Skip to content
All field notes
CTV & OTT
Published
Mar 25, 2026
Updated
Nov 3, 2026
Reading time
17 min read

SSAI and CTV: Why Server-Side Stitching Changes What You Can Verify

Audit SSAI beacon paths, server recognition, device signals and unknown traffic before interpreting CTV impression, IVT or verification reports.

AP
AdProof Research
Independent measurement desk
A television device, streaming server and measurement folder carry distinct context cards, with an amber question at the evidence join.
Server recognition and device-context evidence are separate checks; neither alone establishes a valid impression. Conceptual illustration without measured results.
In this guide (16 sections)

An SSAI beacon tells you that a particular system reported an event. To interpret it, establish who produced the event, which playback evidence triggered it, what device information reached the receiver, and how unknown records were treated. Server recognition, valid-traffic classification, viewability and brand suitability need separate evidence.

This guide maps the delivery path and adds a practical review for SSAI traffic flagged as invalid or left unclassified. It does not assume that every server beacon proves playback or that a recognized intermediary guarantees valid inventory.

How SSAI Changes CTV and Live Streaming#

Server-side ad insertion (SSAI) stitches ads into the video stream before it reaches the viewer’s device—typically at the origin, packager, CDN edge, or dedicated SSAI platform. The player receives one continuous timeline instead of separate content and ad assets orchestrated in the app.

That pattern is common in connected TV, live sports, and large AVOD/FAST apps where client-side overlays are unsupported, undesirable for UX, or inconsistent across device types. For live workflows, SSAI also aligns with broadcast-style ad breaks that must align with cueing in the content pipeline—another reason infrastructure-side stitching wins.

Client-side ad insertion (CSAI) keeps content and ads separate: the player or SDK requests ads, resolves VAST, and coordinates playback. Measurement hooks often originate from the client—closer to classic web and mobile patterns.

Once ads are stitched server-side, who fires the impression beacon—the SSAI layer, the CDN, the packager, or the player—becomes a first-class design question. Independent verification has to follow the same path the money does. Teams that treat SSAI as “just video” often discover late that their counting rules assume a client they do not fully control. For AWS MediaTailor client tracking, test session metadata retrieval, playback mapping and beacon receipt. Preserve the chosen tracking path rather than assuming every playable HLS session supplies equivalent client evidence.

For your campaign

Identify who produced the SSAI beacon

A beacon records a system event; it does not alone establish playback, viewability or traffic quality. Discuss the evidence behind your measurement.

Discuss measurement options

A short inquiry to scope relevant options; no campaign access is needed.

SSAI vs CSAI: How the Ad Path Differs#

AspectSSAI (typical)CSAI (typical)
Where stitching happensOrigin / packager / edgePlayer or app
Stream to deviceOften single manifest or transportSeparate ad + content tracks
Beacon originOften infrastructure-side + client hybridOften client / SDK-heavy
Failure modesManifest, cue, CDN cache, splice timingPlayer, SDK, ad block, OS limits
Who “owns” playback truthSplit across infra + playerMostly player/SDK
Live complexityHigh—cue drift, failover, latencyStill high, but different breakpoints

Ad decisioning—which creative fills which break—still comes from your ad server or programmatic stack (VAST/VMAP-class instructions, pod structure, tracking URLs). In SSAI, that output is consumed by the SSAI service before the affected ad segments are presented to the player. Decisioning can occur during an ongoing viewing session. This changes how you prove that what was decided is what was stitched and what was measured.

When the insertion path uses AWS MediaTailor, diagnose MediaTailor first-request preparation gaps across the ad decision, asset preparation and returned session. Keep a newly selected creative separate from evidence that its rendition was ready for the affected playback request.

If an intermediary consumes VAST before the CTV player, audit the disclosure handoff and Icons acceptance test. Name who preserves the supplied disclosure information and who renders it, and retain a representative response and released-build result alongside the stitching evidence.

Practical implication: Your MSA may still say “impression,” but the operational definition might be “beacon fired from component X under policy Y.” If X is not what finance assumed, you get “true” numbers that nobody agrees how to book.

Decisioning, VAST, and the Handoff to Stitching#

In most stacks, ad decisioning produces instructions the SSAI layer can execute: which creatives, in what order, for how long, with which tracking URLs. Those instructions often arrive as VAST (single ad) or VMAP (multiple breaks) payloads, sometimes with wrappers and redirects.

Where teams get hurt:

  • Wrapper depth and latency — Deep chains delay the moment stitching can finalize; live breaks tolerate less delay than VOD.
  • Creative substitution — A valid decision at T0 may not match what actually plays if substitution, slates, or policy fillers intervene.
  • Tracking URL variability — If the SSAI layer resolves tracking differently than the player would in CSAI, your “same campaign” is not the same measurement contract.

Document not only that VAST is in use, but which system resolves wrappers, where macros are substituted, and which HTTP calls count as the billable or verified impression for each partner.

From Encoder to Stitched Manifest: What Actually Happens#

At a high level—your implementation will vary:

  1. Content is encoded and packaged (for example HLS or DASH).
  2. Cue markers in many broadcast and streaming workflows indicate where ad breaks belong. Ecosystem familiarity with SCTE-35 / SCTE-224-class signaling is common here; exact usage depends on your packager and distribution chain.
  3. Ad decisioning selects creatives for each break or opportunity, subject to business rules, frequency, competitive separation, and geo.
  4. Stitching merges ads into the manifest or transport so the player sees a unified stream—or a seamless experience very close to it.
  5. Beacons fire—depending on architecture—from the stitching service, CDN, SSAI platform, or client.

The “single stream” is excellent for playback. It is also where reconciliation gets hard: the same logical impression must be consistent across manifest structure, actual playback, and tracking calls.

Live vs VOD: Different Verification Temperatures

Live introduces: tighter latency budgets, higher sensitivity to cue drift, failover between encoders or regions, and mid-game changes to break patterns. Your verification window may need to tolerate short-lived inconsistency while still flagging sustained gaps.

Live introduces: tighter latency budgets, higher sensitivity to cue drift, failover between encoders or regions, and mid-game changes to break patterns. Your verification window may need to tolerate short-lived inconsistency while still flagging sustained gaps. For programme material missing after a break, audit the insertion or replacement timeline and content return before treating a completed ad event as complete playback acceptance.

VOD allows: more deterministic replay for QA, more time to reconcile logs, and often simpler recovery from errors—but long-tail device behavior (seek, resume, binge sessions) still breaks naive “one play, one count” assumptions.

Beacons and Events: Who Fires What#

If you are investigating a live difference between publisher and agency totals, follow our CTV impression discrepancy investigation guide. It provides a step-by-step diagnosis, a worked reconciliation, and an evidence request you can send to each partner.

Beaconing records impressions, quartiles, completions, and related events. In SSAI stacks, those signals may not originate where your team assumes.

Common failure patterns:

  • Duplicate paths — The same logical impression triggers beacons from more than one layer (for example client + server both firing “start”).
  • Timing skew — Stitch time vs playout time vs beacon time diverge, especially in live or low-latency modes.
  • Hybrid stacks — Mixed SSAI and CSAI across partners, each with different counting rules.
  • Restart and resume — A user restarts the app mid-break; rules for whether progress events reset differ by vendor.
  • Seek and trick play — Skipping into or across a break can produce edge cases in “did the ad play” logic.

Independent verification should treat SSAI as a cross-signal problem: logs from the SSAI platform, ad server, CDN, and device or app layer need to tell one coherent story—or you need an explicit, audited rule for which layer wins when they disagree.

A Minimal “Beacon Contract” Checklist

Agree in writing on:

  • Which HTTP request (or server-side event) constitutes a billable impression vs a diagnostic event.
  • Quartile and completion rules when the user tabs away, dims the screen, or the app backgrounds.
  • Whether server-fired beacons can count if the client never renders—under what exception policy.
  • Deduplication keys: transaction ID, break ID, session ID, creative ID—what is authoritative.

If SSAI traffic is flagged as IVT, separate server recognition from device evidence#

An archived March 2026 publisher question asks whether a smaller SSAI provider will automatically be classified as invalid traffic, even when it forwards device headers. The replies describe one integration experience. They do not establish a universal whitelist, an automatic classification rule, or a guaranteed route to monetization.

Investigate three separate issues: whether the measurement service recognizes the intermediary, whether it receives the expected device and event inputs, and what its actual classification covers. A familiar server name does not settle all three.

1. Confirm the measurement relationship before changing headers

The MRC's final August 2021 SSAI and OTT guidance, sections 4.2–4.3, explains why disclosed server IP ranges can help identify SSAI but have spoofing, shared-address and changing-address limitations. It also requires verification of claimed SSAI providers and discrete IVT measurability, otherwise treating the traffic as unknown rather than valid. Recognition is therefore a starting point for evaluation, not blanket acceptance of every event from that server.

Ask the receiving service and stitcher to identify the supported integration, approval scope, last validation date and owner of ongoing tests. Record app, player, region, endpoint and version restrictions. If a provider uses “certified” or “registered,” request what was assessed and which traffic it covers. Do not infer that a process used by one measurement vendor is required, available or sufficient at another.

For a hybrid app, trace the WebView or native measurement session alongside the serving path. The correct renderer, event owner and obstruction handling need acceptance evidence before a missing measurable result is blamed on the visibility threshold.

For a concrete platform workflow, the CTV discrepancy guide's server-side integration checks describe CM360's documented onboarding, certificate and call-type requirements. Avoid copying those implementation steps into unrelated recipients.

2. Trace server and device fields as different claims

Inspect an authorized sample at each handoff: the player-to-stitcher input, outgoing tracking request and receiver's parsed record. Keep diagnostic records restricted and share only fields permitted for the investigation. A field being present at the sender does not establish that the receiver accepted or interpreted it as intended.

VAST 4.3, section 1.1.3, distinguishes the server making a request from the client it represents. Its compatibility guidance for pre-4.1 server-side requests specifies X-Device-IP and X-Device-User-Agent, with X-Forwarded-For retained as deprecated backward compatibility. That does not mean you should remove X-Forwarded-For everywhere: CM360's specific server-side integration still documents it for passing an available client IP. Verify the contract for the exact destination and request type.

IAB Tech Lab's SSAI macro guidance provides a complementary way to convey device, environment and content information. Compare resolved values with what the receiver parsed; do not assume the original VAST template proves correct substitution or forwarding.

When a macro value is missing or changes between systems, trace its template, substitution owner, emitted URI and receiver-parsed value. A browser capture covers browser-owned requests; server-owned substitutions need evidence from that server.

Evidence itemCheck on the actual pathWhat it cannot establish alone
Server identity or authenticated senderDoes it match the recipient's supported intermediary relationship?That every reported impression represents valid playback
Device IP and user agentOrigin, permitted availability, forwarding and receiver interpretationA unique person, household, or attentive viewer
App and content fieldsValues match the authorized test session and stated inventoryThat the surrounding content met every buyer's suitability rules
Playback-related eventProducer, trigger and permitted join to the tracking receiptScreen visibility without the required measurement signals

Missing information may be intentional. Google Ad Manager DAI data-sharing rules allow host-specific full, truncated or absent user IP, and full or absent user agent. In that workflow, inspect the effective host rule before calling a missing field a transport failure. Do not fill restricted values with server addresses or guessed device data to improve a report.

3. Keep unknown traffic outside the valid count without calling it proven fraud

Request the starting population, classification coverage, excluded counts and reason definitions. Check whether unknown, unsupported, IVT and duplicate categories overlap or are applied sequentially before summing them.

Hypothetical example: for one completed interval, suppose a receiver has 10,000 distinct impression records after an agreed count-once step. Its mutually exclusive review categories are:

Category in this exampleRecords
Classified, not flagged under the applied IVT rules7,200
Classified as IVT under those rules800
Unknown because required classification evidence is unavailable2,000
Total received distinct records10,000

Classification coverage is 80%: 8,000 divided by 10,000. The flagged share is 10% of classified records, or 8% of all received distinct records. Unknown records represent 20% of receipts. State which denominator a reported rate uses.

The 2,000 unknown records do not belong in the accepted valid count under this example's stated treatment. They are also not 2,000 proven fraudulent impressions. The 7,200 unflagged records do not establish viewability, attention or a sales outcome. Keep any billing adjustment separate from this diagnostic table.

A receiver connection is not necessarily the original measurement location. CM360's methodology distinguishes “Measured,” “Provided” and “Inferred” measurement sources: third-party client observations can arrive through a server-to-server integration. Request that distinction rather than equating every server-origin request with an inferred result.

Finish the review with the original exclusion report, supported integration scope, approved sample requests and receiver records, effective sharing rules, classification coverage, unresolved fields and next test owner. A changed headline rate is insufficient validation; confirm what happened to both the classified and unknown populations after the targeted repair.

SSAI Spoofing and Stitching Integrity#

“SSAI spoofing” is not one bug—it is a family of misrepresentations where inventory, stitching, or beacons do not line up. Conceptual categories:

  1. Environment misrepresentation — Inventory sold or labeled as SSAI when the actual delivery path is different, so expected controls and signals do not apply.
  2. Break and pod integrity — Ads stitched into wrong positions, wrong duration, overlapping content, or mis-timed splices relative to cue markers.
  3. Beacon–to–render mismatch — Beacons fire without a corresponding stitched asset, or fire from a layer that does not match verified playback.
  4. Substitution after decisioning — Creative or tracking redirects change between decision and stitch.
  5. Edge and cache artifacts — Stale manifests, regional variance, or failover behavior so the “same” break presents differently across users or time windows.
  6. Session simulation at the edge — Automated or scripted requests that traverse SSAI endpoints and generate plausible manifests or beacons without a human viewing context.

Symptom → Cause → Next Step (Working Lens)#

Use this as a starting lens, not a diagnosis—always validate against your logs.

SymptomPossible causes to investigateFirst checks
Ad-server count >> verifierDouble beaconing, different dedupe keys, timezone boundariesCompare event IDs, clock alignment, “who fires start”
Verifier >> ad serverStricter render rules elsewhere, filtering differencesDefinition of “rendered” vs “stitched”
Spikes by region onlyCDN cache, geo routing, failoverRegional manifest diff, edge logs
Live-only chaosCue timing, encoder swap, latencyCompare cue timeline vs stitch logs
Sudden jump after partner changeNew beacon owner, wrapper policyDiff the beacon contract pre/post migration

How Independent Verification Approaches SSAI#

Strong SSAI verification programs usually combine:

  • Architecture mapping — Decisioning, stitching, beacon ownership, hybrid CSAI paths, and failover behavior.
  • Signal reconciliation — Matching SSAI platform logs to ad server and, where available, playback or device-level confirmation.
  • IVT logic suited to infra — Distinguishing recognized intermediary traffic from unsupported or invalid activity, with automation and anomaly checks appropriate to the supported signal path.
  • Transparent limitations — Live vs VOD, latency, what is observed vs inferred, and known blind spots.

MRC-oriented reviews often focus on methodology transparency: how stitching and beacons relate to counted impressions, how deduplication works, and how hybrid paths are handled without silent double counting.

In SSAI, the single stream the viewer sees is a strength for UX—and a test for truth: verification is less about trusting one player event and more about proving the same story across the stitch, the manifest, and the beacon.

Log-Level Reconciliation: What to Line Up#

When you can access event-level data (even sampled), prioritize joining on stable identifiers agreed across systems:

  • Break / opportunity identifiers — Same break across decision, stitch, and beacon.
  • Creative and wrapper resolution IDs — So substitution is visible.
  • Session or stream identifiers — So restarts do not look like new people.
  • Timestamps with agreed clock skew — Especially across vendors in live.

If your organization cannot align three systems on one ID, your dashboard will always be a negotiation, not a measurement.

Questions to Ask Vendors and Publishers (Before Scale)#

  1. Draw the decision → stitch → beacon path on one page. Where does it fork?
  2. What happens on encoder failover mid-break?
  3. Who resolves VAST wrappers, and what is the timeout policy?
  4. What is the source of truth for “impression” in your finance system vs ops system?
  5. How are replays, seeks, and backgrounding handled for progress events?
  6. What discrepancy threshold triggers escalation, and who owns remediation?

Industry Pressures (2024–2026)#

Several forces are converging:

  • Vendor fragmentation — Multiple SSAI vendors, CDNs, and SSPs on one campaign increase reconciliation surface area.
  • Live and low-latency — Tighter windows raise the cost of cue timing errors and failover quirks.
  • Privacy and identity — Fewer portable IDs push buyers toward log-level discipline and supply-path transparency.
  • Buyer scrutiny — As CTV spend grows, so does demand for auditability, not only dashboard averages.

A Practical Buyer Checklist (Expanded)#

Before scaling spend in SSAI-heavy inventory:

  1. Architecture map — Decisioning, stitching, beacon ownership, hybrid CSAI paths, and failover.
  2. Cue and break rules — Markers, pods, slates, filler, and “what if the break is short.”
  3. Beacon contract — Single source of truth; dedupe; seek/pause/restart/replay.
  4. Geo and frequency — Edge behavior vs targeting; caps that assume household vs device.
  5. Reconciliation thresholds — What discrepancy triggers investigation, makegoods, or exclusion.
  6. Change control — Versioned docs when partners migrate SSAI or CDN—measurement is not static.

SSAI is not “unmeasurable”—but it is not client-only measurement. The teams that win treat server-side truth as part of the same workflow as creative and audience strategy.

Start by asking your partners for a single diagram: from decision to stitch to beacon. If that diagram does not exist, your verification vendor should help you draw it—and your finance team should agree which box on that diagram counts as the money.

Frequently asked questions#

Does a smaller SSAI provider automatically make traffic invalid?

The provider's size alone does not establish a traffic verdict. Ask the measurement recipient about its supported integration, verification scope and required inputs. Keep unavailable classification evidence distinct from a confirmed IVT finding.

Should we remove X-Forwarded-For because VAST calls it deprecated?

Check the applicable request contract first. VAST's legacy compatibility guidance and a receiving platform's current implementation instructions can differ. CM360 still specifies this header for its documented server-side path. Do not make a global replacement from one specification note.

Does server-to-server measurement always mean inferred measurement?

No. It may transport observations originally measured on a client, or carry other forms of evidence. Ask for the source classification, original signal producer and validation scope. The transport location alone does not answer those questions.

Does an accepted SSAI impression prove viewability or brand suitability?

It establishes acceptance under that report's event rules. Viewability needs the relevant visibility evidence and denominator; suitability needs the applicable content and buyer criteria. Neither follows automatically from a successful tracker request.

Bring the measurement contract into the campaign review#

For an AdProof campaign review, bring the campaign window, placement exports, event definitions and supported tracking path. Confirm what the available event counts can substantiate before requesting audience, visibility or outcome conclusions. Explore the measurement workflow with that scope in hand.

The anonymized streaming-publisher video case illustrates a dated impression-count reconciliation. It is not evidence that every SSAI integration, server-side signal or viewer behavior is covered.

Sources and update notes#

The SSAI recognition and signal-provenance material was checked on 28 October 2026. The linked publisher discussion is an archived March 2026 question, not evidence of a current outage or growing traffic. MRC's final August 2021 guidance and VAST 4.3 are established references, not newly announced rules. The worked counts are hypothetical, not customer measurements.

Your next step

Identify who produced the SSAI beacon

A beacon records a system event; it does not alone establish playback, viewability or traffic quality. Discuss the evidence behind your measurement.

Discuss measurement options

A short inquiry to scope relevant options; no campaign access is needed.

Back to all field notes