We use cookies to improve your experience and for marketing. Read our cookie policy & privacy policy.

Getting started · AI Ad Ops SDK

Pipeline Analysis and the AI Ad Ops report

StackUp TeamUpdated July 20267 min read

Once your SDK is live, StackUp checks your pipeline daily. Each scan captures a full window of live traffic analyzed by the AI SDK across your domains: sessions, events, and signal quality. This guide explains what you see on the Pipeline Analysis page, how the SDK monitors your data, and how to open and read the AI Ad Ops report.

✓
Before you start: this page assumes the SDK is embedded and your domain is verified. If not, follow Set up the AI Ad Ops SDK first.

The Pipeline Analysis page

Open Dashboard › Pipeline Analysis in the console. The page is built around one idea: a daily scan per domain, each summarized in a row you can drill into. From top to bottom you will find:

Headline metrics

Three cards summarize the health of your collection at a glance:

MetricWhat it shows
Active domainsHow many domains had a live SDK in the last 24 hours.
Sessions analyzedTotal sessions captured across the last 7 days.
Events collectedA 7-day rolling total of signal events.

Next to the page title, a live status badge (for example SDK live · event in 24h) confirms that events have arrived recently, so you always know whether the pipeline is receiving data without opening anything.

Last 24h activity

Below the headline cards, an activity module shows the total signal events collected in the last 24 hours, bucketed hourly in UTC, with a mini bar chart of the hourly distribution and a shorter recent window (for example, events in the last 4 hours). Use it to spot collection gaps: a healthy pipeline shows bars in every recent hour, while an empty stretch usually means the SDK stopped firing or traffic dropped.

The scan table

The core of the page is the table of scans for the last 7 closed UTC days, plus today's scan in progress. Each row is one scan of one domain and shows:

  • Scan ID — a unique identifier with the scan date, useful when sharing or referencing a specific scan.
  • Domain — the site the scan covers.
  • SDK name and version — which SDK build produced the data.
  • Generated — the scan's status and timing. Today's scan shows an in progress state with a progress bar, time remaining until the 00:00 UTC close, a projected end-of-day event total, and how the day is pacing against recent days (for example, 26% ahead).
  • Sessions and events — the volume captured by that scan.

Above the table you can search by domain, SDK name, or scan ID, and toggle timestamps between UTC and local time. Scans close on UTC days, so keep UTC selected when comparing rows against quota or pacing numbers.

What the SDK monitors

Every scan is built from the signals your collection configuration allows. Depending on the packages you enabled, a daily scan can cover:

  • Sessions and events — the raw volume of traffic the SDK observed, per hour and per day.
  • Signal quality — whether events are arriving consistently, on schedule, and in statistically meaningful volume.
  • Ad stack behavior — auctions, bidders, timeouts, rendering, ad density, consent state, and page performance, according to your signal packages.

Collection respects the scope, quota, and sample rate you set in the configuration, so scan volumes reflect what you asked the SDK to measure, not necessarily your full traffic.

Checking SDK status

Click the SDK status button above the scan table to open the status panel. It answers one question: is the pipeline healthy right now?

The panel confirms whether pipeline telemetry is active, meaning the SDK has sent events in the last 24 hours. Realtime status turns online when another event lands within the 15 minute health window. Below that, a scan status checklist tracks the four stages of a healthy pipeline:

CheckWhat it means
SDK connectedThe SDK has been detected on your pages, with how long ago it first connected.
Streaming eventsEvents are flowing, with the cumulative event count collected so far.
Statistical significanceEnough data has accumulated for the AI analysis to produce reliable insights.
Insights readyDaily scans are active and reports are being generated on schedule.
💡
Tip: if Streaming events is green but Insights ready is not, nothing is wrong. Reports need statistical significance first, and new integrations typically reach it within the first days of collection.

Viewing a report

The flow from raw pipeline to insight is short:

1

Find the scan. On the Pipeline Analysis page, locate the row for the domain and day you care about. Use search if you track multiple domains.

2

Wait for the day to close. Today's scan shows as in progress until 00:00 UTC. Closed days are final and their reports are complete.

3

Open the report. Click the scan row to open the AI Ad Ops report for that domain and day. Inside the report you can switch the date range (for example, Yesterday) and refresh the data at any time.

What the AI Ad Ops report provides

The report is a deep dive into how your ad stack actually behaved in the analyzed sessions. It opens with an overview header that establishes the context of the analysis: how many deep-dive sessions were captured, which ad libraries and versions were detected on your pages (for example your ad server and Prebid builds), and the device mix covered (desktop, tablet, mobile iOS, mobile Android).

Headline metrics

The header summarizes the day in a row of key numbers:

MetricWhat it tells you
SessionsDeep-dive captures analyzed in this report.
BotsDetected bot traffic, with its share of total traffic.
Bid requestsTotal demand requests across all analyzed sessions.
Response rateShare of bid requests that received responses, with the response count.
Avg bid eCPMAverage bid value, alongside the maximum observed bid.
Auction winsBids that won their auctions.
ImpressionsAds actually rendered.
RevenueSum of winning CPMs for the analyzed sessions.

Reading these together tells the story of your funnel: requests turn into responses, responses into wins, wins into rendered impressions, and impressions into revenue. A drop between any two stages points at where to investigate.

Report sections

Below the header, the report is organized into tabs, each answering a different operational question:

TabWhat it covers
OverviewVolume and yield aggregates for the analyzed sessions: bid requests and responses, response rate, timeouts, average and max bid eCPM, auction wins, impressions, and revenue.
Issues & OpportunitiesWhere the AI analysis found problems or headroom worth acting on.
Advertiser InsightsWho is buying your inventory and how demand is distributed.
Bidder PerformanceHow each bidder performs on volume, win rate, and pricing.
Bidder HealthAdapter reliability: timeouts, errors, and responsiveness per bidder.
Compliance & ConsentCMP detection and consent state across the analyzed traffic.
Ad UnitsPerformance broken down by ad unit and placement.
Page PerformanceThe ad stack's impact on page load and runtime performance.
⚠
Report metrics describe the analyzed sample, shaped by your configuration's scope, quota, and sample rate. When comparing against ad server or SSP totals, compare rates and ratios rather than absolute volumes.

Troubleshooting

  • No scans in the table: open SDK status and check the pipeline checklist. If the SDK is not connected, revisit the setup guide.
  • Today's numbers look low: today's scan is still in progress. Check the pacing indicator on the row before drawing conclusions, and compare only closed UTC days.
  • Report shows fewer sessions than expected: review your configuration's sample rate, scope filters, and daily session quota. All three limit what the scan captures.
Was this article helpful?