currently cover both 'card activated' and 'first transfer' as one band, split into two dedicated bands so postbacks stop being ambiguous
Success criteria
You're done when you can:
- Correctly identifies all 6 sub-threshold ad sets and their null conversion value rates
- Recommends consolidation as the fix, not simply 'increase budget'
- Identifies the conversion value schema's ambiguous top band as a separate, second defect