Systems and records
Advertising approval workflow data: recording a multi-party campaign sign-off
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
Represent a campaign approval as a timeline of decisions, not a single status. Each entry names one party role, one asset version, the decision, the rule or evidence relied on, and any rights or disclosure conditions. The final scope, meaning which versions may run where and for how long, is derived from those entries, which is what makes advertising approval workflow data useful for AI reasoning tasks.
Provenance trail
Advertising approval
Key takeaways
- A final approved status hides which version was approved, the conditions attached and who still had to decide.
- Every decision should point at exactly one asset version, so revisions visibly reopen earlier sign-offs.
- Conditional and deemed approvals are different decisions from affirmative approvals and need their own labels.
- Platform review is a separate party whose limits or disapprovals belong on the timeline unaltered.
- Approval metadata is often the agency's own record, while creative assets, briefs and talent materials frequently are not.
How do you represent a campaign approval that needs several parties' decisions?
Represent it as a timeline of decisions, not a status. Each entry records one party, one asset version, the decision that party made, the rule or evidence it relied on, and the conditions attached. The campaign's final scope is then derived from those entries: which versions may run, where, for how long and with which disclosures.
That structure is what makes advertising approval workflow data useful for AI. A model that sees only the last status learns that campaigns get approved. A model that sees the sequence learns that a legal reviewer asked for a qualifier on a claim, that a talent release limited the territory, that a revision reopened an earlier sign-off, and that a platform review still limited where the ad could show. Those are the reasoning steps a creative operations lead actually performs.
This guide sets out who decides what, a field-level dataset brief, the conditions that change what may run or be reused, an illustrative worked case, the exception where platform review overrides internal approval, and the records that should stay out.
Who decides what before an ad can run?
Most campaign approvals involve at least four kinds of decision: whether the work is on brief and on brand, whether its claims are supported, whether the people and materials in it are cleared for this use, and whether the media platform will accept it. Each is made by a different party with different evidence, and each leaves a different record.
The parties do not decide in a fixed order. Claims review often runs in parallel with brand review, rights clearance can start before the first cut exists, and platform review happens only after the ad is submitted. That is why the timeline needs its own sequence numbers rather than relying on a workflow template's stage names, which describe the intended order and not what happened.
Roles also overlap in smaller teams. One person may act as brand owner and claims reviewer on the same day. Record each decision under the role it was made in, because the role determines which rule applied, and a model should learn that a brand comment does not clear a claim even when the same person wrote both.
| Party | What they decide | Evidence they rely on | Record it leaves |
|---|---|---|---|
| Brand or marketing owner | Whether the work meets the brief and brand standards | Creative brief version, brand guidelines | Comment thread or proofing approval on a specific version |
| Legal or claims reviewer | Whether each express and implied claim is supported and disclosures are adequate | Substantiation file, disclosure rules | Annotated script or proof with required changes |
| Rights and talent reviewer | Whether people, music, footage and images are cleared for the planned media, territory and term | Talent agreements, model releases, stock and music licenses | Clearance log entry per asset element |
| Agency creative lead | Whether revisions preserve the idea after changes | Prior versions and reviewer notes | Version history in the proofing or asset system |
| Platform policy review | Whether the ad complies with the platform's policies and where it may show | Platform policy | Review status on the ad, with any limitation or disapproval reason |
| Traffic or media operations | Which exact asset ships to which outlet | Asset identifier, final approvals | Traffic instructions keyed to an asset identifier |
Why does a single approved status throw away the reasoning?
Proofing and project tools usually end a job with one state such as approved or complete. That state is true but nearly empty: it does not say which version was approved, whether the approval was conditional, or which reviewer's condition changed the final cut.
Conditional approvals are the most valuable records and the most often lost. An approval with changes is really two decisions: a rejection of the version reviewed and a pre-approval of a described fix. If the fix is never checked against the condition, the dataset should show that gap. Likewise, an approval given on version 3 does not carry to version 4 unless a reviewer says so; keeping the version on every decision is what lets a model learn when sign-off must be repeated.
Silence is another trap. Some agreements treat an approval as given if the client does not respond by a deadline. Record that as a deemed approval with the clause and date that produced it, never as an affirmative decision by a named role.
Which fields go in the approval timeline and rule evidence brief?
The asset below is the dataset brief: one row per decision event, grouped by campaign and asset version. Each field names its source and the check that makes it trustworthy. Example values are illustrative.
Build it by joining three exports on the asset identifier and version. Proofing or review tools supply the comments and approvals per version. Clearance logs or a rights spreadsheet supply the talent, music and stock conditions per asset element. Traffic instructions and platform status reports supply what actually shipped and how the platform treated it. A shared asset identifier makes the join reliable; Ad-ID, which describes itself as the industry standard identifier for advertising assets across media, is one common choice, and an internal key works as long as it is stable across versions.
Derive the final scope with a rule, not by hand. Start from the latest version every required party approved, intersect the rights conditions of each element in it, and add every disclosure a reviewer required. When the derived scope differs from what traffic actually shipped, keep both and flag the case; that difference is often the most instructive record in the set.
- Task prompt: given the timeline up to a point, state which versions may run, in which media and territories, with which disclosures, and which party must still decide.
- Expected answer: the final_scope derived from the recorded decisions and the list of open conditions.
- Scoring: an answer that allows a version a reviewer rejected, or omits a rights limit, fails even if the rest is right.
| Field | What it holds | Example value (illustrative) | Validation check |
|---|---|---|---|
| campaign_id and brief_version | Campaign key and the brief version in force when the decision was made | CMP-0412, brief v2 | Brief version date precedes the decision |
| asset_id and asset_version | Asset identifier, such as an Ad-ID code or internal key, and its version | 30-second cut, version 3 | Every decision points at exactly one version |
| party_role | Role of the decider, never a personal name | Claims reviewer | Role is from a fixed list |
| decision | Approve, approve with conditions, request changes, reject, deemed approval, withdraw | Approve with conditions | Conditional decisions have at least one condition |
| rule_or_constraint | The rule applied: claim support, disclosure, brand standard, platform policy, rights term | Implied claim needs qualifier | Each non-approval cites a rule |
| evidence_link | Identifier of the substantiation file, release, license or policy page relied on | SUB-77 test summary | Link resolves to a record dated before the decision |
| rights_conditions | Media, territory, term and any exclusivity attached to people, music or footage | Paid social, United States, 12 months | Conditions present for every asset with people, music or licensed footage |
| timestamp and sequence | When the decision was made and its order in the timeline | Day 6, step 9 | Sequence is strictly increasing within a campaign |
| final_scope | Versions cleared to run, with media, territory, term and required disclosures | Version 4 only, with on-screen disclosure | Derived from decisions, never typed by hand |
| outcome | What ran and what happened: platform status, pulled, re-cut, expired | Ran, platform status eligible | Outcome comes from a dated record, not the project status |
Which revisions, claim rulings and rights limits narrow what may run?
Three kinds of event change what may run or be reused: a revision, a rights condition and a substantive claim or disclosure ruling. Any one of them can narrow the final scope, and a dataset needs all three on the timeline to explain why the scope ended where it did.
Revisions reopen earlier approvals. Claims rulings travel with the claim rather than the version: the FTC expects advertisers to have a reasonable basis for objective claims before an ad runs, so a claim that loses its support during review should be removed or qualified in every version that carries it. Endorsement content adds a disclosure rule; the FTC's guidance says a material connection between an endorser and an advertiser should be disclosed clearly and conspicuously, so a reviewer's disclosure requirement is a condition the final scope must carry.
Rights conditions are the main limit on reuse. Talent, music and stock licenses are usually bounded by media, territory and term, and image metadata standards such as IPTC Photo Metadata include fields for the copyright owner, the licensor and the model release status of people in an image. Ownership of the creative itself also varies: under U.S. copyright law, a commissioned work is a work made for hire only in listed categories and with a signed written agreement, so who owns an agency-made asset depends on the contract. That question decides whether the asset, or only the approval metadata about it, can ever be licensed.
Put these three event types in the rule_or_constraint field with a fixed vocabulary, so that a later reader can count them. A campaign set where most rejections cite brand standards teaches something different from one where most cite claim support or expired rights, and the mix is part of what a buyer of the records needs to understand before using them for training or evaluation.
Worked example: a product claim video with two revisions (illustrative)
All names, durations and counts here are illustrative. A household products brand commissions a 30-second video with an on-screen claim that the product removes most common stains, an employee testimonial and licensed music. Four parties review it over 9 working days.
Version 1: the brand owner requests a shorter opening. Version 2: the claims reviewer approves with conditions, because the substantiation file supports the claim only for a defined list of stains, and requires an on-screen qualifier. The rights reviewer notes that the music license covers online video in the United States for 12 months and that the employee's release covers paid social but not broadcast. Version 3 adds the qualifier, and the creative lead approves it.
The employee testimonial raises a disclosure condition: viewers may not know the speaker works for the brand, so the claims reviewer requires an on-screen statement of the relationship. Version 4 adds it, and the claims reviewer approves version 4 only.
Final scope, derived from the timeline: version 4, paid social and online video, United States, 12 months from first run, with both on-screen statements. A dataset row for this campaign holds 9 decision events. The task asks a model to state the scope after event 6; a correct answer says no version is cleared yet, because the disclosure condition is still open.
What happens when internal sign-off and the platform disagree?
Internal approval does not decide whether a platform will show an ad. Google Ads, for example, shows an ad as under review, eligible, eligible (limited) or disapproved; a limited ad can run but may be restricted by location, age or device under policies such as healthcare or alcohol, and a disapproved ad will not show until it is fixed or a decision is appealed.
Record the platform's decision as one more party on the timeline, with the status and reason the platform gave. When the platform limits or disapproves an ad that every internal reviewer approved, the case becomes an exception worth keeping: it teaches that platform policy is a separate constraint with its own evidence. Do not rewrite the internal approvals to match the platform outcome; the disagreement is the lesson.
A counterexample to keep: a campaign marked complete whose outcome field is empty. Without a dated record of what ran, the case shows decisions but not results, and it should be labeled incomplete.
Which records should be held back, narrowed or sent to a rights reviewer?
Stop and narrow when the record set includes material the company does not control. Approval metadata, such as roles, decisions, rules cited, versions and dates, is usually the agency's or brand's own business record. The creative assets, scripts with unreleased claims, talent agreements and client briefs often belong to someone else or carry confidentiality terms.
Narrowing usually preserves most of the value. A timeline that keeps roles, decision types, rules cited, condition types and durations, with the creative replaced by an identifier and a short neutral description, still teaches the multi-party reasoning. A reviewer records which of the rules below was applied to each campaign so the remaining set can be explained.
- Remove reviewer names and personal contact details; keep the role.
- Hold back creative files, recordings of people and licensed music unless the rights owner and the agreements allow the intended use; keep identifiers and rights conditions instead.
- Treat client briefs, unreleased products and unpublished substantiation as confidential client material that needs the client's permission.
- Route to a rights reviewer any campaign with talent, minors, union performers or third-party footage, and any campaign that drew a regulator or self-regulatory inquiry.
- Exclude campaigns where the version history cannot be reconstructed, because decisions cannot be tied to versions.
Can an agency or brand team license its approval histories?
An agency or in-house team that has kept approval timelines in proofing, project and traffic systems may hold a record set AI developers find useful for multi-party reasoning tasks. Access to a client's assets is not ownership of them, privacy review of reviewer data is a separate question, and redacting names does not create a right to license. Counsel reviews the client and talent agreements before any record set is offered.
If counsel clears them, those timelines can be licensed through SourceX, which qualifies the company and carries the records through rights review, contracting, preparation, delivery and payment. A creative operations lead can test eligibility with a short questionnaire instead of a file transfer, and no timeline leaves the agency until its company has approved the scope and terms and signed.
Frequently asked questions
What is advertising approval workflow data?
It is the record of who reviewed each version of an ad, what they decided, which rule or evidence they relied on, and what conditions they attached, along with the final scope and what actually ran. It lives across proofing, project, clearance and platform systems.
What should the approval timeline and rule evidence dataset brief include?
One row per decision event with the campaign and brief version, asset identifier and version, party role, decision type, rule applied, evidence link, rights conditions, timestamp and sequence, a derived final scope, and a dated outcome. Conditional and deemed approvals need their own decision types.
Does an approval carry over when the asset is revised?
Only if the reviewer says so. Recording the version on every decision lets a dataset show when a revision reopened sign-off and which parties had to decide again.
Can an agency license campaign approval records?
Often the approval metadata is the agency's business record, but the creative assets, client briefs and talent materials are frequently owned by or confidential to others. Whether anything can be licensed depends on the client and talent contracts, which counsel reviews.
How should platform ad review be recorded?
As one more party on the timeline, with the platform's status and reason, such as eligible, limited or disapproved. Keep internal approvals as they were; a mismatch between internal and platform decisions is a useful case, not an error to correct.
Sources
- The FTC says advertisers must have a reasonable basis for their claims before an ad runs. Source
- The FTC says a material connection between an endorser and an advertiser should be clearly and conspicuously disclosed. Source
- Google Ads review statuses include under review, eligible, eligible (limited) and disapproved, and limited ads may be restricted by location, age or device. Source
- A commissioned work is a work made for hire only if it falls in a listed category and the parties sign a written agreement saying so. Source
- Ad-ID describes itself as the industry standard identifier for advertising assets across media platforms. Source
- The IPTC Photo Metadata Standard defines fields including Copyright Owner, Licensor and Model Release Status. Source
See if your company qualifies
A short company assessment. No data uploads are needed.