A teardown series by Guy Alon
Two tracks. The App track takes one real mobile app per issue and walks through its mediation, SDK adapters, formats, supply chain, and (where the publisher files with the SEC) the revenue mix from the 10-K. The Web track takes one real publisher site per issue and captures every bid request fired in the browser, then walks through wrapper-by-wrapper schain coverage, CPM signal, and the same supply-chain audit. The goal of each teardown is to land two or three observations an operator can act on.
2
Teardowns published
1
App track
1
Web track
8
Total surfaces analyzed
Publisher: Warner Bros. Discovery (CNN)
Fifteen wrappers fire in the page lifecycle on a US desktop capture. 3,002 of 3,063 bidder attempts time out before they can compete (98% timeout rate). Six bidders return zero measurable responses. On the premium ATF banner: 5 responses across 1,112 attempts, zero Prebid wins.
Publisher: BuzzFeed, Inc.
Google ad server (AdMob or GAM, same SDK), three SDK-level demand adapters all wired for bidding, four declared formats, and a full TCF v2 + GPP consent layer. The 10-K shows 76% of ad revenue is programmatic.
Every issue follows the same six steps. The depth of each step varies with what the build exposes and what the publisher files publicly, but the structure is constant so you can compare issues to each other.
01
App track: static analysis of the publicly distributed Android Package plus a runtime screen capture via scrcpy. Web track: live wire-level capture of every bid request fired in the browser via the Beamflow Chrome extension.
02
App track: SDK adapters wired into the mediation surface and what each one routes to. Web track: prebid wrappers and direct integrations fired in the page lifecycle, with per-wrapper bid counts.
03
App track: ad formats, placements, refresh timing observed in the runtime capture. Web track: schain coverage per wrapper, CPM signal by schain presence, authorized vs unauthorized bid share.
04
ads.txt and app-ads.txt cross-checked against each declared SSP's sellers.json file via Beamflow.co. Health score, verified vs unverified entries, top problem SSPs.
05
Where the publisher files with the SEC, the 10-K is read against the stack. Revenue mix, customer concentration, programmatic share. Numbers that change how the build reads.
06
Two to four findings worth surfacing to a team that owns the stack. Ordered by estimated yield impact. Specific enough to act on, anchored in what the capture shows.
The series is editorial. Each issue is an independent technical observation of a publicly distributed build, paired with a crawl of the publisher's public ads.txt / app-ads.txt and, where applicable, public SEC filings. SDK presence, demand-partner detection, and placement screenshots are public observations. Operational commentary is editorial. No assertions are made regarding causality, intent, quantitative revenue impact, or regulatory compliance.
If you represent a publisher featured here and want to add context, correct something, or request a teardown of your own stack, get in touch.
Run a real app or publisher site?
If you run a mobile app or a web publisher at meaningful revenue scale and want a full audit (mediation, demand, floors, schain coverage, supply chain, attribution), the free 30-minute call is the starting point. The best engagements often start with the same kind of teardown you see here.
Book a Free 30-Min Call