Why it matters
Install happens at the top of the funnel. Revenue often arrives later through trials, subscriptions, in-app purchases, or ads viewed over months. Campaigns optimized purely on CPI can flood the app with users who never monetize, churn during trial, or refund early purchases.
CPI also interacts with measurement loss. App Tracking Transparency (ATT), SKAdNetwork delays, and MMP postback gaps make install attribution noisy. Teams compare CPI across networks while match rates and incrementality differ. A lower CPI on one channel is not better if trial-to-paid and LTV underperform.
For finance, CPI must connect to payback period and margin LTV. A Head of Performance may hit CPI caps while cohort economics show the install cohort never recovers CAC. That gap is why mobile UA teams graduate from CPI targets to value optimization, tROAS-style goals, or user-level pLTV sent server-side to ad networks.
Cost per install (CPI)
CPI measures cost at install; pLTV estimates value before monetization matures:
- Inputs: Install events, early in-app behavior, trial starts, purchases, and churn from first-party data, MMP feeds, and your data warehouse.
- Proxy layer: CPI and install volume remain useful for pacing and creative testing, but are not sufficient success metrics for subscription LTV.
- Model layer: User-level pLTV scores installs using early engagement signals correlated with trial-to-paid, renewal, and net revenue at cohort maturity.
- Signal design: Time pLTV to the anchor event (install or trial start); calibrate against realized LTV and refund rate where IAP refunds matter.
- Activation: Churney sends predicted values directly to ad networks via Meta Conversions API, Google app analytics / Firebase import paths (for mobile in-app value), or MMP-orchestrated postbacks, enabling value-based bidding beyond CPI optimization.
- Readout: Compare CPI, trial-to-paid, payback, and incremental ROAS for pLTV treatment vs install-only business as usual (BAU).
Lowering CPI without changing the value signal can increase low-quality volume. pLTV shifts optimization toward installs worth paying for in margin terms.
Basic CPI:
CPI = Ad spend / Attributed installsEffective CPI (subscription context, illustrative):
Effective CPI = Ad spend / Paid subscribers acquired (same cohort, at maturity)Interpretation guardrails:
Align attribution window with network reporting when comparing CPI trends.
Report CPI with trial-to-paid, D7/D30 revenue, or LTV at cohort maturity.
Segment by campaign, geo, and network; blended CPI hides quality mix shifts.
Pair CPI with incrementality or holdout readouts when scaling new channels.
Category variants
| Model | How CPI shows up |
|---|---|
| Ecommerce / DTC | Less central; mobile app installs may feed app-commerce funnels. Web CPA often replaces CPI unless app is primary surface. |
| Subscription app | Primary UA metric paired with trial-to-paid and renewal; CPI alone over-rewards low-intent installers. Core pLTV use case. |
| SaaS / PLG | Mobile install rare; desktop signup CPA plays analogous role. Same delayed-value problem with different anchor event. |
Common mistakes
- Treating CPI as success metric for subscription apps. Cheap installs with poor trial-to-paid destroy unit economics.
- Ignoring ATT and SKAN postback delay. CPI numerators and denominators may not match finance cohort views.
- No link from CPI cohorts to LTV or payback. Scaling on CPI without maturity readout.
- Optimizing install campaigns without downstream value events. Platform never learns which installs monetize.
- Comparing CPI across networks with different fraud and match quality. Normalization requires incrementality or holdout checks.
- Using gross IAP value without refunds. Inflates perceived quality of high-CPI segments that refund.
Advertiser lens
| Role | What they ask | What good looks like |
|---|---|---|
| Head of Performance / UA | Are we buying installs we can afford? | CPI paired with trial-to-paid, D30 revenue, and payback at maturity by network. |
| VP Growth / CMO | Should we cap CPI or shift to value bidding? | Documented path from CPI goals to pLTV or tROAS with holdout proof. |
| Marketing Analytics / Data Science | Does install quality vary by source? | Cohort LTV by install campaign; calibration of early pLTV scores. |
| Data Engineering | Are MMP postbacks complete? | Install, in-app event, and revenue streams joined in the data warehouse with ID graph. |
| Finance / Procurement | What CPI still pays back? | CPI ceilings derived from margin LTV and payback, not historical install benchmarks alone. |
FAQ
What is cost per install (CPI)?
CPI is total ad spend divided by the number of attributed app installs in the same period, for a defined network, campaign, or cohort.
Why is CPI considered a proxy metric?
Install is an early funnel event. Subscription revenue, renewals, and IAP often arrive later, so CPI alone does not measure profitable acquisition.
How does CPI relate to pLTV?
pLTV predicts monetization potential at or soon after install so ad platforms can bid toward high-value users, not just cheap installs.
Is lower CPI always better?
No. Lower CPI with worse trial-to-paid, LTV, or refund rate can produce worse unit economics than moderate CPI with strong downstream value.
How do MMPs affect CPI measurement?
MMPs attribute installs and forward in-app events via postbacks. Match gaps, SKAN delay, and ATT opt-out skew CPI and require cohort validation in your data warehouse.
When should mobile UA move beyond CPI optimization?
When install volume is sufficient but payback or LTV varies materially by source, and you can send calibrated value signals (pLTV) with adequate signal volume and match quality.
How does CPI connect to payback period?
Payback uses CAC at install (or trial) against cumulative margin over time. CPI is the cost input; margin LTV and delayed value determine whether payback occurs.
Not the same as
| Term | Difference |
|---|---|
| CPA (cost per acquisition) | CPA usually means cost per converting user or customer; CPI is install-specific. |
| CPC (cost per click) | CPC is pre-install; CPI includes install conversion rate from click to install. |
| Cost per trial start | Trial is a downstream event; CPI stops at install unless defined otherwise. |
| Customer acquisition cost (CAC) | CAC often fully loaded and may use paid subscriber as denominator, not install. |
| Platform ROAS | Revenue efficiency metric; CPI is cost-per-unit at funnel top. |
