Skip to content
All field notes
Measurement
Published
Nov 5, 2026
Reading time
14 min read

CTV App Bundle IDs: Verify Which App Your Ads Ran In

Resolve CTV app IDs against the correct store, distinguish bundle and store fields, audit delivery joins, and report unknown or conflicting inventory honestly.

AP
AdProof Research
Independent measurement desk
A television app-ID tag is matched through one of two distinct app-store catalog cards to a delivery report, while another report row remains unresolved.
App identity needs a store-qualified match. A catalog lookup alone does not prove playback or seller authorization. Conceptual illustration.
In this guide (10 sections)

To identify an unfamiliar CTV app ID, establish which store and identifier field it belongs to, then verify that value against the relevant store listing. Keep the raw value, store URL, device platform and observed delivery evidence together. An app name returned by a search engine is a lead, not a completed match.

Treat app identification, seller authorization and actual ad playback as separate checks. A successful catalog lookup identifies a claimed app; it does not authenticate the delivery record that supplied the ID or prove which show carried the ad.

This practical distinction answers an older ad-operations question asking for a master CTV bundle-ID lookup table. A separate question about missing OpenRTB app bundles and store URLs shows why a name-only report cannot finish the job. Neither question establishes that missing identifiers necessarily mean fraud.

First ask what the report's “app bundle” column contains#

Request the column's data dictionary before enriching it. A report label may refer to a store-assigned ID, an application package identifier, an exchange's internal ID or an already-normalized value. Those are different joins.

IAB Tech Lab's OTT/CTV app-identification guidance recommends store-assigned identifiers for this inventory and describes how they are communicated in OpenRTB and VAST. It is guidance updated in October 2020, not a new announcement. Current receiver requirements still need checking.

For example, DV360's OpenRTB specification references the store-assigned-ID guidance for CTV and says app.bundle, device information and app.storeurl should be consistent. It also distinguishes app names used for reference from the bundle field used for targeting. That is a concrete platform example, not proof that every exchange accepts the same extensions or normalizations.

Ask the reporting provider:

  • Is this the raw received value or an enriched output?
  • Which request field supplies it, and under which protocol version?
  • What identifies the app store or platform?
  • Can you provide the original store URL and observation time?
  • Does the value describe the app hosting playback, an advertised app or an internal inventory package?

Keep a named answer next to the export. A platform-wide statement that “bundle IDs are supported” does not define the specific column you received.

Related guide

Reconcile delivery on the same CTV population.

After checking app identity and device class, align campaign counts and reporting boundaries.

Reconcile CTV impression counts

Opens a practical guide. No email required.

Do not copy every “bundle” field into every other field#

A particularly costly mistake is assuming that similar labels always require identical strings.

The IAB VAST macro definitions distinguish [APPBUNDLE], a platform-specific application bundle or package identifier, from [STOREID], the app-store ID. [APPBUNDLE] should not be filled with an iTunes store ID. [STOREURL] describes the installed app's store URL. Some platforms use the same value for package and store identity; that is not a universal rule.

Context you are inspectingIdentify the expected meaningAudit action
OpenRTB 2.x CTV app.bundleStore identity under the applicable CTV guidance and receiver contractCheck the actual received value against that contract
VAST [STOREID]Platform-specific store IDMatch against the correct store namespace
VAST [APPBUNDLE]Application bundle/package identityVerify the package field rather than substituting an unrelated numeric ID
VAST [STOREURL]Store listing URL for the installed appCheck the listing and any redirect target
Report display nameA label for human referenceKeep it separate from the join key

Apple's App Store Connect reference documents both a Bundle ID and an automatically generated Apple ID used in the desktop App Store URL. They are different properties. Android's application-ID documentation explains how the package-like application ID identifies an app on the device and in Google Play.

Do not replace a value simply because a second field has a more recognizable format. Verify the intended translation at each boundary and record the rule used. A correction appropriate for an OpenRTB receiver can be wrong for a differently defined VAST macro.

Build a store-qualified mapping table#

Use a key with an explicit namespace, such as (store, store_assigned_id), rather than a bare ID or app name. Add the device platform as a consistency check. The same brand can publish several apps, and a single app may have different identifiers across stores.

For each mapping, retain:

FieldWhy it belongs in the record
Raw received ID and field namePreserves what the delivery system actually claimed
Store namespace and platformStops an ID from being resolved in the wrong catalog
Verified store ID and listing URLMakes the identity match reproducible
App title and listed developerGives context without turning the title into the key
Lookup time and evidenceSeparates a current lookup from a historical observation
Mapping status and reasonPreserves ambiguity instead of forcing a name
Normalization rule, if anyLets another analyst reproduce the transformation

Store IDs as text. Preserve the original spelling and punctuation; apply only normalization supported by the relevant integration's rules. Keep a raw column even when you create a normalized join column.

If a partner supplies only an internal ID, request its dated crosswalk to the store identity. Record that as a partner-provided mapping until you can verify the corresponding listing. Do not silently promote it to independently verified evidence.

A paid or free commercial app database can help generate candidates, but evaluate its store coverage, refresh dates and provenance before using it as the source of record. The useful output is a traceable match for your actual inventory, not a large spreadsheet whose unknown coverage is hidden.

Verify the listing and preserve the historical gap#

Open the listing for the identified store and check the ID, app title, developer and supported platform information that is actually exposed. Preserve the resolved URL and evidence. If the supplied URL redirects, verify the destination rather than treating a successful HTTP response as the match.

A listing unavailable in your region, a removed page or an inaccessible catalog leaves an evidence gap. It does not automatically establish a nonexistent app. Record the limitation and ask the store or publisher for an appropriate supporting record. Do not bypass access controls.

Be careful with time. A listing checked today is evidence of what you can observe today. It does not prove the app's name, owner or availability on the campaign date. If that historical distinction matters, request retained records or a dated change explanation.

Treat screenshots of a branded television interface similarly. They can help document a test, but they do not by themselves prove the exact installed app, the store ID sent downstream or campaign-wide distribution. Attach the test's platform, app version and relevant identifiers instead of relying on appearance alone.

Join delivery records without changing the total#

Once the mapping is ready, join it to an agreed delivery cohort: the same campaign, event definition, dates, timezone and counting policy. A store lookup is not a reason to change the impression population halfway through the analysis.

Check join cardinality before summarizing results. Each input delivery row should receive at most one resolved mapping for the applicable key and period. If the lookup table contains two candidate rows for one key, classify the ambiguity rather than duplicating the delivery row.

A useful validation sequence is:

  1. Save the input row count and measured-impression total.
  2. Confirm that mapping keys are unique for the intended observation period.
  3. Perform the join while retaining unmatched rows.
  4. Compare the post-join row count and impression total with the original.
  5. Put conflicting candidates into a separate review table.
  6. Recalculate coverage from the preserved original denominator.

Do not drop unmatched rows through an inner join and then present the remainder as the whole campaign. That can make inventory coverage appear better by removing the cases that need investigation.

If you are comparing two measurement providers, first follow the CTV impression-reconciliation workflow. Resolve event scope and duplicate counting before attributing their difference to app identity.

A worked example with explicit unknowns#

The figures below are hypothetical. They illustrate a mapping audit, not AdProof customer results, an industry benchmark or a fraud rate.

Suppose a completed campaign cohort contains 100,000 measured impressions under an agreed counting definition. After joining a store-qualified mapping and comparing it with a dated inventory plan, the analyst assigns one status per impression:

ResultImpressionsShare of the original cohort
Store identity resolved and inside the plan60,00060%
Store identity resolved and outside the plan15,00015%
Multiple plausible mappings remain10,00010%
Required identity information missing10,00010%
Supplied ID, store or platform conflict5,0005%
Total100,000100%

The store-identity resolution rate is 75,000 / 100,000 = 75%. Confirmed in-plan identity coverage is 60,000 / 100,000 = 60%. The remaining 25% has no resolved identity in this audit.

Among resolved identities, the in-plan share is 60,000 / 75,000 = 80%. That is a valid description of the resolved subset, but it must not replace the 60% whole-cohort result. Both numerator and denominator belong in the report.

“Outside the plan” needs investigation against the actual agreement and targeting setup. It is not automatically unauthorized selling, invalid traffic or refundable spend. Likewise, conflicting fields require a provenance review; they do not establish intentional spoofing on their own.

If a later lookup resolves an ambiguous ID, update the mapping status with its evidence and timestamp. Do not describe that reporting improvement as proof that ad delivery itself improved.

Reconcile the app identity with the reported device class#

A store-qualified app match does not settle whether delivery belongs in a purchased CTV bucket. A streaming service can be available on several kinds of device. An older programmatic discussion asks how buyers and publishers handle the CTV-versus-OTT distinction in actual buys. Its market-share percentages are not verified campaign evidence and are not adopted here.

IAB Tech Lab's terminology guidance uses CTV for television devices and devices connected to televisions; mobile and desktop are outside that device category. OTT describes content rather than requiring one device class. Ask the buyer to state the device requirement separately from the content or service being purchased.

Create a device-class appendix to the app mapping rather than changing the store ID until it looks like a television app:

Evidence layerPreserveQuestion to resolve
Bought definitionExact campaign or deal wording and applicable device inclusionsDoes the agreement require a television screen, or allow other OTT devices?
Raw requestActual device field, schema/version and relevant supplied contextWhich classification did this integration receive?
Platform reportReturned device label and report settingsHow was the raw signal grouped into this label?
Independent observationAvailable, permitted test-device and player evidenceDoes the controlled example agree with the declared context?

Use the correct field namespace. The current AdCOM device-type list distinguishes Connected TV, Connected Device, Set Top Box, Phone and Tablet. An implementation's aggregation of those categories still needs documentation. Do not treat every non-phone value as television delivery, or paste an AdCOM object field into an OpenRTB request without checking the applicable schema.

For a concrete receiver example, the DV360 OpenRTB specification describes device.devicetype as highly recommended and used to improve recognition. It separately recommends consistency among app identity, store URL and device information. That does not make a supplied device code independent proof of the physical screen, nor establish that every vendor uses Google's recognition behavior.

A useful controlled test fixes the installed app, store listing, actual test hardware, integration version and permitted emitted request. Compare the supplied device classification with the receiver's exposed report result. If the application changes device class, repeat on that class rather than assuming that a television test validates its phone version. Keep a missing signal separate from a conflicting one; identify the producer and the receiver of the disputed field.

Separate hypothetical device audit: suppose a 1,000-impression campaign cohort has 900 records with a matched app identity. Within those 900, the comparable evidence classifies 650 as television delivery and 150 as phone/tablet; 100 remain unclassified. Another 100 lack the required app match. The observed television share is 65% of the original cohort: 650 / 1,000. Among the 800 app-matched records with a classification, it is 81.25%: 650 / 800. Report the 100 unclassified and 100 unmatched records beside those rates. Neither number is a verified physical-screen census, an attention measure or a fraud rate.

Escalate a conflict with the exact mapping row, device field, report bucket and controlled evidence. Do not relabel the unknown population as television inventory or as fraud. The finding may be a reporting-definition difference, missing context or a verified configuration mismatch; retain whichever distinction the evidence supports.

App identity does not identify the show or selling rights#

An app can contain multiple channels, programs or content categories. Matching the app cannot establish that the ad ran in a particular live event, episode or suitability category. Request the appropriate content-level records when the buying commitment is more specific than an app.

This is why a public question about access to premium CTV inventory cannot be settled by recognizing a familiar app name in a report. The placement, deal and content scope still need evidence. Do not adopt commenters' pricing, sourcing or margin claims as proof about your campaign.

Seller authorization is another join. Use the app-ads.txt, sellers.json and SupplyChain investigation to connect the app's discovery path, relevant seller accounts and actual transaction. Keep that result separate from the catalog identity match.

Finally, a valid-looking identifier in a request does not authenticate the claimed playback. Where your supported integration permits it, trace an authorized test from the actual app environment to the receiving measurement records. Use the SSAI provenance guide when a server acts on the player's behalf. Label whether the identifier was observed in the app, supplied by an intermediary or enriched later.

Turn an unresolved ID into a specific evidence request#

Send the responsible partner a small redacted sample rather than an undifferentiated list of unknown app names. Include the original value, field, expected store/platform, supplied URL, affected campaign interval and why the match failed.

Ask who supplied the value, where it was transformed, which mapping version was used and what evidence links it to the originating app. Request a corrected sample and its effective release time before changing a campaign-wide inclusion list.

For campaigns measured by AdProof, retain the supported campaign, package-name, date, timezone and report filters with your export. Use that scoped record alongside the catalog evidence in the review. It does not imply automatic access to every app store or the ability to reconstruct missing identifiers. The streaming-publisher video case provides context for inspecting delivery populations; it does not certify the identity of this campaign's apps.

Request an AdProof measurement review with the cohort, raw app identifiers, mapping evidence and unresolved questions. The useful deliverable is a clear account of which identities are supported, which claims require other evidence and who owns the next check.

Frequently asked questions#

Does a valid streaming-app ID prove the impression was on a TV?

No. Verify the app's store context and the delivered device classification separately. A service name does not establish the physical screen, and a declared device code needs its integration and reporting context checked. Keep unclassified and conflicting records visible.

Is there one universal CTV app bundle-ID format?

No. Identify the store and the protocol field before interpreting the value. A numeric store ID, a package-style identifier and an exchange's internal ID require different evidence. Do not join on display names alone.

Can I paste OpenRTB app.bundle into VAST APPBUNDLE?

Not without verifying their meanings for the integration. CTV OpenRTB guidance uses store identity, while the VAST APPBUNDLE macro expects an application bundle or package identifier and distinguishes it from STOREID. The fields can coincide on some platforms and differ on others.

Does an app-store match prove my ad ran there?

No. It resolves a claimed identifier to an app record. Seller authorization, authenticity of the delivery claim, counted playback and any show-level commitment require separate evidence.

Should unknown or conflicting app IDs be counted as fraud?

Keep them as unresolved identity findings until evidence supports a more specific conclusion. Report their share against the original delivery cohort and identify the missing check. Do not quietly drop them or label them valid solely because an app name looks familiar.

Your next step

Reconcile delivery on the same CTV population.

After checking app identity and device class, align campaign counts and reporting boundaries.

Reconcile CTV impression counts

Opens a practical guide. No email required.

Back to all field notes