- Published
- Nov 3, 2026
- Updated
- Nov 12, 2026
- Reading time
- 12 min read
SSAI Ads Skip Content? Audit the Replacement Timeline
Diagnose content skipped around SSAI ad breaks. Compare source, cue and player timelines, separate insertion from replacement, and verify the return boundary.

In this guide (11 sections)
When an ad ends and part of the programme is missing, first establish whether the workflow replaces a running content interval or adds time between preserved content pieces. Trace the source, cue boundaries, returned manifest and actual player position. An ad-complete event cannot establish that the intended programme resumed.
Start with one authorized test break and a source recording covering both sides of it. Identify the last expected content before the break and the first expected content afterward. That gives the investigation a concrete acceptance condition instead of a vague report that playback “jumped.”
An older AWS re:Post question about lost content after a MediaLive splice asks whether ads can finish and then resume the underlying programme without skipping it. The discussion explains that adding a marker does not pause the incoming live source; changing the marker command does not supply that missing pause. The page displays three-year age labels. This is an evergreen architecture question, not evidence of a new service incident.
The workflow below uses current AWS documentation to make the distinction testable. Other stitchers need their own contracts; do not transfer a MediaTailor configuration recipe to every SSAI platform.
Separate delivery mode from the content timeline#
AWS's MediaTailor ad-behavior reference distinguishes replacement, which covers existing content segments, from insertion where segments do not exist. It also says InsertionMode controls stitched versus server-guided delivery independently of that distinction. A setting called “insertion mode” therefore does not promise programme preservation.
Write down two decisions separately:
| Decision | Evidence to request |
|---|---|
| How does the player receive ads? | Effective stitched or guided session and the requests actually made |
| What happens to programme time? | Approved replacement interval or a preserved continuation point |
| What marks the return? | Source cue, returned timeline and observed next content |
| What is counted? | Separate break, ad-event and receiver definitions |
The server-guided implementation documentation describes ads referenced through separate playlists and player selection during session initialization. Preserve the effective session choice, rather than treating a configuration screenshot as proof of the path that a particular device used.
This guide does not prescribe a switch between stitched and guided delivery. First determine whether the current path obeyed the intended timeline. A migration introduces additional player and tracking acceptance work.
Trace the content timeline around ad replacement
An ad-complete event does not prove that the intended programme resumed. Discuss which source, cue and player signals are available.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
Identify the source as it reaches the stitcher#
Record whether the insertion service receives live content or VOD. A stored file somewhere upstream is not enough to describe the final input. A channel assembled from files can still present a linear stream to the next service.
AWS's MediaTailor concepts describe Channel Assembly as scheduling source manifests into a linear stream. Use that architecture information to ask the right question: what timeline reaches this playback configuration, and where has room for advertising actually been created?
For its live workflow, the ad-behavior reference documents replacement and return to underlying content. For VOD, it documents zero-duration EXT-X-CUE-OUT signaling for insertion, and VAST/VMAP behavior when markers are absent. These are scoped behaviors, not interchangeable instructions.
Have the stream owner label the expected programme interval before altering anything. Is it a commercial break, a slate, a section deliberately replaceable under the distribution plan, or editorial content that must remain? The same observed splice can be correct under one plan and a serious acceptance failure under another.
Capture the marker that was emitted, not only the scheduled action#
A scheduled splice is an intention. Preserve the corresponding emitted marker and the downstream representation for the affected session.
The MediaLive schedule documentation describes inserting SCTE-35 messages and applying configured processing such as manifest decoration, blanking, blackout and passthrough. Review those effective settings with the channel owner. Avoid assuming that seeing a schedule row establishes every downstream output.
The separate supported-output reference explains that manifest decoration, blanking and passthrough have different roles. Inspect the output consumed by the next component, then the manifest consumed by the player. A marker visible upstream but absent at the insertion boundary is a different investigation from a correctly signaled replacement.
For one break, save a restricted evidence row with the action reference, emitted cue, duration, cue-in or equivalent return information, manifest capture time and relevant media sequence or period. Record unavailable fields explicitly. Do not manufacture a matching break ID simply to make several logs join.
Keep wall clock, media position and duration units separate#
A scheduled time, a source timestamp and a player position are not the same clock. Put the clock definition beside each value before subtracting them.
The MediaLive splice-insert fields define a fixed start as UTC wall clock, rather than input timecode. They define the splice duration in 90-kHz ticks. In an illustrative conversion, 30 × 90,000 = 2,700,000 ticks represents 30 seconds. Do not interpret that integer as milliseconds or assume a UI elsewhere uses the same unit.
Compare the decoded marker duration with the returned boundary and player evidence. Preserve any known clock offset, timestamp reset or discontinuity. When the relationship is unknown, report it as unknown instead of presenting a precise timing error.
For HLS, the supported-marker documentation describes cue-out/cue-in forms and SCTE-35-bearing dateranges. For its documented daterange handling, MediaTailor ignores START-DATE; when duration and an explicit cue-in both define the end, it uses the earlier cue-in setting. A visually plausible wall-clock label alone is therefore insufficient evidence of this service's break boundary.
Reconstruct the expected return before examining ad counters#
Build three short timelines for the same test:
- Source: what programme material was present before, during and after the intended break?
- Returned media: which content, ad or filler segments were available to this session?
- Playback: which material did the device actually present, and where did it resume?
Use sanitized asset references and recognizable test frames or audio cues where authorized. Preserve enough context to distinguish a repeated scene from the intended continuation. A player progress bar can use a transformed timeline; its displayed number should not be your only content reference.
If source content advanced during a replacement interval, determine whether that was planned. If the returned media contains the intended continuation but playback misses it, investigate the player branch. If the continuation is absent from the returned timeline, repeatedly changing a player callback will not explain its absence.
For DASH, apply the DASH marker reference to the actual period structure. It distinguishes single-period handling from the default multi-period case, where the first qualifying event in a period governs ad insertion. Do not count every event element as a separate accepted break without checking that scope.
A hypothetical thirty-second replacement#
Hypothetical example only. A test source contains editorial material continuously from source position 600 through 660 seconds. The approved requirement says every part of that material must remain available in sequence.
Suppose a replacement covers source positions 600–630. The device presents 30 seconds of advertising, then resumes at source position 630. The missing interval is 630 − 600 = 30 seconds. That establishes a content-preservation failure against this example's stated requirement. It does not establish an invalid impression, a refund or a lost-revenue amount.
Now compare a deliberately authored insertion test. The expected continuation remains source position 600 after the ad. If one 30-second ad is added to a complete 900-second VOD asset, the idealized presentation duration is 900 + 30 = 930 seconds, ignoring packaging adjustments. With a 30-second replacement inside the same asset, the idealized total remains 900 seconds. Those totals illustrate different timeline models; actual segment boundaries and player behavior still need inspection.
An ad-complete signal could occur in either test. It cannot tell you which content model was approved or prove that the next editorial frame was preserved. Keep that acceptance evidence beside the measurement trace.
Test return conditions and name the failed boundary#
Include a normal filled break, the relevant unfilled or filler branch, and a supported early-return case in an authorized test plan. Repeat on the player and rendition involved in the complaint. Preserve the response actually returned on each attempt; ad decisions can differ.
The MediaLive SCTE-35 actions reference distinguishes splice-insert, time-signal and return-to-network actions. The last can end a splice-insert with or without a duration. Treat that as a supported signaling contract to test with the owner, rather than an instruction to change an active production schedule during diagnosis.
For each branch, document the planned return, emitted return, returned content and observed content. Record which layer first departs from the plan. If a test was not captured through the return, label it incomplete; a successful ad start does not close it.
A useful acceptance table contains one row per selected journey, with separate pass, fail or unknown entries for content continuity, ad playback, supported event sequence and receiver evidence. Do not collapse these into one “healthy SSAI” score.
Keep content loss, ad fill and measurement separate#
A break can preserve its intended timeline while serving fewer paid-ad seconds than planned. It can also produce ad events while covering editorial material that should have remained. These outcomes need different owners and different calculations.
Use the SSAI verification guide to map event producers and evidence handoffs. Use the CTV impression discrepancy workflow when reconciling defined campaign populations. Neither count comparison establishes the return frame by itself.
When a selected creative never becomes playable, follow the MediaTailor first-play readiness investigation. Preserve asset preparation as a separate hypothesis from a content interval intentionally replaced by the current timeline.
For supported campaigns already sending relevant events to AdProof, retain the available campaign/date/timezone-scoped event or aggregate export alongside the external stream trace. Confirm event availability and filters before comparison. AdProof's received records do not recover missing programme media or establish a native stitcher integration.
The anonymized streaming-publisher case provides delivery-reconciliation context. Discuss the available measurement evidence with the stream plan and tested session before interpreting matching campaign totals as complete playback acceptance.
Close with a reproducible acceptance packet#
Keep the approved insertion or replacement requirement, input type, effective session path, source excerpt, emitted marker, returned manifest and device recording together. Add the clock dictionary, duration units, first unexplained boundary and next owner.
The proposed correction should address that boundary: source playout planning, cue generation, downstream interpretation, asset preparation or player continuation. Validate the same requirement afterward with a fresh authorized trace. A changed aggregate counter alone cannot show that the editorial interval was preserved.
When a refreshed HLS interstitial and Media3 ad group disagree#
A Media3 issue opened November 9, 2026 describes an HLS interstitial whose asset list had already resolved when a later playlist refresh added X-RESUME-OFFSET and X-PLAYOUT-LIMIT to the same interstitial ID. In the reporter's Media3 1.11.1 reproduction, parsing reflected the new values while the corresponding AdPlaybackState.AdGroup retained a 26-second resume offset and asset durations of 2 and 24 seconds. The test expected a 6-second resume offset and 6-second playout limit; a separate playback reproduction continued beyond that limit. Original Media3 issue #3530.
A Media3 maintainer replied November 10 that parsing accepts the update, but HlsInterstitialsAdsLoader does not update its internal HLS ad sessions or ad playback state. The team said it would consider the issue for its roadmap. This is evidence about the reported Media3 version and implementation, not proof that SSAI, every HLS player or a publisher's replacement timeline is wrong.
For a similar symptom, preserve one full playlist revision before and after the update, the stable EXT-X-DATERANGE ID, asset-list response and durations, parsed interstitial values, player ad-group state, and content position at the boundary. Compare the schedule/cue metadata with the state the player actually resolved. Once assets have resolved, do not assume a later playlist refresh automatically mutates the existing ad session. Reproduce in the exact Media3 version and check issue and release history before attributing the discrepancy to the stitcher or changing cue timing.
The issue cites the HLS 2nd Edition document draft-pantos-hls-rfc8216bis-22, whose cover lists a November 2, 2026 expiry. Datatracker labels it an active individual Internet-Draft, not an IETF-endorsed document or a published standard. RFC 8216 is a 2017 Informational RFC and does not define these interstitial extension attributes. Keep the cited work-in-progress document, current playlist behavior and normative status separate in an escalation.
| Layer | Capture | Diagnostic value |
|---|---|---|
| Playlist | Full old/new EXT-X-DATERANGE with the same ID and update time | Confirms which metadata changed on refresh |
| Asset resolution | Asset-list response and resolved durations | Shows what the player had already scheduled |
| Parsed object | New values after refresh | Separates parsing from downstream state propagation |
| Player state | Existing AdPlaybackState.AdGroup, resume offset and durations | Shows whether player state adopted the update |
| Playback | Actual ad end and content-resumption position | Tests the reported boundary behavior |
Treat this as a Media3-specific diagnostic branch. Do not generalize it into a claim about all SSAI integrations.
Frequently asked questions#
Does a MediaLive splice marker pause the source programme?
The older question's answer explains that the marker does not pause incoming live content. Inspect the current source and output plan; do not assume changing the marker command creates a gap for advertising.
Does MediaTailor InsertionMode select preservation versus replacement?
No. The cited reference separates stitched versus guided delivery from insertion versus replacement. Verify the applicable live/VOD and marker contract, then test the actual content return.
Does an ad-complete event prove the correct programme resumed?
No. Retain the last expected source content, the intended continuation and the device's observed return. Event and receiver evidence remain separate checks.
Should I fix skipped content by increasing a VAST timeout?
First locate the failed boundary. A loading deadline addresses a different question from which content interval was replaced. Preserve both investigations if the same session also contains a genuine media-loading failure.
Trace the content timeline around ad replacement
An ad-complete event does not prove that the intended programme resumed. Discuss which source, cue and player signals are available.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
