Mobile ad networks explained for beginners

A mobile ad network is any company that sits between a game buying installs and the inventory, the ad slots inside other apps, where those installs come from. That's the loose definition, and it hides a lot. In practice the term covers several distinct kinds of business: self-attributing networks that run their own inventory and report their own results, SDK networks whose code sits inside publisher apps to serve ads directly, demand-side platforms that buy inventory across many exchanges on a bidder's behalf, and mediation layers that let a publisher app run several ad sources at once and pick the best-paying bid for each impression. A buyer's job is to know which kind of company they're dealing with. It changes who sets the price and who owns the targeting, and it changes whose report of the result everyone else has to take on trust.

What is a self-attributing network?

A self-attributing network, often shortened to SAN, is a platform large enough to both serve ads and claim credit for the installs those ads produce, based on its own view of the user. Meta and Google are the best-known examples, with AppLovin's Axon alongside them. These platforms report which installs they believe they drove, and mobile measurement partners like AppsFlyer or Adjust generally accept that claim unless a competing, better-qualified signal exists; that acceptance is the basis of the industry's long-running "last click wins" attribution convention. Where does the buyer's control sit? Mostly at the level of campaign objective and budget, plus targeting and creative. The bidding mechanics inside the platform's auction stay largely opaque, and that opacity is the trade a buyer makes for access to a very large audience and a heavily automated, AI-driven optimisation engine.

What is an SDK network?

An SDK network supplies a software development kit that publisher apps install directly, which lets the network serve ads from inside that specific app's inventory rather than buying access to it through an exchange. Historically ironSource and Unity's Vector business (and its predecessor Unity Ads) grew this way, as did Vungle before it moved under Liftoff and AppLovin's own network business before Axon. An SDK network usually pays the publisher app directly for showing the ad and takes its own cut on top of what it charges the advertiser, which means it sits on both sides of the transaction at once, sourcing supply from publishers and demand from advertisers. In the UI this looks much like a self-attributing network. The reach is different, though: it comes from which publisher apps chose to install that specific SDK, not from an exchange-wide inventory pool.

What is a DSP, and how is it different?

A demand-side platform, or DSP, doesn't own or operate the ad-serving inventory itself. Instead it buys impressions on a buyer's behalf across many exchanges and supply sources in real time, bidding into auctions the DSP doesn't control. Moloco is the clearest example built specifically for app install and re-engagement demand, alongside platforms like Digital Turbine's on-device and DSP offerings. The appeal is reach: a very wide set of supply sources through one interface and one optimisation model, plus in most cases a first-party view into how the DSP's own machine learning is weighting signals, which some buyers find more transparent than a fully closed SAN. The trade is dependency. A DSP relies on the exchanges and supply-side platforms it connects to for the inventory quality it can offer, so its result is only as good as the supply behind it.

Where do mediation and in-app bidding fit in?

Mediation and in-app bidding sit on the publisher's side of the transaction rather than the advertiser's, but every buyer needs to understand them because they determine which of the demand sources above actually get a chance to win an impression. A publisher app running mediation, through a layer like LevelPlay or AppLovin MAX or Google's mediation stack, plugs in several ad networks and DSPs as demand sources for its own inventory and lets them compete for each impression, either through a waterfall that asks sources in a fixed order, or through in-app bidding, where multiple sources bid simultaneously in one unified auction and the highest bid wins. In-app bidding has become the more common model industry-wide because it lets a publisher capture closer to the true market value of each impression rather than the fixed price a waterfall's next-highest source was willing to guarantee. Why should a buyer care? Because it explains why win rates and effective costs can shift when a publisher app changes its mediation setup, even if the buyer changed nothing on their own side.

Who controls what, and where?

Control sits unevenly across this stack. A buyer directly controls budget and targeting inside whichever platform they're using, along with creative and bid strategy, whether that platform is a SAN or an SDK network or a DSP. What a buyer doesn't control is the auction mechanics inside a closed platform, the specific inventory mix a DSP's supply partners make available on a given day, or how a publisher's mediation stack ranks competing demand sources. Reading a cost or performance shift correctly starts with asking which of these layers actually moved: the buyer's own settings, the platform's internal optimisation, or the underlying inventory and mediation environment the ad lands in. Conflating them is the most common reason a buyer pins a change in results on the wrong cause.

An exchange, the layer underneath a DSP, deserves a separate mention because buyers rarely touch it directly but its behaviour still shapes results. An exchange is simply the marketplace where impressions from many publisher apps go up for real-time bidding, and a DSP's job is to decide, impression by impression, which of those available slots deserve a bid and at what price. Two DSPs connected to overlapping exchanges can still produce different results for the same advertiser, because each DSP's own model decides which impressions in that shared pool are worth pursuing and how aggressively to bid for them.

One question, then, for a newcomer buying media across several of these categories at once, whenever a number moves: which layer of the stack changed? A rising cost per install inside a self-attributing network's dashboard is a signal about that platform's own auction dynamics or the buyer's own targeting choices. The same rise showing up simultaneously across every source, including channels bought through entirely separate DSPs and SDK networks, points instead to something further down the stack, in the publisher inventory itself or in a mediation configuration that changed how it ranks demand sources. Telling those two situations apart quickly is what separates a buyer who diagnoses a genuine problem from one who reacts to noise inside a single dashboard.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Why self-attributing networks always look better
• Programmatic DSPs for games, explained