The AdTech auction, from request to impression ↗
Trace the handoffs between publisher, exchange, DSP, and ad server.
Publisher × Buyer × Product × Checks
Understand the auction.
Build with confidence.
Practical AdTech guides and focused tools for the people building, testing, and operating advertising systems.
The field guide
From concepts to practical checks.
Header bidding explained ↗
Understand bidder requests, timeouts, targeting, and rendering.
OpenRTB testing checklist ↗
Build positive, negative, performance, and reconciliation coverage.
sellers.json and SupplyChain ↗
Trace seller identity and supply-path consistency.
AdTech QA checklist ↗
Test the full chain instead of only the user interface.
Website readiness checklist ↗
Review content, trust, navigation, and technical fundamentals before applying.
Original practical focus
Evidence over assumptions.
Reproducible troubleshooting
Each investigation starts with a defined scope, expected result, observable evidence, and a smallest-change approach.
Transparent examples
Illustrative numbers are labeled as examples. The site does not present synthetic results as client results or platform guarantees.
Independent publishing
Learn about the editorial approach, sources, corrections, and expertise behind the site.
Authors & expertise →Tools, without account access
Start with the evidence you already have.
Check seller-record syntax and calculate reporting differences locally in your browser.
Use the tools →About the platform
Product thinking. Practical checks.
Skills Labs Tech covers Ad Exchanges, DSPs, publisher workflows, measurement, and QA from an independent educational perspective.
About Skills Labs Tech →Choose a learning path based on the failure you see
No bids received: begin with the OpenRTB request/response guide, then test campaign eligibility, bidder timeouts, and floor rules in the no-bid lab.
Bid returned but ad not visible: use the Prebid debugging guide to separate bidder response, key-value targeting, ad-server decision, and creative rendering.
Impressions differ between systems: start with metric definitions and reporting cutoffs, then reproduce the discrepancy using a common UTC window and event IDs.
Video ad fails after an XML response: work through VAST retrieval, wrapper resolution, media compatibility, and player event logging before blaming demand.
AdTech Engineering Research & Labs
Ten original engineering investigations with synthetic case studies, evidence models and reproducible debugging methods. These are educational methods, not reports of production incidents.
- The Invisible Revenue Leak: Requests That Never Reach Auction — Causal measurement for monetization changes
- Publisher Revenue Forensics: A Reconciliation Workbook — Causal measurement for monetization changes
- AdTech Failure Injection: Safe Experiments for Auction Infrastructure — Causal measurement for monetization changes
- A Supply Path Evidence Graph for Ads.txt, Sellers.json and Schain — Causal measurement for monetization changes
- The Hidden Cost of Auction Latency: An Experimental Model — Causal measurement for monetization changes
- Designing an AdTech Incident Replay System — Causal measurement for monetization changes
- A Bidstream Data Quality Observatory — Causal measurement for monetization changes
- Privacy-Safe AdTech Debugging Without Raw User Identifiers — Causal measurement for monetization changes
- Detecting Silent Configuration Drift in Ad Servers — Causal measurement for monetization changes
- The AdTech Experiment Design Handbook — Causal measurement for monetization changes
Practical investigation notes and worked methods
Start with the failure you can actually observe
A blank ad slot, a missing bid, a lower-than-expected CPM and a reporting mismatch can look like one problem from a publisher dashboard. They are not the same problem. Start by identifying the earliest observable failure in the delivery chain. If the page never issues an auction request, investigate tag loading, consent gating and slot initialization. If the exchange receives the request but the DSP returns no bid, investigate eligibility and auction parameters. If the auction succeeds but nothing renders, inspect ad-server targeting, creative loading and browser restrictions. This distinction saves time because each symptom points to different evidence.
A practical path for publisher operations
Publisher teams can begin with seller authorization and inventory identity, then move to auction setup, rendering and reporting. Check ads.txt against the actual seller account, verify sellers.json and supply-chain consistency where applicable, and confirm that ad units use the expected formats and sizes. Next, capture a controlled browser auction and record whether requests reach the configured partners. Finally, compare the rendered impression with the reporting event. The objective is not to pass a checklist blindly; it is to find the first handoff where expected behavior diverges from observed behavior.
A practical path for exchange and DSP testing
For exchange integrations, begin with a sanitized OpenRTB fixture and a documented request-response contract. Validate required fields and supported media types before introducing commercial conditions such as bid floors or targeting. Separate malformed requests from valid requests that legitimately produce no bid. Record transport status, response time, protocol errors and any exchange-side filtering. For DSP investigations, distinguish bidder eligibility, campaign pacing, creative approval and budget decisions. A reproducible mock bidder is useful for deterministic regression tests, while a live DSP is better treated as an integration smoke test because campaign decisions change.
What makes a troubleshooting result trustworthy
A credible investigation records the initial symptom, scope, expected behavior, observed behavior, evidence, hypothesis, controlled change and result. A screenshot showing an empty slot is a symptom, not a root cause. A single successful request after a code change is encouraging, but it is not a full regression test. Re-run the original failure and at least one known-good scenario. Keep synthetic examples clearly labeled, and remove user identifiers, tokens and proprietary account details before sharing logs. Skills Labs Tech uses this evidence-first approach across its guides and labs.
Choose a focused learning sequence
New readers can start with the auction lifecycle and header bidding foundations to understand how requests travel between systems. Publisher operations readers can continue to ads.txt, sellers.json, floor prices and discrepancy testing. QA engineers can follow the OpenRTB testing checklist, Prebid debugging and the five failure labs. Engineers designing observability systems can explore the research section on incident replay, configuration drift, latency economics and bidstream quality. These paths overlap intentionally because real advertising incidents often cross organizational boundaries.
What the site does not claim
The examples are educational and many datasets are synthetic. They do not represent measured outcomes from a named publisher, advertiser or demand partner. Browser-side tools cannot inspect private DSP campaign rules or verify that Google will approve a site. Documentation can also differ across OpenRTB versions, Prebid adapters and ad-server configurations, so use the applicable vendor specification for production decisions. The goal is to help readers ask better questions and design safer tests, not to imply that a generic checklist can diagnose every production incident.