One SDK connection, every network competing for the same impression.
Mediation replaces a stack of direct network SDKs with a single integration through which all of your demand competes at once. Networks that support bidding bid in real time on every impression; networks that do not are managed in an automated waterfall that runs inside the same decision. This page explains how the two paths cooperate, how the auction stays honest when our own network demand is one of the bidders, and what a migration off your current setup actually involves.
Bidding and waterfall · one decision, logged
Bidding demand competes, the waterfall fills the gaps.
Demand sources with RTB endpoints bid into the unified auction on every impression. Networks without bidding endpoints participate in an automated waterfall that runs in the same decision, and the optimizer keeps its CPM targets current on its own. One request leaves your app; every source you have connected gets a fair shot at the impression inside it.
One SDK, many sources
Each additional demand source is added in the dashboard, and your app build never changes. The mediation layer owns the messy parts of multi-network life: per-source timeouts, fallback behavior when a network answers late or not at all, and credential hygiene across SDK versions. One binary, however many networks end up behind it.
Automated CPM adjustment
Waterfall targets are recalculated continuously for every country and placement from how your inventory actually clears, so the price ordering that someone used to maintain in a spreadsheet every Monday maintains itself. When a network's effective CPM drifts, its position drifts with it on the next cycle. Manual floor maintenance stops being a job.
Per-network reporting
Bidding revenue and waterfall revenue are reported separately per network, so each source's contribution carries a number you can act on. Impression-level data is available through the reporting API for publishers who want to join it with their own analytics. When something changes, the change has a name attached to it.
Sequential waterfalls leave money behind, here is the mechanism.
A traditional mediation waterfall calls networks one at a time, in the order of their last known CPM, and stops at the first fill. The order reflects yesterday's prices, not this impression's value, and every network earlier in the chain enjoys a privileged position it did not pay for. Worse, each network in a waterfall typically sees the impression and the price context before deciding, which lets buyers shade bids they would otherwise have to make honestly.
A unified auction removes both problems by construction. Every bidding network answers the same request at the same time without knowing what anyone else bid, and the highest valid bid wins. Networks that genuinely value your audience reveal it in price, and the impressions your waterfall used to sell cheaply to whoever happened to be first in line get exposed to everyone at once. That is the entire argument for in-app bidding, and it is also why non-bidding networks stay in an automated fallback: the mediation layer earns them a real share of the run.
Our own demand gets no thumb on the scale.
SkyBoosts network campaigns compete in the same decision as external demand and are subject to the same validity and policy checks. There is no last-look stage, no implicit floor advantage and no sequencing that favors our own line items. If a network campaign wins an impression, it did so on price, and the log shows it.
The same policy holds on the waterfall path: the optimizer alone sets priority, with no operator-owned boost applied to our own lines. Publishers can audit any decision from the logged bid data, and that audit trail is what gives the policy teeth.
You decide what your users see, network by network.
Mediation concentrates decisions, so it has to concentrate controls with them. Blocks and creative review apply across both the auction and the waterfall path.
Per-placement blocks
Block categories, advertisers and destinations per placement, and the blocks hold across every connected source, auction bidders and waterfall networks alike. A block list you can only apply to half your demand is half a block list, so enforcement is centralized in the mediation layer.
Creative review
Creatives pass policy review before serving and are re-checked in flight when destinations or payloads change. Problematic creatives are flagged automatically, and the flag is visible to you with its reason, so store reviews stop being the first sign of a bad creative.
See what served
Impression-level records show which network served, at what price, into which placement, for any session you need to investigate. When a user complaint or an app store review references an ad, you can reconstruct the full ad journey from the log and answer it with facts.
Connect every source, test every change, switch on your own evidence.
The current list
The mediation-supported network list with setup guides is maintained in the documentation, so it reflects the adapters that actually shipped. Coverage for a specific network is confirmed with your integration contact before anything is promised.
Managing sources
Adding a source is a dashboard configuration: enter its account ID, set its tier and timeouts, and the adapter enters the next SDK release cycle. Per-source settings cover timeout budget, fallback behavior and country availability. Every change is visible in per-network reporting, so a misconfigured source appears as its own named line and gets fixed the same day.
What can be tested
Format mixes, network groups and price points can run against a control cohort, with users split at the device level so the comparison is clean. Results are read from per-network revenue and retention deltas, which keeps audience shifts from contaminating the read on a setup change. Tests run until they reach significance or you stop them, no silently abandoned experiments.
Moving over
Migration runs placement by placement: integrate the SDK, map existing placements, run bidding and the legacy stack in parallel on a test cohort, compare like-for-like, then cut over. Nothing switches while blind, the parallel period is the evidence, and you keep it.
Honest limitation: mediation cannot reconcile cross-network discrepancies. Networks count impressions by their own definition, so totals can differ. We report both sides and explain the gap rather than hiding it.
Mediation in practice.
No. After the initial SDK integration, additional networks are added from the dashboard. New adapters ship inside regular SDK versions and are adopted on your release schedule, so your app store review cadence stays yours.
Yes, and it keeps competing honestly. Non-bidding networks run in the automated waterfall alongside the unified auction, with CPM targets maintained by the optimizer from observed performance on your inventory.
Run the migration as a cohort test. Bidding and waterfall revenue are reported separately per network, and the parallel-run period gives you a like-for-like comparison on your own inventory, seasonality and audience shifts show up in both cohorts, so the setup change is the only variable.
Yes. Every source, our own network campaigns included, bids under the same policy checks in the same decision window, and the winner is decided on price alone. The decision log records the competing bids for every impression, so you can verify fairness impression by impression.
Floors are imported per placement at migration and then managed like any other auction input. The optimizer proposes changes from the clearing prices it observes on your inventory; you approve what you want, lock what you care about, and everything the optimizer did is visible in reporting.
Mediation is part of the SkyBoosts platform rather than a separately licensed product. Commercial terms are set out in your publisher agreement, including any share applied to mediated revenue, before you integrate anything.
The same unified decision covers every format, but each one has its own trigger: banner placements re-auction on your refresh interval, interstitials auction at the natural transition points you define, and rewarded ads auction at the moment the user chooses to engage, with the mediation layer confirming completion before the app-side reward callback fires. Format pages carry the specifics for each.
One connection, every source competing.
Bring your current setup, and the integration team will map it to a unified auction with a parallel-run plan.
Unified auction · managed waterfall · auto-optimization