Why is my CPI increasing when the campaign still looks familiar? Start with the arithmetic: cost per install rises when impressions become more expensive, the recorded click rate falls, or the recorded install-to-click ratio falls. Then establish whether the movement comes from performance inside a segment, a change in traffic mix, or a measurement change. A CPI spike is an outcome to explain, not a reason to change every setting.

In this article

Why is my CPI increasing in this account? · A three-part CPI spike, worked through · Define the denominators once

Trace the increase through the account · Match the action to the component · Decide whether the signal is strong enough

How PvX Delta checks the surrounding signals · Related reading

Why is my CPI increasing in this account?

CPI equals spend divided by installs. Using compatible totals, it can be decomposed into CPM divided by 1,000 times CTR times CVR, where CTR and CVR are fractions. This identity identifies the numerical contributors. It does not prove the causal mechanism behind them.

The distinction matters for mobile attribution. If installs include view-through or modeled outcomes, installs divided by clicks is not a literal click-to-install conversion probability. The identity still cancels algebraically when the same totals are used, but describing every movement as store conversion would be misleading.

A three-part CPI spike, worked through

Here is an illustrative example, with all metrics on a consistent reporting basis.

Input or result

Earlier period

Later period

CPM

$10.00

$12.00

CTR

2.0%

1.6%

Install-to-click ratio

25%

20%

CPI

$2.00

$3.75

The first CPI is $10 ÷ (1,000 × 0.02 × 0.25) = $2.00. The later CPI is $12 ÷ (1,000 × 0.016 × 0.20) = $3.75. The increase is 87.5%, not the sum of the three percentage changes.

The multiplicative factors are 1.20 for CPM, 1.25 for the CTR effect, and 1.25 for the conversion-ratio effect. Their product is 1.875. This is a compact way to identify that all three components deserve investigation rather than blaming auction inflation alone.

At 1,000 installs, the cost difference would be $1,750 if those rates held. That is an illustrative planning comparison, not a forecast. Volume and auction selection can change when you alter spend.

CPI increasing from 2 dollars to 3 dollars 75 as CPM, CTR, and conversion-ratio factors multiply.

Define the denominators once

Metric

Formula

Watch for

CPI

Spend ÷ installs

Network versus MMP installs

CPM

Spend ÷ impressions × 1,000

Currency and inventory mix

CTR

Clicks ÷ impressions

Link clicks versus all clicks

CVR in this decomposition

Installs ÷ clicks

Attribution compatibility

IPM

Installs ÷ impressions × 1,000

A useful click-independent cross-check

Use Google's CTR definition and AppsFlyer's CPI definition as starting references. The full CPI formula guide explains how to preserve units when exporting reports.

Recompute ratios from summed numerators and denominators. Averaging the CPI of a ten-install campaign and a thousand-install campaign gives them equal weight and usually answers the wrong question. Never average rows merely because the spreadsheet makes it convenient.

Trace the increase through the account

Aggregate and OS

Confirm spend and install totals before interpreting the ratio. Check the source refresh time, reporting day, currency conversion, and install definition. Separate iOS from Android; an attribution or app-release change can affect one without affecting the other.

Country and network

Inspect the five largest countries by spend and the rest-of-world group. Calculate whether a shift toward historically expensive countries explains the aggregate increase even when their individual CPIs are stable. Keep source and destination mix separate from within-country changes.

Then split Meta Ads and Google App Campaigns. A blended CPI can rise because budget moved to a higher-cost network whose users also monetize better. That is not automatically deterioration. Pair the cost result with comparable cohort returns before reversing the allocation.

Campaign and ad set

Rank segments by absolute spend and install contribution, then by deterioration. Record which component changed most in each material segment. If the aggregate explanation disappears after segmentation, revisit the mix weights before recommending a campaign edit. The worst percentage move may be operationally trivial. Check recent changes in optimization, budget, targeting, and attribution settings around the affected campaign or ad set.

Audience and creative

If CTR falls in one creative while comparable assets remain stable, investigate the hook and audience fit. If multiple creatives weaken inside one audience, investigate saturation. If the conversion ratio falls across assets, inspect the store, app availability, onboarding-related attribution, and tracking before replacing every ad.

Google App Campaigns may expose a different level of asset reporting from Meta Ads. Keep conclusions at the granularity supported by the export; do not manufacture ad-level CPI from an asset rating.

Match the action to the component

When CPM drives the change, investigate auction and delivery mix. A budget increase, narrower eligibility, or a move into expensive inventory can alter prices. The CPM diagnosis separates broad auction movement from a local delivery problem.

When CTR drives it, compare creative response within similar segments. Inspect message relevance and creative mix, then run a controlled test. A sudden CTR decline is a symptom; it is not conclusive proof of creative fatigue.

When the install-to-click ratio drives it, check attribution and the post-click journey. A localized store-page problem, incompatible device population, or changing click mix can matter. Follow the conversion-drop diagnosis before treating the entire fall as ad quality.

When spend is stable but the denominator changes abruptly, investigate an install drop. Missing installs can inflate CPI even if real acquisition performance has not moved. Compare source availability and attribution definitions without expecting network and MMP totals to match exactly.

Decide whether the signal is strong enough

Compare complete periods of equal length and similar weekday composition. A partial morning often has a different relationship between recorded spend and attributed installs from a completed day. Attribution windows and reporting delays can create a temporary spike.

Use the sample-size gate to distinguish a material pattern from a small denominator. Review seasonality and the attribution-window guide. Set the observation horizon before looking for a favorable answer.

An efficient CPI is not necessarily a profitable CPI. For a game, cheaper users may retain poorly or generate less ad revenue. For a subscription app, install cost can improve while paid conversion weakens. Check mature D7 or D30 revenue on a consistent basis rather than assuming every reduction in CPI is progress.

How PvX Delta checks the surrounding signals

PvX Delta runs nightly deterministic checks across supported account, OS, country, network, campaign, ad-set, and audience performance. Its approved CPI-trend families sit alongside CPM, CTR, conversion, and cohort-return checks. Connected MMP and media data give the morning review both acquisition-cost and user-quality context.

Delta provides advice; your team applies campaign changes. A detected movement is not proof of causality, and this article's illustrative decomposition is not a claim that every alert includes a specific calculation display. Use the audit checklist to define the broader review and compare manual with automated auditing when assigning investigation time.