A Blueprint for a Sustainable Marketing & Analytics Engine

---

For eight years, I ran a marketing consultancy. While I specialized in SEO and Content, I was inside dozens of businesses, and I kept seeing the same story. A company would hire me to fix a symptom, like low traffic, but I’d immediately see deeper issues: their tools disconnected, their analytics were a mess, they’d have critical bottlenecks in their systems, etc.

The incentives were often misaligned. Companies usually hired me to execute specific tactics, not to fix their infrastructure. Even when I identified the deeper data leaks, the scope of work (and the urgency for immediate traffic) rarely allowed for a full systemic rebuild.

But I was always aware that just “driving more traffic” to a leaky funnel wasn’t going to move the needle the way they hoped. In the back of my mind, I was mentally re-architecting their systems. Over those eight years, I got much better at designing those systems (for myself as well), gaining insights into what works and what doesn’t.

With the rapid rise of generative AI, however, I saw the writing on the wall. The tasks clients had paid me for, like writing standard SEO content, were quickly becoming commodities.

I realized my biggest advantage moving forward wasn’t just fixing one-off problems, but in my ability to build these types of big picture engines that businesses can operate and depend on. From my perspective, in this new world of infinite, AI-generated noise, a key durable advantage is a system that masters first-party data, proves attribution, and delivers a personalized experience.

This would finally let me solve the “bigger picture” problems. I realized I needed to evolve from the specialist hired to fix a specific leak, to the architect responsible for the integrity of the entire structure.

So, I closed my agency. I’m now building GetViajo.com as my personal R&D lab to build that exact system from a blank slate. It serves as a live production environment for my methodology. It allows me to demonstrate exactly how I architect a modern marketing foundation when given the freedom to prioritize data integrity from day one. (Explanation: I recently moved to Portugal and needed to learn Portuguese, so this project is both personal and professional).

This document is the technical breakdown of that engine. It details the system architecture, explains the business logic behind the design decisions, and connects each component to a tangible outcome.

My “Foundation First” Methodology

Most startups are forced to build their airplane while it’s in mid-air. They duct-tape analytics and content pipelines together in a rush to find product-market fit. By the time they scale, they’re drowning in technical debt and data they can’t trust. This is the type of chaos I would see constantly in my consulting work.

Given the opportunity to build from a blank slate, I made a deliberate choice to architect the foundational layer correctly from the start. The goal wasn’t to build a perfect, rigid system that would never change but rather a clean foundation that enables future adaptability, avoiding that initial mountain of technical debt.

It’s far easier to add a new tool or feature when your core data is already clean, reliable, and flowing correctly.

The system automatically storing the user’s first_touch_attribution from LinkedIn and tracking their behavior_score (currently at 10) as they engage with the site.

The purpose was to ensure that from the very first visitor, every data point would be clean, every attribution reliable, and the business would be ready to scale as intelligently as possible. This blueprint is the result of that line of thinking.

[!tip] Author’s Note I recognize that building from a blank slate is a unique opportunity.

The challenge in an established business is rarely about tearing everything down. It is about surgically applying these principles to existing infrastructure.

The work involves architecting a single, reliable source of truth for attribution or refactoring a leaky pipeline. The objective remains the same: ensuring the data flowing into your decision-making tools is actually accurate.

The Core Challenge: Building a Cohesive System

My goal with GetViajo.com was to avoid the long-term headaches I saw companies struggling with as a consultant. This meant deliberately architecting a cohesive system from the start, as opposed to just individual pieces; strategically planning out how each would play nicely together, being as utilitarian as possible.

When I looked at the full picture, I saw four core, interconnected problems that many modern businesses face:

1. The Data Problem: Many companies rely on a patchwork of third-party tools (analytics, CRM, email) that don’t talk to each other. This creates data silos and conflicting reports, making it impossible to get a single, reliable answer to “What’s working?”
2. The Content Problem: In the new age of AI, how do you produce content at scale without sounding like a robot or, even worse, publishing false information and damaging your brand’s credibility?
3. The Engagement Problem: Traffic without intent signals is just noise. I needed a reliable mechanism to automatically distinguish casual browsers from high-intent users, and then use that flag to drive targeted promotions through ads and email.
4. The Revenue Problem: Since I’m selling digital products, how do I ensure only paying customers get access and prevent casual link-sharing from eating into revenue?

These aren’t four separate problems; they are four sequential stages of a single, continuous customer journey. Solving one (like data) is useless if your content is unreliable or your revenue is unprotected.

The Solution: A Four-Part System for Clarity

For GetViajo.com, I designed and built-out a four-part system where each component solves one of these core problems, creating a single, logical flow of data from the first visit to the final, secure purchase.

The complete system is designed to create a clear, linear flow from accurate data capture all way to secure product delivery.

Part 1: Architecting a Single Source of Truth

To solve the data problem, I had to take full control of my most critical data hand-offs. My approach wasn’t to replace my analytics tools, but to control the quality of the data they receive.

• The Fix: I moved critical data hand-offs, like email tagging and consent tracking, from client-side scripts (which are prone to being blocked by ad-blockers or misfiring) to my own server-side API endpoints. This gives me top-down control of the entire data pipeline.
• The Result: I can be highly confident in my performance metrics. This doesn’t replace my analytics platforms; it just ensures the data fed into them is reliable. I’m not risking my attention (or budget) on bad data.

Part 2: A Content Process You Can Actually Trust

With tracking in place, I needed to lean on my content strategy chops and begin to create content. But, as we all know, generative AI has a major trust issue: it often makes things up, which is a huge risk for any business. The default “AI voice” is a brand-killer, and hallucinations are a non-negotiable liability.

My solution was to design a system that first solves for credibility, then for authenticity. It separates the research from the writing.

First, a research engine performs targeted searches, scrapes verifiable facts from live sources, and builds a research brief before the AI is allowed to write a word. This ensures every claim is grounded in verifiable, real-world data.

Second, once the facts are locked, the system uses a custom model which I fine-tuned on my entire archive of human-written articles from my consulting days to draft the content. This ensures the voice is authentic and avoids the generic, robotic tone of default AI.

• The Fix: My system first separates research from writing. The AI is only allowed to work with pre-verified facts, and it does so using a custom-trained model built on my own unique writing style.
• The Result: I can produce content at scale without risking my brand’s credibility on false information. The system isn’t just prompted to sound authentic; it is specifically architected for authenticity.

This focus on a sound, repeatable process is the same core principle I used at my consultancy to get results for clients. For example, for one B2B client, French Meadows, a new SEO-optimized website and a thorough on-page optimization strategy grew their organic traffic by 1,977% in 15 months (from 693 to over 14,000 monthly visitors).

While that was admittedly before I built this specific AI system, the methodology is the same: build a sound process, and the results follow.

Part 3: Identifying the Most Interested Users

Good data and fact-checked, well-researched content is a great start, but you still need to know what to do with it all. I needed a simple way to see which users were just browsing and which were getting serious.

• The Fix: I created a simple, points-based system that tracks a handful of key user behaviors. This flag is then passed via API to all my downstream tools.
• The Result: I can now easily (and automatically) identify my most engaged users. This allows me to focus sales efforts and run targeted promotions to the lowest-hanging fruit: users who are high-intent but not yet customers.

This system is custom-built, but the logic is simple. A 20% scroll on my premium pricing page might get a user 5 points, while opening an email is 2, and scrolling to the bottom of a blog post is another 3. Listening to a complete audio lesson is 8 points, as it’s a strong indicator of engagement with my core product.

When a user’s score crosses a certain threshold (20 points), they are automatically flagged as a hot lead.

This is the ‘outcome.’ After the same user’s score crossed the 20-point threshold, the system automatically flags them by setting ‘high_intent_identified’ to ‘true’.

The best part is that this flag isn’t just stuck in the browser. Via an API, it’s automatically synced to my other core tools, including my analytics platform and my email marketing platform, Kit (formerly ConvertKit).

The ‘high_intent’ flag is automatically sent via API to PostHog for analytics purposes (left) and to ConvertKit (right) for immediate marketing automation. I now have a user’s email address and their level of engagement with my core content.

Because the score is stored locally, the system tracks this engagement continuously. It doesn’t matter if a user hits the threshold in a single session or over many months. The system will flag them regardless, applying that ‘high-intent’ tag to their email record.

This, combined with the first-touch attribution, means I can identify not only who my most engaged users are but also which channels (like Reddit or LinkedIn) brought them in, even if it takes them months to convert.

Part 4: Protecting the Product

Finally, if you’re selling digital products, you have to make sure only paying customers can access them. To solve this, I used AWS to generate temporary, expiring download links (pre-signed URLs) for each purchase. A direct link to a file can be shared anywhere, but these links are secure and stop working after a short time.

• The Fix: Instead of direct links, my system generates secure, time-limited URLs for every digital purchase.
• The Result: It prevents unauthorized sharing and directly protects the revenue from my products.

The Outcome: From Guesswork to a Single Source of Truth

Building this four-part system moved the project from strategic chaos to operational clarity. I went from having conflicting, siloed data to having a clear, reliable view of what’s happening across the entire business.

The most important outcome is that the entire system is now fully functional. I’m extremely confident that this system can accurately track distinct channels like Reddit and LinkedIn, move users into a credible content funnel, flag them based on behavior, and securely deliver the final product to my premium subscribers.

The ‘after’ state. This dashboard shows the system working as designed, populated with data from my own testing and an initial validation post on Reddit. This test successfully proved that all attribution channels (e.g., ‘reddit-rportugal’), conversion funnels, and customer tracking are fully operational and reliable—confirming the blueprint is sound and ready for scale.

How This Applies to a Growth Team

This approach isn’t about buying a specific tool, but about thinking through the flow of data in your own business.

• Map Your Data’s Journey: Look at your own reports and ask: where does this number actually come from? Try to identify any steps in the process where you lose visibility.
• Define Your “Hot Lead” Actions: Pick one or two user actions that signal real interest (like scrolling down a certain amount on your pricing page, or an abandoned cart checkout process). Create a special segment in your email list, or send targeted ads just for these users and see if they engage differently.
• Insist on a Research Brief: Before your team creates any more content, have them fill out a simple brief with links to the core research. It’s a low-effort way to improve quality and make sure AI is used responsibly.

[!warning] Pro Tip LLMs respond really well to JSON formatting. Format your cited data in JSON, along with a url, a snippet of key information, and instruct the LLM to outline the article first using those snippets to avoid hallucinations.

Future-Facing Questions

[!faq]- How does a system like this handle data privacy regulations (like GDPR)? Because I control the entire data pipeline, I have a major advantage. I can be very precise about what data I collect and how it’s handled. This makes it much easier to adapt to new privacy rules compared to relying on a third-party vendor to be compliant.

[!faq]- What’s the biggest challenge in scaling a system like this? Keeping the data clean as you add more tools. The best way to manage this is to have a strict rule: no new tool gets connected until you can prove its data matches what your core system is reporting.

[!faq]- How could machine learning improve the behavioral scoring? The next step would be to use a predictive model instead of a simple points system. A model could analyze user behavior patterns to predict who is most likely to convert, which would be an even more powerful signal for a sales team.

If this way of thinking about marketing systems resonates with you, I’m always happy to connect with people who enjoy building things and solving these kinds of puzzles. Feel free to reach out.