A first-pass claim file for every ad

The workflow, step by step

1. 1

   Upload finished ad

   Start from the cut you intend to run, not the brief. Add the script if you have one, the product specification and the studies offered as support. The claim file is built against that exact version.

2. 2

   Detect express and implied claims

   Claims are collected from the voiceover, captions, on-screen text and imagery, each with its timestamp and source. Implied claims are those carried by pictures, context and testimonials rather than stated outright.

3. 3

   Match to product evidence

   Each claim is set beside the study or specification offered to support it, so the reviewer sees the claim and its evidence on one screen.

4. 4

   Flag mismatches

   Evidence is compared with the claim on four points: the finished product (or a single ingredient), the dose, the population studied, and the outcome measured. Each is marked Match, Partial or Mismatch.

5. 5

   Reviewer decides

   Your reviewer approves, asks for a change, or rejects, and records why. The decision is stored separately from the finding: ForgeStack flags, the reviewer decides.

6. 6

   Re-check the revision

   When the next cut arrives, it is compared against the earlier decisions: what was resolved, what is still open, and anything new.

7. 7

   Keep the audit record

   Every finding, decision, reason and revision is kept in one record per ad, showing who decided what and when.

How it looks

Illustrative demo: fictional product, ad and evidence. Findings are pre-written to show the workflow, not live model output.