Install campaign types by network
Every large network sells an app campaign type, and they differ in where the ads show, what the advertiser controls and what the network needs in order to optimise. The platform sections of this site cover each product in depth; the table is the map.
| Network | Campaign type | Where ads show | What the advertiser controls | Where it is documented in depth |
|---|---|---|---|---|
| Google Ads | App campaigns (installs, engagement, pre-registration) | Search, Google Play, YouTube, Display and Discover placements, chosen by the system | Assets (text, images, video), budget, bid target, locations and languages; placements and audiences are automated | Smart Bidding strategies for how the bid targets behave |
| Meta | App promotion objective, including the automated Advantage+ app campaigns | Facebook, Instagram, Messenger and the Audience Network | Creative, audience (or none, under automation), optimisation event, budget, placements where not automated | Campaign objectives for how the objective shapes delivery |
| TikTok | App promotion (install and retargeting) | TikTok feed and partner placements | Creative, targeting at ad-group level, optimisation goal, budget | Campaign structure for the three levels and the optimisation goal |
| Apple Search Ads | Keyword and placement campaigns inside the App Store | Search results, the search tab, the Today tab and product pages, as the product documentation describes | Keywords and match types, audiences, bids, custom product page as destination | The product's own documentation; there is no platform section on this site |
| In-app networks and demand-side platforms | Rewarded, interstitial and playable formats inside other apps | Inventory across many publisher apps | Creative, bids, often sub-publisher controls; quality varies by partner | Treated generically here; measured through the partner, never through the network alone |
The networks divide into those that capture intent, where the person is already looking for an app, and those that create it. Apple Search Ads and the search placements of Google App campaigns are the first kind; feed and video networks are the second. Audience tools differ too: Meta and TikTok let an advertiser seed a Lookalike AudienceAudience similar to a seed list. from existing users where the measurement partner shares them, while Google App campaigns automate the audience entirely. Most apps that scale end up on more than one, with the mix set by the media planning framework rather than by whichever network's representative called last.
Optimising to installs vs in-app events
Every app campaign asks what it should optimise towards, and the answer changes the users it finds. Install optimisation tells the network to find people likely to install, which it does well and cheaply, and which includes a large share of people who install, open once and leave. Event optimisation tells it to find people likely to perform a chosen action after installing (a registration, a first order, a trial start, a subscription), which costs more per install and far less per user who matters.
The constraint is volume. A network learns from the events it receives, and the help centres for each one describe a learning phase that needs a steady flow of the optimisation event before delivery settles. An event that happens rarely starves the system, so the usual sequence is to start on installs, let events accumulate, move to a shallow event that most useful users perform early (the activation eventThe share of new users or sign-ups who complete the activation event, the action marking first real value, within a set window.), and only then consider a deeper revenue event if it occurs often enough. The cost per acquisition (CPACost of each conversion.) that matters throughout is the cost at the event the business cares about, not the cost per installAverage ad cost of each app install., and the CPA calculator turns installs, event rates and spend into that figure.
On iOS the choice is narrowed by privacy. Where the person has not allowed tracking, networks receive only the limited, delayed postbacks the operating system provides, and event optimisation works through the coarse signals those postbacks carry. The app attribution guide explains what that allows; the practical consequence for campaigns is that iOS event optimisation needs simpler event choices and more patience than Android. Before any install campaign goes live, the app install campaign launch checklist confirms the optimisation event, the measurement chain and the store listing are ready. The illustrative TikTok app-install case shows one invented team making exactly this move from installs to a trial-start event, week by week.
Creative for app ads
App ad creative has one job most other ads do not: it has to show the product being used, because the person is deciding whether to put something on their phone. The formats that dominate are vertical video in feeds, playable and rewarded units inside other apps, and the assembled text-and-image units of search and display placements. Across all of them the patterns that hold up are the same.
Open on the moment of value. The first second shows the thing the app does, not the logo. A habit app shows a streak being kept; a banking app shows a payment sent; a game shows the mechanic. Click-through (CTRShare of impressions that result in a click.) is decided before most people have read anything. Games lean on creators and gameplay video more than most apps, and the entertainment and gaming marketing guide covers how install ads fit alongside them.
Show the interface honestly. Screen recordings and framed captures outperform concept films, and both stores and networks police creative that shows features the app does not have. What the ad promises, the listing and the first session have to deliver, or the install converts and the user does not.
Make it work silently and vertically. Feeds autoplay without sound; captions carry the message. Assets produced for one network should be cut for the others' aspect ratios rather than letter-boxed.
Produce in volume and rotate. Automated campaigns want many assets to assemble from, and feed creative fatigues within weeks. A monthly batch of several concepts, each with alternative hooks, is the working rhythm, and the store's own screenshots and preview are a legitimate source of assets that are already approved and on-message.
Creative is where most of the performance variance in app campaigns lives once the optimisation event is set. The audience and bidding levers are largely automated; the assets are not.
Budgets, bidding and pacing
Most app campaign types bid automatically towards a target: a target cost per install, a target cost per in-app action, or a return target where revenue events flow. The target is a statement of what a user is worth, so it should be derived from the economics rather than from what the network suggests. Work backwards from the value of a retained or paying user, the share of installs that reach that state, and the payback period the business can tolerate; the result is the most the campaign may pay per install, and the marketing budget calculator helps size the total against the installs the plan needs.
Pacing follows from the learning constraint. A campaign that is starved of budget never leaves its learning phase; one that is flooded spends before anything is known. The pattern that works is a daily budget large enough to produce a steady flow of the optimisation event, held stable for long enough to read activation and early retention by cohort, with changes made rarely and recorded. Networks price inventory per thousand impressions (CPMCost per 1,000 impressions.) underneath, so seasonal competition for the same audiences, especially in the weeks before major shopping periods, raises the cost of every install whatever the bid strategy says.
Two habits keep budgets honest. Set the campaign's length and spend before launch and decide at the end, not during; and read cost per activated user by week rather than reacting to daily cost per install, which moves with inventory prices and says little about who was bought.
Quillhaven sets a bid target from the economics
Quillhaven, an invented budgeting app with a paid annual plan, knows from its billing data that a paying subscriber is worth about £54 over the first year and that it wants acquisition cost back within that year. Historically 28% of installs register and 9% of registrations start a paid plan, so one install is worth roughly £1.36 in expected first-year revenue, and the team sets a target cost per install of £1.10 to leave margin for creative and tooling. The campaign runs for six weeks at £300 a day on one network. The first fortnight optimises to installs at £0.92 each, but only 17% register; moving the goal to the registration event raises cost per install to £1.31 and the registration rate to 36%, which lowers cost per registration from £5.41 to £3.64. At the end of the run, cost per paying subscriber sits at £41, under the £54 ceiling, and the team scales the daily budget by a third rather than doubling it.
The store listing as the landing page
An install campaign's landing page is a page the advertiser does not own. On the App Store and Google Play the ad opens the listing, the listing converts or does not, and the conversion rate (CVRShare of visits or clicks that convert.) from listing visit to install is the step between media spend and the user. Campaign managers who ignore this step end up blaming the network for a listing problem.
Three things close the gap. First, the listing has to make the same promise the ad makes; a campaign built around one feature that sends people to a listing leading with another will convert below organic. Second, both stores allow variant pages built for a purpose: Apple's custom product pagesA variant of an app's store page, with its own address and screenshots, built for a specific campaign, audience or feature. and Google's custom store listings each have their own URL, so an ad for a specific audience or feature can land on a page whose screenshots lead with it, and Apple Search Ads can use a custom product page as the ad's destination. Third, the console's conversion-by-source reports show whether paid visitors convert like organic ones, which is the quickest diagnosis of a mismatch. The listing itself, and how to improve it, is the subject of the app store optimisation guide; the app marketing hub explains why the two guides have to be worked together.
Because campaigns land on the listing, the app store screenshot test treats cost per install for those campaigns as a guardrail when new screenshots are tested.
Before paid traffic is pointed at a new or updated listing, the app store listing checklist confirms that its screenshots, claims and campaign links are ready.
Building those variant pages step by step, from the campaign themes that earn one to retiring them when the campaign ends, is the subject of the custom store listings tutorial.
Install fraud and quality
Install fraud exists because installs are paid for and easy to fake. The common forms are click spamming, where a fraudulent source fires large numbers of clicks so that it is credited for organic installs that follow; click injection, where malware on a device fires a click moments before an install completes; device farms and emulators that generate installs with no person behind them; and SDK spoofing, which fabricates the install signals themselves. The symptoms are familiar to anyone who has looked: implausibly short or implausibly uniform click-to-install times, installs that cluster in odd hours or on odd device models, and cohorts with no activation and no retention from a particular sub-publisher.
The defences are structural. Measure through a measurement partner rather than trusting a network's own count, and use the partner's fraud detection and the rules it offers for rejecting installs that fail its checks. Prefer networks that disclose sub-publishers and allow them to be excluded. Judge every source on activation and retention, never on install volume or cost per install alone, because a fraudulent source is cheapest of all. And write a quality clause into any agreement with a smaller network: what counts as an invalid install, how it is detected, and that it is not paid for.
Quality problems are not all fraud. A network that genuinely finds people likely to install but unlikely to use the app produces the same cohort signature with no malice, and the fix is the same: move the optimisation event deeper, exclude the placements that produce the empty installs, and let the cohort data decide where the budget goes.
When install goals are set as OKRs, the north star and OKRs for app installs explains why cost per activated user, not install volume alone, should carry the paid team's key result.
To see how campaign results fit a wider metric structure, the KPI tree for app installs roots growth in activated users and treats cost per activated user as the paid efficiency guardrail.
For a Google-only plan, the Google Ads setup for app installs shows how an install campaign and an in-app action campaign work together, with assets themed and operating systems split.
On Meta specifically, the Meta Ads setup for app installs walks through the App promotion objective, SDK events and the switch from installs to an activation event.
The store page's own install rate can be worked out with the conversion rate calculator for app store listings before it feeds a cost per install estimate.
To decompose install costs into audience price, ad appeal and store page performance, the CPA calculator guide for app installs then converts cost per install into cost per active user.
To set install campaigns inside the wider plan, the app installs objective hub defines the objective as activated users and compares every channel that can deliver them.
Because every campaign lands on the listing, app stores as an install channel explains why listing work should be funded before store search ads.
When a campaign delivers installs but few of them reach the activation event, the diagnosis of installs that never become active users reads activation by source and cohort to separate the optimisation goal and fraud from listing and onboarding causes.
When it is the price of each install that keeps climbing, the diagnosis of a rising cost per install splits the rise into impression price, click-through and store conversion before naming a cause.
Where many of the people a campaign would reach already use the business's website, the comparison of web-to-app journeys and app install campaigns explains when moving those customers into the app should come before buying installs.
Common mistakes
- Buying installs before the activation event is defined and firing. The campaign cannot be judged, so it is judged on install cost, which rewards the wrong sources.
- Optimising to a deep event with no volume. The network never learns and delivery stalls.
- Treating cost per install as the result. It is the price of a visitor, not of a user.
- Creative that describes instead of shows. The person is deciding whether to put the app on their phone; show the screen.
- Sending every ad to the default listing. The promise in the ad and the first screenshot should match.
- Changing budgets and goals daily. Each change restarts learning and makes cohorts unreadable.
- Trusting a network's install count on its own. Networks count what they claim; the partner deduplicates.
- Ignoring iOS postback limits when setting goals. Event optimisation on iOS needs simpler events and more time.
- Scaling the winner by doubling. Inventory prices rise with spend; scale in steps and re-read the cohorts.
