Offer sheets and event reports for CPI campaigns come with their own shorthand. This glossary defines the payout, event, retention, fraud and measurement terms so you can read both without guessing.
Main points
- CPI is total ad spend divided by installs, and the effective CPI you actually pay often differs from the negotiated rate on the offer sheet.
- Install offers that pay on in-app events or retention days pay more because they filter for users who engage after installing.
- Offer sheets state whether incentivized traffic is allowed, and rewarded installs sent to non-incentivized offers usually fail validation.
- Apple’s SKAdNetwork reports installs through aggregated postbacks, with only 64 possible conversion values and privacy thresholds that hide detail.
- Install farms, click flooding and click injection are fraud schemes that misattribute installs, so count installs through a measurement partner.
App install offers have a vocabulary of their own. One sheet pays on the install itself while another pays on a seventh-day purchase event, and the reports behind both assume you already speak the shorthand. This glossary defines the terms you will meet on CPI offer sheets, network dashboards and event reports, so you can read them without guessing.
Payout Terms: CPI, eCPI and the Rest
Cost per install (CPI). The price an advertiser pays a publisher each time a user installs the app as a result of that publisher’s ad. Total ad spend divided by total installs gives the rate, so a $2,000 campaign producing 500 installs has a CPI of $4.
Effective CPI (eCPI). The price you are actually paying per install as the campaign runs or after it ends. Spend and installs rarely match the negotiated rate, so eCPI shows what the traffic truly cost.
Target CPI vs. actual CPI. Target CPI is the most an advertiser will pay for an install, and actual CPI is the average amount paid. Some media sources agree up front to optimize toward a target eCPI goal.
Cost per mille (CPM). The cost of 1,000 ad impressions. Self-reporting networks such as Facebook, Google and Snap often charge CPM while optimizing toward an eCPI goal, so their numbers can differ from an attribution provider’s.
Cost per action (CPA). Pays when the user completes a further step, such as registering or purchasing. The required action filters for intent more strictly than CPI, and per-conversion payouts usually run higher.
Lifetime value (LTV). The total revenue an installed user generates across the time they keep using the app. A healthy CPI stays under that figure, because a cheap install that never opens the app again is expensive at any price.
CPI also shifts by country, device platform, ad network and app category, and the US tends to be the most expensive market for installs.
Pro tip
Count installs through an attribution or measurement partner rather than raw ad-platform totals. More than one network can sometimes take credit for a single install, so platform numbers can overstate what your spend produced.
Installs, In-App Events and Postbacks
Install. The user downloading and opening the advertised app. Sheets differ on what triggers payment, so check whether the payout counts the store download or the first app open.
In-app event. An action taken inside the app after install, such as finishing an onboarding tutorial, making a purchase or buying a subscription. Higher-paying offers often trigger on a specific event rather than the bare install.
Conversion value. A number the app sends after install to represent different levels of engagement, set by the advertiser. In SKAdNetwork reporting, conversion values have only 64 possible values, and Apple’s privacy measures can leave some values unobserved.
Postback. The message that carries attribution data back to the advertiser or network. Under Apple’s AdAttributionKit, it can arrive 24 to 48 hours after the user launches the app, which is why event reports lag installs. Our guide to postback URLs and server-to-server tracking covers the mechanics.
Retention and Day-N Terms
D1, D7, D30. The share of installed users who return one, seven or 30 days after installing. Event-based payouts often set minimum bars here, because an install with no return visit rarely holds value.
Activity window. In SKAdNetwork, the Apple-defined period during which conversion values are observed. It varies between SKAN versions, so post-install data arrives on no fixed schedule.
Incentivized vs. Non-Incentivized Traffic
Incentivized install. An install the user completes to earn a reward, such as currency in another app. Volume comes cheap and fast, but users often install for the reward and never open the app again.
Non-incentivized install. An install from a user who chose the app with no reward attached. Payouts run higher because intent, retention and event rates are generally stronger.
Offer sheets state which type they accept, and advertisers enforce it with event data. Rewarded traffic sent to a non-incentivized offer tends to fail quality checks, and the credit disappears at validation.
Compliance alert
Confirm in writing whether incentives are allowed before you buy traffic for an install offer. Rewarded installs sent into a non-incentivized campaign are a fast way to lose a month of earnings to reversals.
Fraud Terms: Farms, Flooding and Injection
Install farm. An operation that generates installs from real devices or emulators controlled by scripts or paid workers. The installs look real in a tracker, but no user behind them ever engages.
Click flooding. Sending huge volumes of fake clicks to an ad network so that a share of genuine installs gets attributed to the fraudster’s campaign instead of the legitimate source.
Click injection. A device-level scheme that fires a click at the moment an install completes, stealing attribution for an install the user was already making.
Advertisers screen for all of this through event data, which is one reason event-based payouts pay more than bare installs. The wider toolkit appears in our guide to fraud and invalid traffic in CPA campaigns.
Measurement Terms: MMP and SKAdNetwork
MMP (mobile measurement partner). An independent service that attributes installs and in-app events across networks, giving every source one comparable count. Google Ads calls approved partners on the Apple side App Attribution Partners; they receive SKAdNetwork data but no credit for modeled conversions.
SKAdNetwork (SKAN). Apple’s privacy-preserving framework for measuring app installs at an aggregated level. It involves three actors, the ad network, the publisher app and the advertiser app, and reports results through postbacks rather than device-level data.
Coarse conversion value. The less specific attribution Apple returns when an install misses privacy thresholds. SKAdNetwork 4.0, available in iOS 16.1 and iPadOS 16.1, also allows up to three postbacks per install to show engagement over time.
Apple’s AdAttributionKit sets two attribution windows for installs:
| Attribution type | Window |
|---|---|
| Click-through | User taps the ad and installs within 30 days |
| View-through | User views the ad (no link) and installs within 24 hours |
Once the sheet makes sense, the next decision is where the traffic comes from. Our beginner’s guide to native ads for CPA offers and our comparison of push versus native traffic both cover how those sources behave on install campaigns.
Frequently asked
How is CPI calculated?
Divide total ad spend by total installs. A $2,000 campaign that produces 500 installs has a CPI of $4, though the effective CPI you actually pay often differs from the rate printed on the sheet.
What does D7 mean in an app install offer?
D7 is the share of users who return to the app seven days after installing. Event-based offers often pay only when installs clear retention or in-app event thresholds.
What is the difference between incentivized and non-incentivized installs?
Incentivized installs are made to earn a reward, so they convert cheaply but retain poorly. Non-incentivized installs come from users with genuine intent and command higher payouts.
Why does SKAdNetwork reporting look different from my ad platform’s numbers?
SKAdNetwork reports aggregated postbacks from Apple with privacy thresholds, limited conversion values and reporting delays. Ad platforms also run modeled numbers based on different logic, so discrepancies are normal.
This article is general information for marketers, not legal advice. Advertising, privacy and consumer protection rules vary by country and change over time, so check the current rules and your network’s terms, and get legal advice where needed before running campaigns.
