What Is eCPI? Effective Cost Per Install Explained

What is eCPI (effective cost per install)?

eCPI, or effective cost per install, is the average amount you actually paid for each install of your app, whatever way you bought the traffic. You calculate it by dividing your total ad spend by the installs that spend brought. It turns CPM, CPC, CPI and CPA campaigns into one number you can compare side by side.

The difference from plain CPI is about intent. A CPI is often a price you agree in advance (“$1.20 per install”). An eCPI is the result you measure after the fact. AppsFlyer’s cost per install glossary page describes eCPI as the de facto price you pay for installs as they come in or after the campaign, and gives the example of a $10,000 budget that drives 5,000 installs, an eCPI of $2.

If you are new to install pricing, start with What is CPI?, which covers what counts as an install and what drives the price. This guide picks up where that one stops: converting other buying models, paid vs blended eCPI, and why the number can mislead you.

The eCPI formula

The core formula is short:

eCPI = total spend ÷ installs

Both sides need the same scope. If the spend is one network’s cost for one country in one week, the installs must be that network’s installs, in that country, in that week, attributed under the same rules. Mixing a monthly invoice with a weekly install count, or one network’s cost with installs from everywhere, is the most common way eCPI goes wrong.

Three questions define any eCPI you see in a report:

• Which spend? Media cost only, or media plus agency fees, creative production and tools?
• Which installs? Installs attributed to the campaign (paid eCPI), or every install in the period, organic included (blended eCPI)?
• Counted by whom? The ad platform, your mobile measurement partner (MMP), or your own backend? They rarely agree, as the reporting section below explains.

The CPI calculator does the division for you once you have the two numbers.

Why eCPI exists: one basis for every buy

App marketers rarely buy everything the same way. A video network might sell impressions (CPM), a search or social campaign might charge per click (CPC), an affiliate or offerwall partner might charge per install (CPI) or per in-app event (CPA). Each price tag is in a different unit, so a $6 CPM, a $0.25 CPC and a $1.40 CPI cannot be compared directly.

eCPI converts all of them to the same unit: dollars per install. That lets you rank sources, set budgets and spot a campaign that looks cheap on its own price but expensive per install. For the general version of this idea across any result, not just installs, see CPM vs CPC vs CPA.

Platforms use it as a target too. Google Ads lets App campaigns bid with a target cost per install, which its help page on bidding in App campaigns describes as the average amount you would like to spend each time someone installs your app (checked on October 20, 2026). The word “average” matters: you set a goal, and the eCPI you actually get is measured afterwards.

How to convert CPM, CPC and CPA buys into eCPI

You can always use spend ÷ installs. The conversion formulas below are useful before a campaign runs, when you know the price and a few rates from past campaigns and want to predict the eCPI.

From CPM: use IPM

IPM means installs per mille: installs per 1,000 ad impressions. Unity’s installs per mille glossary entry defines it as installs multiplied by 1,000, divided by impressions. Because CPM is the price of 1,000 impressions and IPM is the installs from 1,000 impressions:

eCPI = CPM ÷ IPM

If you prefer to think in a per-impression install rate (installs ÷ impressions), the same formula becomes eCPI = CPM ÷ (1,000 × install rate per impression). They give identical results; just do not apply the 1,000 twice. A $10 CPM with an IPM of 5 gives $10 ÷ 5 = $2.00 per install. Doubling IPM with better creatives halves eCPI without any change in what the network charges. What is CPM? covers the impression side.

From CPC: use the click-to-install rate

eCPI = CPC ÷ click-to-install rate

The click-to-install rate is installs ÷ clicks, as a decimal. A $0.30 CPC with 20% of clickers installing gives $0.30 ÷ 0.20 = $1.50. Your store page drives this rate, which is why a weak listing raises eCPI across every click-based source at once. CTR vs CVR explains the two rates in the funnel.

From CPA: installs that did not convert are free

When you pay per in-app event, such as a finished tutorial, you pay nothing for installs that stop before the event. Your eCPI is lower than the CPA price:

eCPI = CPA × (events ÷ installs)

If you pay $2.50 per finished tutorial and 400 of 520 installs finish it, your spend is $1,000 and your eCPI is $2.50 × (400 ÷ 520) = $1.92. For an install-priced deal, eCPI simply equals the agreed CPI, as long as your count of installs matches the partner’s. The trade-offs between paying per install and per event are covered in CPI vs CPA for apps.

Illustrative example: four buys, one table

Sample figures only. Suppose a casual game spends $1,000 on each of four sources in the same week and country. The prices and rates below are invented for the arithmetic, not benchmarks; for published install prices by market, see cost per install by country.

On eCPI alone, the CPI partner wins and the CPA partner looks worst. Keep reading before you move the budget.

Paid eCPI vs blended eCPI

Here the industry uses the same word for two numbers, so always check which one you are looking at.

• Paid eCPI divides spend by installs attributed to paid campaigns. This is the definition in AppsFlyer’s glossary and what most dashboards show per network.
• Blended eCPI (often called blended CPI) divides paid spend by all new installs in the period, organic included. Some glossaries build this into the definition of eCPI itself: Unity’s eCPI entry describes it as a metric that accounts for organic factors like viral growth, and Singular’s eCPI glossary page says it takes into account installs from other channels such as organic or direct traffic.

Paid campaigns often lift organic installs too. A burst of paid installs can push an app up the store charts, more people see it, and friends of new users hear about it. That is the organic uplift the blended number tries to capture.

Blended eCPI flatters spend because it credits organic installs that would have arrived anyway. Paid eCPI can do the reverse and miss real uplift, or it can flatter a network that claims installs which were going to happen without it. The only way to know how much a channel truly adds is to test it, for example by pausing it in some regions and comparing; incrementality testing walks through the methods. Use paid eCPI to compare networks with each other, and blended eCPI as a sanity check on total spend against total growth.

Why a low eCPI can still be expensive

An install is not a user, and it is certainly not a customer. eCPI stops counting at the first open, so it says nothing about who stays or who pays. Two further ratios fix that:

• Cost per retained user = spend ÷ installs still active on day 7 (or day 30).
• Cost per payer = spend ÷ installs that made at least one purchase in your chosen window.

Back to the four sources in the table. Suppose, again with invented rates, that day 7 retention and the share of installs that buy something look like this:

Sample figures only. The ranking flips. The source with the cheapest installs brings the most expensive retained users and payers, and the one with the highest eCPI is the cheapest way to get people who stay. That pattern is common when a source sends low-intent users or when the price is set on the shallowest event, but your own data decides it. For realistic retention ranges, see app retention benchmarks.

The step after cost per payer is revenue per install over 30 or 90 days, which turns into return on ad spend. What is ROAS? and CAC vs CPA vs LTV take it from there.

How MMPs and ad platforms report eCPI

You will usually see at least two eCPIs for the same campaign, and they will not match.

Ad platforms

Large self-attributing platforms report their own cost per install based on the installs they believe they drove. AppsFlyer’s glossary notes that self-reporting networks such as Facebook, Google and Snap charge advertisers on CPM while optimizing towards eCPI or the advertiser’s maximum bid, and that their eCPI will often differ from the attribution provider’s because they charge on engagement whether or not they won the last-touch credit.

MMPs

Your mobile measurement partner pulls cost from each network and divides it by the installs it attributed to that network under one set of rules. Because every network is judged by the same method, MMP eCPI is usually the better number for comparing sources. It can still differ from the network’s figure because of:

• Attribution windows and models. A longer click window credits more installs to ads. What is attribution? explains the rules.
• Install definition. Most tools count the first open, not the download, so installs that are never opened do not count.
• Cost data gaps. Some networks do not pass cost, pass it late, or pass it in a currency or time zone that does not line up with your install counts.

How to lower eCPI without buying worse users

• Raise IPM. Test new creatives often; on CPM buys, every gain in installs per 1,000 impressions lowers eCPI one for one.
• Raise the click-to-install rate. Better screenshots, a clear first line and a short video help every click-based source.
• Pay for the event, not the install. Pricing on a post-install goal can raise eCPI and still lower the cost per retained user, as the table showed. Multi-event offers show how to pay for several goals in one campaign.
• Cut fraud. Fake installs lower your apparent eCPI and destroy every downstream ratio. Watch for installs that never open twice.
• Price by country. A single global target CPI overpays in cheap markets and loses volume in expensive ones.

Frequently asked questions