Traffic Manager Prover

Traffic Manager Prover MCP Connector for Claude

A+

A startup spent $180K on Meta Ads and reported ROAS 4.2x. The board celebrated. Then someone ran an incrementality test — a 10% holdout that saw no ads. 38% of 'attributed' conversions were organic users who would have purchased anyway. True incremental ROAS: 2.6x. $68K spent on people who needed no convincing. Platform-reported ROAS is fiction. This tool forces five axes: unit economics per channel, attribution integrity with incrementality testing, funnel diagnostics at every stage, creative performance with fatigue analysis, and audience architecture with saturation awareness.

1 tools Official Updated Jun 28, 2026 Official Vinkius Partner

The Problem

LLMs commit five traffic management failures:

  1. Vanity Metrics — impressions and CTR instead of CAC, LTV, payback period.
  2. Attribution Naivety — trusts platform-reported ROAS (inflated 30-60%).
  3. Funnel Blindness — knows total conversion but not where it leaks.
  4. Creative Laziness — 'good ad' instead of hook rate, fatigue curves, rotation.
  5. Audience Saturation — assumes more budget always means more results.

The 5 Traffic Axes

Axis Pivot Rule
Economics Calculated CAC, LTV, ratio ≥1:3, payback, marginal ROAS per channel.
Attribution Tested Incrementality holdout, iOS adjustment, cross-platform dedup.
Funnel Diagnosed Conversion rate at EACH stage. Biggest leak identified.
Creative Analyzed Hook rate >30%, fatigue curve, rotation schedule, format A/B.
Audience Architected Segments by CPA, reach %, frequency
traffic-managerperformance-marketingpaid-adsroascac-ltvattributionfunnel-optimization

1 tools expose this connector's capabilities to your AI agent.

validate_traffic_manager

(1) UNIT ECONOMICS — CAC (cost to acquire one customer), LTV (total revenue from that customer), LTV:CAC ratio (must be ≥ 3:1 for sustainability), payback period (months to recoup CAC), blended ROAS (total revenue ÷ total ad spend) AND marginal ROAS (revenue from last dollar spent). If blended ROAS = 5:1 but marginal ROAS = 0.8:1: you are overspending — the last dollar loses money. (2) ATTRIBUTION — model choice (last-click, multi-touch, incrementality), cross-platform dedup (platforms double-count), iOS 14.5+ adjustment (30-40% signal loss), incrementality holdout results (the only honest measure of true ad value). (3) FUNNEL — conversion rate at EACH stage (impression → click → land → lead → MQL → opp → close). Identify the biggest leak (lowest conversion between stages). Fix the leak before adding traffic. (4) CREATIVE — hook rate (>30% first 3 seconds), hold rate (50%+ watched), fatigue curve (when CTR drops 30% from peak → kill creative), rotation schedule, format A/B testing. (5) AUDIENCE — segmentation (LAL 1%/3%/5% tiers), reach % of total addressable, frequency (3-7/week optimal), saturation point (when incremental reach < 5%), exclusion lists (existing customers, converters). Structured reflection tool for expert-level performance marketing reasoning — forces unit economics rigor, attribution integrity, funnel diagnostics, creative analysis, and audience architecture before spending a single dollar. Based on the disciplines of direct-response marketing (Claude Hopkins, "Scientific Advertising," 1923), modern attribution science (Randall Lewis, Google incrementality research, 2011), and media buying mathematics (LTV:CAC ratio, payback periods, marginal ROAS). Catches Vanity Metrics (celebrating impressions instead of calculating unit economics — a hotel spends $50,000/month on social media ads. Report: "2.3M impressions! 45,000 clicks! 1,200 likes!" Manager: "Great, how many bookings?" "We do not track that." The hotel has no idea if the $50K generated $0 or $500K in bookings. Impressions are the metric you report when you do not have a metric that matters. Fix: CAC (cost per booking) = $50,000 ÷ bookings. If CAC > booking profit: you are losing money. If LTV:CAC < 3:1: your acquisition is not sustainable. The only metrics that matter: how much does a customer cost, how much are they worth, and how long until you recoup the acquisition cost?), Attribution Naivety (trusting platform-reported ROAS as truth — a furniture retailer runs ads on 3 platforms simultaneously. Platform A claims: "We drove $200K in sales." Platform B claims: "We drove $180K in sales." Platform C claims: "We drove $150K in sales." Total claimed: $530K. Actual total revenue: $300K. Every platform takes credit for the same customer. The customer saw an ad on A, clicked on B, and bought through C — all three claim the sale. Fix: incrementality holdout test. Hold out 10% of audience from each platform. Measure: what revenue do you LOSE by not advertising? That difference is the TRUE incremental value. If holdout group revenue = treatment group: the ads are taking credit for sales that would have happened anyway), Funnel Blindness (not diagnosing which stage is leaking — a cooking class business spends on ads. "We get plenty of clicks but no bookings." Funnel: impression → click → landing page → class selection → checkout → booking. Click rate: 3.2% (good). Landing page → class selection: 62% (good). Class selection → checkout: 8% (DISASTER — this is the leak). Checkout → booking: 91% (good). The problem is not awareness, not the landing page, not checkout — it is class selection. Investigation: too many options (47 classes), no filtering, no "most popular" sorting. Paralysis of choice. Fix: reduce to 12 featured classes, add filters, show reviews. Class selection → checkout: 8% → 34%. Bookings 4x. Without stage-by-stage diagnosis: the business would have spent more on ads to push more traffic through a broken funnel), Creative Laziness (running the same ad until it dies without testing rotation — a wine subscription runs one ad creative for 6 months. Month 1: CTR 4.2%, CPA $28. Month 3: CTR 2.1%, CPA $52. Month 6: CTR 0.8%, CPA $134. This is creative fatigue — the same audience has seen the same ad 12+ times. Hook rate (first 3 seconds of video): dropped from 38% to 11%. The audience has literally stopped watching. But the campaign is still running. Fix: creative rotation schedule — new creative every 2-3 weeks. Test hooks (first 3 seconds), hold rate (watched 50%+), and CPA per creative variant. Kill any creative when CTR drops 30% from peak. Minimum 3 active variants at all times), and Audience Saturation (ignoring reach and frequency limits — a local bakery targets "women 25-45 interested in baking" within 10km radius. Audience size: 12,000 people. Monthly budget: $5,000. CPM: $15. Monthly impressions: 333,000. Frequency: 333,000 ÷ 12,000 = 27.8 impressions per person per month. Each person sees the ad 28 times. After 7-8 exposures: returns diminish sharply. After 15: negative brand perception ("why do they keep showing me this?"). Fix: expand audience (remove interest targeting, go broader), reduce budget, or rotate creative aggressively to vary the message. Optimal frequency: 3-7 exposures per week. Beyond that: diminishing returns → annoyance). Call once per traffic strategy, campaign launch, or performance review

See how to talk to your AI agent using Traffic Manager Prover.

Our Meta Ads have great CTR of 3.2% and impressions are growing 40% month-over-month. We scale by increasing budget. Google Analytics shows strong attributed conversions. Our ads look professional and well-designed.

VANITY_METRICS — Five failures: CTR and impressions are vanity (where is CAC/LTV?), platform attribution untested, no funnel stage analysis, creative assessed subjectively, audience saturation ignored.

Economics: Meta CAC $42, LTV $380, ratio 1:9, payback 2.1mo, blended ROAS 4.2x, marginal 2.8x. Google CAC $38, LTV $420, ratio 1:11, payback 1.8mo. Attribution: data-driven MTA, iOS +35% modeled, holdout showed 22% organic overlap, true incremental ROAS 3.1x. Funnel: CTR 1.8%→72% land→8.4% lead→34% MQL→18% opp→22% close. Biggest leak: MQL→opp at 18%. Creative: UGC hook rate 38%, hold 62%, CPA $28, fatigue at day 14, 5 variants/week. Audience: 1% LAL $22 CPA, 3% LAL $34, interest $48. 64% reach 1% LAL at freq 4.2. Exclusions active.

TRAFFIC_PROVEN — All five axes validated at expert level.

Economics: Meta CAC $55, LTV $290, ratio 1:5.3, payback 3.8mo. Google CAC $42, LTV $310, ratio 1:7.4. Attribution: last-click via GA4. Funnel: 2.1% CTR, 12% landing→lead. No MQL breakdown. Creative: 3 static ads running for 6 weeks. No rotation. Audience: broad interest targeting, $65 CPA, no LAL, no exclusions.

ATTRIBUTION_NAIVE — Economics pass (CAC:LTV healthy). Attribution FAILS: last-click GA4 does not deduplicate, no iOS adjustment, no incrementality test. Fix attribution before trusting the numbers. Then: funnel needs full stage breakdown, creative is fatigue-dead at 6 weeks, audience needs segmentation.

Google and Meta both claim the same conversion. iOS 14.5 made 60-80% of iOS conversions invisible, so platforms model them — often generously. Incrementality tests consistently show 30-40% of attributed conversions are organic (Measured.com data). Platform-reported ROAS 4x might be 2.5x incremental. The only way to know: run a 10% holdout group that sees no ads and measure the difference.

Related Connectors