How do you know whether a CPI is good?

Only against a benchmark for the same country, OS, and category. An absolute number carries no verdict by itself.

Should you optimize CPA or CPI?

CPI when installs are the goal, CPA when an in-app action is. Their denominators differ, so they should not be mixed.

How do you judge app marketing KPIs?

Unify the definitions first, then compare against country, OS, and category benchmarks.

4 Analysis · 4-1

Core App Marketing KPI Analysis

Unify the definitions of CPI/CPA/ARPU/ROAS and judge campaign performance against country/OS/category benchmarks. Absolute numbers without a benchmark are meaningless.

Measurement basis · MMP attributionBenchmarks · 2026 Q1

Summary

Measure KPIs in 5 layers: (1) efficiency (CPI/CPC), (2) conversion (CPA/CVR), (3) value (ARPU/ARPPU), (4) payback (ROAS), (5) durability (LTV/Retention). Never judge on a single metric — monitor all five together.

§1 Metric definitions (SSOT)

Metric↕	Definition↕	Formula↕	Measurement timing↕
CPI	Cost per install	cost / installs	Per campaign, daily
CPC	Cost per click	cost / clicks	Per campaign, daily
CPA	Cost per action	cost / conversions	Typically D7 basis
CTR	Impression→click rate	clicks / impressions	Network report basis
IPM	Installs per 1000 impressions	(installs / impressions) × 1000	Used on TikTok etc.
ARPU	Avg revenue per user	revenue / installs	At D7 / D30
ARPPU	Avg revenue per paying user	revenue / paying_users	Per cohort
ROAS	Revenue vs ad cost	revenue / cost	D7 / D30 / D90
LTV	User lifetime value (predicted)	cumulative ARPU × retention curve	12-month projection

§2 Industry benchmarks (2026 Q1)

USD basis, Adjust + AppsFlyer aggregate data. Standard deviations across category/country/OS are large — treat as reference only.

Country/OS↕	Category↕	CPI↕	D7 ROAS↕	D7 retention↕
KR iOS	Commerce	$3.20 – $4.80	12% – 25%	22% – 30%
KR iOS	Games	$4.50 – $7.00	8% – 18%	18% – 25%
KR AOS	Commerce	$1.80 – $3.20	10% – 22%	19% – 27%
KR AOS	Games	$2.50 – $4.50	6% – 15%	15% – 22%
JP iOS	Commerce	$4.80 – $7.50	15% – 30%	25% – 33%
JP iOS	Games	$6.00 – $9.50	12% – 22%	22% – 30%
US iOS	Commerce	$4.50 – $6.80	10% – 20%	20% – 28%
US iOS	Games	$5.50 – $9.00	10% – 22%	18% – 26%
US AOS	Commerce	$2.50 – $4.20	8% – 18%	17% – 24%
SEA AOS	Commerce	$0.60 – $1.50	6% – 12%	12% – 20%

§3 Analysis methodology

Step 1: efficiency — confirm CPI/CPC is within ±30% of benchmark. If not, review campaign structure or bidding.
Step 2: quality — within the same campaign, D7 retention should be ≥70% of benchmark. Below that, suspect network traffic quality gaps or a missing deferred deep link.
Step 3: payback — D7 ROAS ≥30% of the break-even LTV/CAC = 1.0 level means learning is on track. At 50%+, bids can go up.
Forbidden pattern: judging campaigns by CPI alone → budget drifts to low-quality-traffic networks. Always co-evaluate with D7 ROAS.

§4 Troubleshooting

Network dashboard and Adjust differ by 30%+

1) Attribution window mismatch (Meta 1d_click vs Adjust 7d_click). 2) SKAN postback delays land iOS installs on D-2~D-3. 3) Missing cost-data sync. Use Adjust numbers as the SSOT for all analysis.

ARPU is good but ROAS is bad

You're catching high-value users but acquisition is too expensive. 1) Lower bids −10%. 2) Try the same LTV group on cheaper networks. 3) Narrow the Lookalike seed for tighter targeting.

Loading…

Related resources

Core App Marketing KPI Analysis

Unify the definitions of CPI/CPA/ARPU/ROAS and judge campaign performance against country/OS/category benchmarks. Absolute numbers without a benchmark are meaningless.

Measurement basis · MMP attributionBenchmarks · 2026 Q1

Summary

§1 Metric definitions (SSOT)

Metric↕	Definition↕	Formula↕	Measurement timing↕
CPI	Cost per install	cost / installs	Per campaign, daily
CPC	Cost per click	cost / clicks	Per campaign, daily
CPA	Cost per action	cost / conversions	Typically D7 basis
CTR	Impression→click rate	clicks / impressions	Network report basis
IPM	Installs per 1000 impressions	(installs / impressions) × 1000	Used on TikTok etc.
ARPU	Avg revenue per user	revenue / installs	At D7 / D30
ARPPU	Avg revenue per paying user	revenue / paying_users	Per cohort
ROAS	Revenue vs ad cost	revenue / cost	D7 / D30 / D90
LTV	User lifetime value (predicted)	cumulative ARPU × retention curve	12-month projection

§2 Industry benchmarks (2026 Q1)

USD basis, Adjust + AppsFlyer aggregate data. Standard deviations across category/country/OS are large — treat as reference only.

Country/OS↕	Category↕	CPI↕	D7 ROAS↕	D7 retention↕
KR iOS	Commerce	$3.20 – $4.80	12% – 25%	22% – 30%
KR iOS	Games	$4.50 – $7.00	8% – 18%	18% – 25%
KR AOS	Commerce	$1.80 – $3.20	10% – 22%	19% – 27%
KR AOS	Games	$2.50 – $4.50	6% – 15%	15% – 22%
JP iOS	Commerce	$4.80 – $7.50	15% – 30%	25% – 33%
JP iOS	Games	$6.00 – $9.50	12% – 22%	22% – 30%
US iOS	Commerce	$4.50 – $6.80	10% – 20%	20% – 28%
US iOS	Games	$5.50 – $9.00	10% – 22%	18% – 26%
US AOS	Commerce	$2.50 – $4.20	8% – 18%	17% – 24%
SEA AOS	Commerce	$0.60 – $1.50	6% – 12%	12% – 20%

§3 Analysis methodology

Step 1: efficiency — confirm CPI/CPC is within ±30% of benchmark. If not, review campaign structure or bidding.
Step 2: quality — within the same campaign, D7 retention should be ≥70% of benchmark. Below that, suspect network traffic quality gaps or a missing deferred deep link.
Step 3: payback — D7 ROAS ≥30% of the break-even LTV/CAC = 1.0 level means learning is on track. At 50%+, bids can go up.
Forbidden pattern: judging campaigns by CPI alone → budget drifts to low-quality-traffic networks. Always co-evaluate with D7 ROAS.

§4 Troubleshooting

Network dashboard and Adjust differ by 30%+

1) Attribution window mismatch (Meta 1d_click vs Adjust 7d_click). 2) SKAN postback delays land iOS installs on D-2~D-3. 3) Missing cost-data sync. Use Adjust numbers as the SSOT for all analysis.

ARPU is good but ROAS is bad

You're catching high-value users but acquisition is too expensive. 1) Lower bids −10%. 2) Try the same LTV group on cheaper networks. 3) Narrow the Lookalike seed for tighter targeting.