Mobile is where over 70% of digital media is consumed globally, and where the majority of programmatic impressions are served. But mobile advertising is not just "display on a small screen" — it has its own inventory types (in-app vs mobile web), its own measurement stack (MMPs, SKAdNetwork), and its own creative and attribution rules. This guide covers what actually works in mobile advertising in 2026.
1. What mobile advertising actually is
Mobile advertising splits into two fundamentally different environments. In-app is served inside apps via the app's own ad SDK — measured with mobile identifiers (IDFA on iOS, GAID on Android) and attributed through Mobile Measurement Partners. Mobile web is served inside mobile browsers — measured with cookies (increasingly limited) and attributed like desktop.
The two environments are bought through different DSPs, use different creative specs, follow different privacy rules, and often report to different teams. Treating them as one channel is the single most common mistake in mobile planning.
In-app takes 65–75% of mobile ad spend in most markets because time-in-app now exceeds time-on-mobile-web by 4-to-1. Mobile web remains critical for retail, news and travel journeys where the browser is still the destination.
- In-app and mobile web are separate environments — plan them separately.
- In-app takes 65–75% of mobile ad spend globally.
- Different identifiers, different attribution, different creative specs.
2. In-app advertising: SDKs, formats and MMPs
Every ad-supported app integrates one or more ad SDKs — AppLovin MAX, Google AdMob, Meta Audience Network, Unity, ironSource, Vungle. These SDKs manage inventory, connect to DSPs via mediation, and pass impressions upward. The mediation layer decides which SDK serves each impression in real time.
Formats include banners (adhesion 320×50 or 320×100), interstitials (full-screen between app screens), rewarded video (opt-in ads in exchange for in-app currency), playable ads (interactive mini-games, dominant in gaming), and native.
Attribution is handled by Mobile Measurement Partners (MMPs) — AppsFlyer, Adjust, Singular, Kochava. The MMP sits between ad networks and your app, deduplicates installs, provides last-touch attribution and passes post-install events to marketing platforms. No MMP means no reliable mobile measurement.
3. iOS privacy: SKAdNetwork, ATT and what actually works
Since iOS 14.5 (April 2021), Apple's App Tracking Transparency (ATT) prompt gates access to the IDFA. Opt-in rates settled at 20–30% globally. Without IDFA, individual-level attribution on iOS breaks.
Apple's replacement is SKAdNetwork (SKAN 4). It provides aggregated, privacy-preserving install attribution with delayed and coarse-grained postbacks. Configure conversion values carefully — you get 64 SKAN 4 values across a 35-day window to capture your key post-install events.
iOS advertising now operates in two modes in parallel: deterministic (for the ATT-opted-in minority) and probabilistic + SKAN (for the majority). Every serious mobile stack runs both. Losing sight of either loses money.
- ATT opt-in rates plateaued at 20–30%. Plan for the 70% you don't track deterministically.
- SKAN 4 conversion values are 6 bits (64 values) across a 35-day window. Design them intentionally.
- Deterministic + probabilistic + SKAN run in parallel on modern iOS stacks.
4. Android: GAID, Privacy Sandbox and Play Install Referrer
Android still supports the Google Advertising ID (GAID) at the app level, giving richer attribution than iOS. But Google's Privacy Sandbox for Android is rolling out and will constrain GAID access over 2025–2026 on a similar arc to iOS ATT.
The Play Install Referrer API provides deterministic install attribution for Google Play installs — critical for closing the loop between Google Ads campaigns and Play Store installs.
Plan Android for the same privacy trajectory as iOS: build your SKAN-equivalent aggregated measurement now, don't wait for the deprecation to hit.
5. Mobile creative: what works on a 5-inch screen
Portrait first. 90%+ of mobile use is portrait; landscape creative on portrait screens either gets letterboxed (wasting real estate) or auto-cropped (killing the composition). Ship portrait as the primary asset, landscape as fallback.
Thumb-reachable CTAs. The bottom third of the screen is the natural tap zone. CTAs at the top of the creative get 30–50% fewer taps than the same CTA in the thumb zone.
Playable and interactive formats deliver 5–10× higher install rates than static video in gaming and utility apps. Budget for a playable asset if you're driving app installs — the CPM premium is more than paid back in CPI reduction.
6. Measurement: MMPs, deep-linking and post-install
Attribute installs and re-engagements through your MMP. Attribute in-app events (registration, purchase, subscription) through the same MMP so the full funnel lives in one report.
Deep-linking (Universal Links on iOS, App Links on Android) sends users from an ad directly to the right screen in the app — the difference between a 30% install-to-registration and a 60%. Use deferred deep-linking so the destination survives the install step.
For paid ROAS analysis, integrate cohorted revenue data back from the MMP into your BI stack. Attribution reports alone under-count re-engagement value and over-count last-touch. Cohort-based LTV is the only honest ROAS view for mobile.
7. Buying mobile: DSPs, ad networks and self-attributing networks
For open-market mobile, buy through a DSP with strong in-app supply (DV360, The Trade Desk, Adikteev, Liftoff, Moloco). For scaled user acquisition, layer in performance-focused networks (AppLovin, Unity Ads, Meta App Install, Google App Campaigns).
Self-Attributing Networks (SANs — Meta, Google, TikTok, Snap) do not accept third-party attribution and report their own installs. Reconcile SAN self-reports with MMP attribution monthly — the delta is usually 15–35%.
For CTV-adjacent mobile campaigns (cross-device retargeting from CTV to mobile), integrate with a CTV-to-mobile ID graph (LiveRamp, Adstra) or run a device-graph-enabled DSP.
8. Launching your first mobile campaign
Pick the objective first. User acquisition (CPI/CPA) → performance networks + DSPs measured through MMP + SKAN. Re-engagement → DSPs with device-graph targeting, capped frequency, MMP attribution. Brand → in-app video + rewarded formats with brand-lift study.
Set the MMP up before you spend anything. Missing SDK integration on day one means missing attribution forever on those cohorts — you cannot backfill.
Test at least two networks in parallel for the first two weeks. Mobile inventory quality varies dramatically between sources; the same $10 CPI can be $5 real CPI or $20 real CPI once you account for fraud and cohort quality.
See Mobile CPM benchmarks in your market
The Global CPM Index shows 2026 average Mobile CPMs for every market Adyllic activates in — filtered and ready to compare.
Open Global CPM IndexFrequently asked questions
What is the difference between in-app and mobile web advertising?+
In-app runs inside apps via ad SDKs, uses mobile identifiers (IDFA/GAID) or SKAN for attribution, and is measured through MMPs. Mobile web runs inside mobile browsers, uses cookies (increasingly limited by ITP and Chrome changes), and is measured like desktop. In-app takes 65–75% of mobile ad spend because time-in-app dominates time-in-browser 4-to-1.
How has iOS ATT changed mobile advertising?+
App Tracking Transparency reduced IDFA-based attribution to the 20–30% of users who opt in. The rest are measured through Apple's SKAdNetwork — aggregated, privacy-preserving, delayed. Every iOS ad stack now runs deterministic (opted-in) + probabilistic + SKAN in parallel. Campaigns designed for pre-ATT deterministic attribution consistently under-perform because they can't see the 70% majority.
Do I need a Mobile Measurement Partner (MMP)?+
Yes, for any app that runs paid user acquisition on more than one network. The MMP deduplicates installs across networks, resolves last-touch attribution, integrates with SKAN, and passes post-install events into marketing platforms. Without an MMP, you can't compare networks fairly and you can't optimise. AppsFlyer, Adjust and Singular are the leading options.
What CPMs and CPIs should I expect on mobile?+
In-app banners: $0.50–$3 CPM. In-app interstitials: $3–$10 CPM. Rewarded video: $8–$25 CPM. CPI (cost per install) ranges from $0.50 for tier-3 markets on gaming to $8+ for tier-1 markets on high-value verticals (finance, subscription apps). ROAS-based bidding usually outperforms CPI targeting once you have 200+ conversions.
What are playable ads and when should I use them?+
Playable ads are short interactive experiences (usually HTML5-based) that let a user try a slice of the app before installing. They deliver 5–10× higher install rates than static video in gaming and utility apps. Production is more expensive ($5–20k per playable) but the CPI reduction pays back within days on any campaign above $50k budget.
How do I stop mobile ad fraud?+
Mobile fraud (install spamming, click injection, click flooding, SDK spoofing) is heavily concentrated on cheap open-exchange networks. Buy through your MMP's fraud-protection layer (Protect360 in AppsFlyer, Fraud Prevention Suite in Adjust), audit new networks in a $1–5k test, and require ads.txt / app-ads.txt compliance from every supply source.
Plan your next mobile campaign with Adyllic
In-app and mobile web supply across premium networks, MMP integration, SKAN configuration, and creative production tuned for portrait-first, sound-off, thumb-reachable formats — in 100+ markets.