Sources and Provenance

This module exists because AA 01's original module map promised it, and because a course making claims about real money deserves a place that says plainly what those claims trace back to. Read this alongside the confidence-tag definitions in AA 01 — this lesson doesn't repeat that framework, it applies it to the whole course at once.

---

#How this course was actually built, and what that means for this page

This course was built as a single research pass in August 2026, working from general web research rather than one disclosed operator's private playbook (unlike AMZ, IDS, or LUCE, which compile a named practitioner's own documents — see AA 01). That has a direct consequence for this reference page: most of the course's [Directional] claims — margin bands, churn rates, capital ranges, hiring-order guidance — were synthesized across multiple agency-consultancy, financial-modeling-vendor, and trade-press sources that were cross-checked against each other for consistency at the time of writing, but whose individual URLs were not preserved as a formal citation list during that first pass. This is stated plainly rather than papered over with a fabricated bibliography: a precise, per-claim citation list for that first research pass does not exist, and reconstructing one after the fact — searching for a source that happens to match a number already written down — would produce false confidence, not real verification. AA 01 already flags this course as "a first-pass, barebones research build," and this is the specific, honest shape that limitation takes.

What this page can do instead, and does: name every source this course cites specifically by name or publication in its own body text (the disclosed-tier sources below), name the sources behind the one lesson built with full citation discipline from the start (AA 10, added in a later pass specifically to close a gap AA 01 had promised), and give you a straight list of what's flagged [Speculative] across the whole course — the claims most worth independently re-checking before repeating them to a client or investor.

#Named, disclosed-tier sources (cited specifically in the course body)

These are sources the course names directly, not folded into "agency-consultancy sources" — treat these as the strongest evidence in the course, and the ones worth reading in full if you want the underlying detail:

• ANA (Association of National Advertisers) / K2 Intelligence, the 2016 media-transparency investigation — the definitive disclosed-operator-tier report on programmatic markup opacity (principal transactions, rebates, ad-server markups), reported via AdExchanger's direct trade-journalism coverage. Cited in AA 03.
• ANA, Q1 2025 Programmatic Transparency Benchmark — the "41% of programmatic spend resulted in quality impressions" figure, the ANA's own reported methodology and number. Cited in AA 03.
• ANA and 4A's, 2025 Client-Agency Relationship Tenure Report — the tenure-by-agency-type breakdown (7.3 years independent/full-service vs. 5.8 years holding-company; the mandatory-review-clause finding), reported via MediaPost's direct coverage. Cited in AA 10.
• 4A's, 2024 Compensation Methodologies Survey — the "72% of agencies use fixed fee as primary compensation" figure. Cited in AA 03, tagged [Directional] there because it's one of two disagreeing survey populations discussed side by side.
• US Bureau of Labor Statistics, Business Employment Dynamics programme — the general new-business survival data (≈76.8% one-year survival, ≈51.2% five-year survival) cited in AA 09 as the grounded comparison point for the untraceable "75% of small agencies fail" claim, the same primary source the SMMA course uses for the same purpose. Covers all new employer establishments, not agencies specifically — a floor for comparison, not a direct agency-failure figure.

#Sources behind AA 10 (Holding Company Disruption), added in this build's second pass

This lesson was added specifically to close a gap the original nine lessons left open — AA 01 promised the course would treat the AI-disruption question as central to Module 4, and the original module 4 (AA 08–AA 09) didn't directly address it. It was built with full source discipline from the start:

• Storyboard18, reporting on the Omnicom–IPG post-merger layoffs, cost-savings targets, and retired agency brands (DDB/MullenLowe into TBWA, FCB into BBDO) — March 2026.
• eMarketer, "Omnicom's latest report signals an agency shift from human talent to AI" and "How the Omnicom IPG merger sets a new standard for AI-driven advertising" — the 91%/73% agency-leader survey figures and the Publicis growth-rate comparison.
• AI CERTs News and VideoWeek, reporting on WPP's Elevate28 restructuring and its AI-partnership framing.
• MediaPost, direct coverage of the ANA/4A's 2025 tenure report (see disclosed-tier list above).
• Challenger, Gray & Christmas, cited via trade coverage for the 150,000-plus AI-cited US job-cut count since early 2025 (a broader labor-market figure, not agency-specific, used here only for industry-wide context).

#What's flagged [Speculative] across the whole course — the numbers to re-check before repeating

Gathered here from where each first appears, so you don't have to hunt through nine lessons to find the weakest claims:

• The "~$415,000 CAPEX / ~$840,000 total cash" fully-staffed-agency-at-launch scenario (AA 05) — one financial-modeling vendor's aggressive planning scenario, explicitly flagged as an outlier rather than a typical starting point.
• "Nearly 40% of businesses using an agency are already planning to switch within six months" (AA 04) — single-source, no traceable primary study, flagged as easy to generate from an unrepresentative survey.
• "Roughly 80% of companies now outsource at least one part of their digital media" (AA 02) — untraceable to a named study, and worth extra scrutiny precisely because it's a number that validates the business model this course teaches; caveated directly in AA 02 rather than presented as a plain fact.
• The 91%/73% AI-headcount-impact survey figures and the Publicis 6%-vs-2% growth comparison (AA 10) — both single-source-reported figures from trade coverage, not independently re-derived by this research.
• The "$21,000/month" fixed-running-cost scenario (AA 05) — added to this list in a later pass, since it turned out to be the same class of source as the $415,000-CAPEX outlier scenario immediately above it in AA 05's capital table, but had originally been given the lighter [Directional] tag rather than that scenario's [Speculative] one. Both are now tagged and caveated consistently.
• The "60% of PDF-attachment proposals never opened" figure (AA 06, referenced again in AA 07) — this specific statistic circulates heavily in interactive-proposal-software vendor marketing, a category of source with a direct commercial interest in agencies believing static PDFs underperform. Downgraded from [Directional] to [Speculative] in this pass; the underlying recommendation (use a trackable proposal format) survives independent of the number's accuracy.
• The "60–87% reduction in tactical-optimization time" and the "$50,000/month" incrementality-tool cost threshold (both AA 07) — neither traces to a named, vendor-independent source; both are the kind of figure a platform or a tooling vendor benefits from readers accepting at face value, flagged in AA 07 itself rather than only here.

#What was deliberately not re-researched for this reference pass

Per this build's own scope — fill genuine gaps, don't re-litigate content that's already well-sourced — the margin, churn, capital, and staffing benchmarks in AA 04, AA 05, and AA 08 were read and assessed as consistent, properly-tagged research, not re-verified against fresh primary sources. If you intend to rely on a specific number from those lessons for a real financial decision, re-verify it directly rather than treating its presence in this course as independent confirmation — the same standard AA 01 asks of every [Directional] figure in the course.