INDUSTRY FIELD GUIDE
Marketing consulting campaign measurement and claim evidence for AI
A campaign can show a rise in inquiries during a test period while leaving its causal impact uncertain.
NAICS 541613 · Marketing Consulting Services
A campaign can show a rise in inquiries during a test period while leaving its causal impact uncertain. FTC guidance describes the need for substantiation of objective advertising claims. A consultant can build an original brief that tests a model’s ability to preserve metric definitions and evidence limits. FTC advertising substantiation policy
Make two dummy funnel reports. One counts unique inquiry forms and the other counts submission events, including repeats. Add a campaign memo that attributes the entire change to a new channel. Ask which comparisons are valid and what evidence would be needed for the claim. Score identification of the numerator and reporting window.
Use synthetic aggregates and your own definitions. Keep audience identifiers, pixels, client budgets and campaign credentials out.
Describe an original marketing evidence sample
Original Fiund exercise
More submission events do not establish more unique inquiries or causal lift
Two dummy funnel reports use synthetic aggregates and original definitions. No audience IDs, pixels, client budgets, account credentials or actual campaign appears.
| Record | Source | What it records |
|---|---|---|
| MK1 | Practice period A report | 100 unique inquiry forms, deduplicated by the exercise rule. |
| MK2 | Period B export | 140 submission events, including repeats. Unique inquiry count and cross-period deduplication absent. |
| MK3 | Draft channel memo | Attributes a 40% inquiry increase entirely to new channel C. No controlled comparison or source-attribution evidence supplied. |
Question for the model
What reported counts can be stated, and does the packet establish a 40% rise in unique inquiries caused by channel C?
Show the reference answer and checks
MK1 reports 100 unique inquiries in A, while MK2 reports 140 submission events in B under a different counting basis. Without B's deduplicated inquiry count, their difference does not establish a 40% unique-inquiry increase. MK3 also supplies no design isolating C's causal effect. Preserve counts, definitions and periods, and flag both the metric-comparability and causal-attribution gaps rather than repeating the memo's claim.
- Comparing unique entities with repeat-including events as the same metric.
- Attributing an observed period difference entirely to one channel without supporting design or evidence.
Fields to preserve
- Metric numerator and deduplication rule
- Defines what the count represents.
- Reporting period
- Keeps the time window explicit.
- Channel attribution and comparison design
- Separates descriptive movement from causal claims.
Test another case
Supply B's unique count but still no causal design. Require the model to resolve the descriptive change while preserving the attribution limitation.
Use synthetic aggregates and owned definitions. Exclude people, pixels, budgets and credentials.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.