Technology

Infrastructure that thinks per-impression.

12 data centers, 2.4M bids/second, sub-50ms p99 latency. The AdFlux platform was built from scratch for in-app programmatic — not adapted from web display.

ML pipeline visualization
0
Data centers
0
Peak bids
0
P99 latency
0
Core uptime (trailing 12mo)
Real-Time Bidding

A first-price auction, every impression.

The AdFlux RTB engine processes bid requests in parallel across all connected DSPs and internal campaigns, returning a winning creative in under 50ms — including fraud screening.

01

Bid Request

Publisher SDK fires an ad request with device signals, geo, app bundle, ad-unit specs, and user-level IDFA/GAID (where available).

02

Scoring & Fraud Check

The ML model scores per-user eCPM across all eligible campaigns and runs 14 fraud signals — IVT, click injection, device fingerprint — pre-bid.

03

Winner Selection

First-price unified auction. The highest predicted-eCPM creative wins, ad markup is returned to the SDK, and impression/click/install callbacks are registered.

ML Engine

Models retrained every 15 minutes.

Our prediction engine runs gradient-boosted decision trees + neural embeddings on device, geo, and behavioral signals to score each impression individually — not by segment.

  • eCPM Prediction — per-user expected value across ad format, creative, and campaign.
  • Pacing — budget allocation model prevents over/under-delivery within 3% tolerance.
  • Creative Ranking — multi-armed bandit selects the highest-CTR variant with exploration budget.
  • LTV Estimation — D1/D7/D30 projected lifetime value for cohort optimization.
ML optimization
Fraud Detection

94% of invalid traffic, caught pre-bid.

Fraud Shield processes every bid request against a device reputation graph of ~2.8 billion IDs, real-time CTIT distributions, and behavioral anomaly models.

  • GIVT & SIVT classification per TAG/MRC guidelines.
  • Click-injection & click-spam detection via timing analysis.
  • SDK spoofing protection with cryptographic attestation.
  • Post-attribution fraud reconciliation reports for MMP alignment.
Fraud Shield
API & SDK

Build on AdFlux.

RESTful Reporting API, Server-to-Server postbacks, and native SDKs for iOS & Android.

Reporting API

Pull campaign or publisher data by day, country, ad-unit, creative. JSON responses, OAuth2 auth, 15-second data freshness.

GET /v2/reports/campaigns
?from=2026-04-01&to=2026-04-16
&group_by=country,creative
&metrics=spend,installs,roas

S2S Postbacks

Fire server-to-server install and event callbacks to your BI or MMP. Supports macros for click_id, device_id, publisher_id, payout, and 40+ parameters.

https://track.adflux.network/pb
?click_id={click_id}
&event=install&payout={payout}

Mobile SDK

iOS (Swift/ObjC) and Android (Kotlin/Java). CocoaPods, SPM, Gradle. 3-line init, supports banner, interstitial, rewarded, native, and playable.

AdFlux.init(appKey: "YOUR_KEY")
AdFlux.loadRewarded(adUnitId: "main")
AdFlux.showRewarded()

See it in action.

Request a live sandbox environment or schedule a platform walkthrough with our engineering team.

Get API access Book a walkthrough