Keeping EPGs accurate across every destination
Your channel is showing the right programme. The electronic programme guide (EPG) says something else. For the viewer deciding whether to stay, that mismatch is already a service problem. Keeping EPGs accurate becomes harder when one channel reaches several FAST platforms, connected-TV apps and broadcast destinations, each with its own metadata workflow.
A successful export gives the operations team one piece of evidence. You still need to know which revision the destination accepted and what its guide actually displays.
Build an EPG workflow you can trace
Give every scheduled occurrence a stable identity, linked to its programme, episode, service and regional feed. Keep the approved schedule revision alongside it. Two transmissions of the same episode need separate occurrence identifiers, even when they share artwork and descriptions.
Maintain a destination contract covering identifiers, required languages, field lengths, artwork dimensions, update deadlines and acknowledgement behaviour. Record who owns the mapping and who can resolve a rejection. An integration becomes difficult to support when those answers live in somebody’s inbox.
Standards help define exchanges, but destinations still need their own mappings. DVB-I covers service discovery and retrieval of electronic programme data. Check the specification and interface your partner actually supports before designing its adapter.
Arrows represent metadata publication, separate from video delivery. Compare the destination's listing with the approved schedule; use as-run records separately to check what actually aired.
Validate meaning as well as syntax
A document can pass schema validation while describing the wrong episode. Check identity mappings, overlaps, gaps, missing artwork and language coverage with ordinary deterministic rules. These checks don’t require an AI model.
For example, XMLTV’s DTD requires programme start and channel attributes, while stop is optional. Your destination may still require an explicit end time. XMLTV also defines zero-based season and episode numbering for xmltv_ns; copying a one-based episode number directly can produce a convincing but incorrect listing.
Keep editorial and technical metadata ownership explicit. A missing parental rating should trigger a request to the authorised source. A model-generated guess can create a compliance problem that a perfectly valid XML file won’t reveal.
Make time and versions explicit
Store unambiguous instants for comparison, retain the scheduling timezone, and test daylight-saving changes and midnight boundaries. RFC 3339 provides an Internet timestamp format with UTC or numeric offsets; XMLTV has its own timestamp syntax, so convert deliberately at the boundary.
Include a revision and publication time in your internal delivery ledger, even when the receiving format cannot carry them. Record the exact payload and destination response. Reject a proposed correction if its source revision has changed since approval, then regenerate the difference against the current schedule.
Where an agent can save an operator time
The more interesting task begins when a valid update disappears between systems. An agent could inspect the approved schedule, exported payload, partner acknowledgement and available guide readback, then investigate why they disagree.
Consider a hypothetical sports channel whose live match overruns by 12 minutes. The scheduler moves the following programme from 20:00 to 20:12. One destination updates correctly; another still shows 20:00 and yesterday’s artwork. Several explanations fit: a rejected upload, an older payload arriving late, a cache delay or an incorrect regional-feed mapping.
An agent can choose the next evidence source based on what it finds, correlate a partner’s rejection message with the relevant field, and prepare a correction or support ticket with the affected occurrences attached. That investigation across unfamiliar exceptions is where model-assisted reasoning earns its place. Existing adapters should continue handling predictable retries and field validation.
- 1Identify the difference
Hypothetical example: approved revision 42 says 20:12; the destination guide still says 20:00.
- 2Approve and publish
Confirm the affected feed and occurrences. Publish only if revision 42 is still current; record the response.
- 3Verify the listing
Read back the guide after its agreed refresh interval. Close when it shows 20:12, or escalate with evidence.
The 12-minute change is illustrative. Refresh intervals and correction permissions depend on the destination agreement; no universal instant-update mechanism is assumed.
If a partner exposes no readback API, use an authorised operational portal or a documented device check. Record the device, territory, account context and observation time. Sampling one guide view provides evidence for that view; it cannot prove every receiver’s cache has refreshed.
Keep correction authority narrow
Start an agent with read-only access. Let it assemble evidence and propose a field-level difference. Give any publishing tool a tightly defined destination scope, revision check and audit trail, enforced outside the model.
OWASP’s excessive-agency guidance recommends limiting permissions and requiring approval for high-impact actions. Apply that discipline to changes affecting rights, ratings, regional availability or an imminent live schedule. Treat programme descriptions and partner messages as untrusted data, including any instructions embedded in them.
Retries need bounds, too. Use a stable operation identifier to avoid duplicate submissions where supported, and escalate repeated failures. Preserve the previous payload for investigation, but restore it only after checking that it remains valid for the current schedule.
Measure the guide viewers receive
Track time from an approved schedule change to verified destination visibility. Separate delivery failures, partner rejections and stale displayed listings. Report these by service, territory and destination so a healthy aggregate doesn’t hide a broken feed.
Agree operational targets with each partner rather than inventing a universal freshness promise. Keep unresolved discrepancies visible at shift handover, with an owner and the next check due. The first rollout can cover one service and two destinations before expanding.
Accurate EPGs depend on clear ownership and a traceable correction loop. For teams reviewing their wider channel workflow, explore Evrideo Broadcast and our guide to scheduling metadata. The agent-assisted workflow described here is an implementation approach, not a claim of an existing Evrideo integration with every destination.