Building in Public

Chapter 12

Learning from the Best

CampaignForge AI · May 2026 · SimQuant LLC

CampaignForge AI - The Journey

Chapter 12: Learning from the Best

Date: 2026-05-10 Vertical: Performance Marketing SaaS | Budget: $500/month

Where We Left Off

Chapter 11 gave CampaignForge a memory loop.

The system can now store campaign outcomes, retrieve similar prior examples, and use that context before the Strategist and Creative agents write their next recommendations. That solved one half of the cold-start problem: once CampaignForge has run campaigns, it can learn from itself.

But the first real campaign still starts with an empty internal memory store.

The obvious next question is whether CampaignForge can learn from public market intelligence before it has its own verified performance data. Not by copying competitors. Not by scraping private tools. Not by pretending public benchmarks are proof. But by converting public, observable patterns into structured priors the system can use.

Chapter 12 starts that work.

The first company studied: Benly.ai.

Why Benly

Benly is close enough to matter and different enough to learn from.

It describes itself as an AI operating system for performance marketing. Its public positioning is built around unifying ad data, generating winning creatives, and producing AI-powered insights for modern marketing teams. The homepage frames the product around Meta, TikTok, Google Ads, GA4, creative intelligence, reporting, and a copilot that answers questions about performance data.

That makes it useful as a reference point for CampaignForge.

Not because CampaignForge should become Benly. It should not. Benly is positioned as an intelligence layer for connected ad accounts: connect platforms, analyze performance, generate insights, produce creatives, report across channels. CampaignForge is an operator-gated campaign creation and execution system: start from a landing page and brief, audit readiness, build strategy, create assets, prepare launch, monitor performance, diagnose failures, and route rework.

But the overlap is real:

If CampaignForge is going to compete in the same conversation, it should learn the vocabulary of the category.

What We Can Learn Automatically

The first pass over Benly's public pages produced seven structured categories CampaignForge can learn from.

1. Market category

Benly is clearly B2B performance marketing software. The public pages speak to marketing teams, agencies, D2C brands, in-house teams, freelancers, and growth teams. That category signal matters because it is exactly what we fixed in CampaignForge after TrainrAI: the system must not assume every page is B2B, but when a page is B2B it should recognize it quickly.

Benly gives us clean examples of B2B performance-marketing language.

2. Offer and funnel pattern

The funnel is standard B2B SaaS:

This is useful for landing-page analysis. If a B2B paid-media SaaS page has no demo path, no trial path, no pricing signal, and no sales contact, CampaignForge should flag that. If a consumer mobile app page has no demo form, that should not be a penalty. The Benly study reinforces that CTA recommendations must follow market context.

3. Feature taxonomy

The product modules are easy to map:

This becomes a taxonomy for competitor profiling. CampaignForge can compare a landing page against the expected feature language of the category without copying anyone's language.

4. Trust signals

Benly uses public trust markers: customer count, managed ad-spend scale, rating, retention, support promises, contact information, location, legal pages, privacy terms, and data-deletion language.

These are useful because Agent 00 already audits trust signals. A B2B SaaS page for cold paid traffic should usually include some combination of proof, contact, privacy, security, pricing, customer logos, case studies, or operational credibility. A page without those signals may be technically healthy and still not ready for paid acquisition.

5. Pricing architecture

Benly's public pricing communicates several category norms:

CampaignForge should not blindly recommend pricing pages for every product. But for B2B SaaS pages, pricing clarity is part of conversion readiness. If pricing is not shown, the page should compensate with a strong demo CTA and high trust.

6. Messaging patterns

The abstract patterns are valuable:

The words themselves are not the point. The reusable structure is.

CampaignForge can store these as hook categories:

7. Competitive graph

Benly publicly compares itself against tools like Foreplay, Motion, and Weavy and targets agencies, D2C brands, and in-house teams. That gives CampaignForge a starting graph for the paid-media intelligence category.

The graph is not a verdict. It is a map: which competitors exist, what feature clusters define the category, which alternatives buyers may already know, and what objections are likely to appear.

What We Should Not Learn

This boundary matters.

CampaignForge should not copy Benly's design, copy, assets, or interface. It should not use Benly's service to build a competing product. It should not reverse engineer private app behavior. It should not claim access to data it does not have.

Public pages are fair competitive intelligence. Private product behavior is not.

The right artifact is not "Benly clone." The right artifact is a structured competitor profile derived from public material, source-linked, timestamped, and treated as a market prior.

The Competitor Profile Artifact

The next memory object CampaignForge needs is not a campaign outcome. It is a public competitor profile.

Initial shape:

{
  "source_url": "https://benly.ai/",
  "source_type": "public_website",
  "observed_at": "2026-05-10",
  "company": "Benly",
  "category": "B2B performance marketing SaaS",
  "positioning": "AI operating system for performance marketing",
  "icp": [
    "agencies",
    "D2C brands",
    "in-house marketing teams",
    "freelancers and small teams"
  ],
  "primary_ctas": [
    "free trial",
    "demo"
  ],
  "feature_clusters": [
    "creative generation",
    "AI copilot",
    "ad library",
    "creative vision",
    "unified reporting",
    "cross-channel integrations"
  ],
  "trust_signals": [
    "customer count",
    "ad spend managed",
    "rating",
    "retention",
    "support promise",
    "privacy and deletion terms"
  ],
  "pricing_pattern": "tiered SaaS with free trial and enterprise path",
  "hook_categories": [
    "tool consolidation",
    "creative fatigue",
    "cross-channel reporting",
    "AI-powered insights",
    "creative scale"
  ],
  "allowed_use": "market prior for strategy, audit, and creative ideation",
  "disallowed_use": "copying assets, copy, private workflows, or reverse engineering"
}

This object can live beside the Performance Memory RAG, but it should not be treated as verified campaign performance. It is not ROAS data. It is not proof that a hook works. It is category intelligence.

That distinction should be explicit in state:

{
  "is_real_performance_data": false,
  "is_public_competitive_intelligence": true,
  "confidence": "market_prior"
}

How CampaignForge Should Use It

The profile can support five parts of the graph.

Agent 00: Website Auditor

The auditor can compare a page against category expectations.

For B2B performance-marketing SaaS, it should look for:

For consumer apps like TrainrAI, those expectations change. The correct CTA is download/install. The correct trust signals are platform fit, privacy, support, and use-case clarity. The Benly profile is useful partly because it shows what B2B looks like, making it easier to avoid applying B2B rules to consumer pages.

Brief Prefill

When a landing page resembles the Benly category, the prefilled brief should not say "General." It should infer something closer to:

B2B performance marketing SaaS for agencies, D2C brands, and in-house teams.

When it resembles TrainrAI, it should infer:

Consumer fitness app for iPhone and Apple Watch users.

That is the smarter behavior we started adding after the TrainrAI audit.

Brief Analyst

The analyst can use public competitor profiles to ask better questions:

Strategist

The Strategist can use the category profile as a targeting prior:

This should be a starting point only. The campaign's own landing page, budget, and product claims still control the final strategy.

Creative

Creative can turn the hook categories into tests:

Again: not copied lines. Categories of persuasion.

The Bigger Lesson

The best systems in this market are not just automating tasks. They are compressing media-buying judgment into workflows.

Benly's public story is coherent:

  1. 1. Connect the platforms.
  2. 2. Analyze the data.
  3. 3. Generate or act on recommendations.

CampaignForge's story needs to be just as coherent:

  1. 1. Audit the landing page.
  2. 2. Build the campaign.
  3. 3. Get operator approval at every risk point.
  4. 4. Launch.
  5. 5. Diagnose performance.
  6. 6. Learn into the next campaign.

The difference is the control plane. Benly's public positioning is intelligence for media buyers. CampaignForge's positioning is an operator-gated agent system that can assemble and improve the campaign itself.

That distinction needs to appear everywhere:

If the system only says "AI for ads," it collapses into the category noise. If it says "audited, approved, launched, diagnosed, and learned," it has a sharper claim.

What Got Added From This Study

The first set of findings:

The immediate product implication:

CampaignForge needs a competitor_profile artifact and a public-intelligence memory lane separate from real campaign performance memory.

Performance memory answers:

What worked in campaigns we have run?

Competitive intelligence answers:

What does this market publicly reward, claim, price, and prove?

Those are different signals. The system should use both.

Sources Used

These are public pages. The findings above are paraphrased market observations, not private product analysis.

What Comes Next

Turn this chapter into code.

The next implementation step is a small competitor_profiles table or JSON store. It should support:

Then wire retrieval into the same places that already use performance memory:

The first profile should be Benly.

Not because Benly is the product to copy. Because Benly is a clear public example of the category CampaignForge has to compete inside.

The system should learn that category before it spends another dollar trying to enter it.

This post was drafted by AI and reviewed by the operator. Content is published as part of the CampaignForge AI build-in-public journey.
All chapters SimQuant LLC