Marketing Strategy Engine

A complete marketing strategy built from real data, following a structured MBA-level framework from orientation through tactics. No dashboards, no gut feel, just 4 APIs and the discipline to follow the steps in order.

The Problem

Marketing strategy was being built on dashboards and intuition. We had data everywhere, GA4, Klaviyo, Google Ads, Shopify, but nobody was connecting it into a coherent strategic framework. Decisions were made based on whatever metric happened to be on screen that morning. We needed a proper strategy, not a campaign plan or a media brief, but the full arc from market orientation through segmentation, targeting, positioning, and finally tactics. The kind of thing you produce at a Fortune 500 company or in an MBA program, except with our actual data instead of a case study.

The Approach

I followed a structured 7-phase marketing strategy framework and refused to skip steps. Seven phases, in order: orientation, research, segmentation, targeting, positioning, objectives, tactics.

For each phase, I pulled real data from production APIs. GA4 gave us traffic patterns, channel performance, and conversion paths. Klaviyo provided subscriber records with engagement histories. Google Ads data covered campaign performance and spend efficiency. Shopify contributed order histories with full purchase patterns and product affinities.

Where real data wasn’t available (specifically customer attitude and motivation data that would normally come from surveys), I generated synthetic survey data and labeled it clearly as synthetic throughout the documents. No pretending modeled data is real.

The segmentation phase produced customer profiles grouped into behavioral clusters based on actual purchase patterns, not personas invented in a workshop. From those segments, I built Google Ads Customer Match audience CSVs from purchaser emails, so the targeting recommendations were immediately executable rather than theoretical.

Everything compiled into professional strategy documents. Fifteen documents, roughly 50 pages total, with budget allocation tied to specific segments and channels.

Under the Hood

The strategy pulls live data from four production APIs. Here’s the GA4 connection pattern:

from google.analytics.data_v1beta import BetaAnalyticsDataClient

client = BetaAnalyticsDataClient()
response = client.run_report(
    property=f"properties/{GA4_PROPERTY_ID}",
    dimensions=[
        Dimension(name="sessionDefaultChannelGroup"),
        Dimension(name="country"),
    ],
    metrics=[
        Metric(name="sessions"),
        Metric(name="conversions"),
        Metric(name="totalRevenue"),
    ],
    date_ranges=[DateRange(start_date="2024-01-01", end_date="today")],
)

Segmentation uses actual purchase data, not personas:

# Behavioral clustering from real order data
segments = {
    "high_value_repeat": customers.query(
        "order_count >= 3 and total_spent > threshold"
    ),
    "single_purchase": customers.query(
        "order_count == 1"
    ),
    "discount_acquired": customers.query(
        "first_order_discount == 'WELCOME15'"
    ),
}

Key Decisions

• Followed the framework strictly, no shortcuts. The temptation with AI is to jump straight to tactics because generation is fast. I deliberately forced each phase to complete before moving to the next. Orientation before research. Research before segmentation. The discipline of sequence matters more than the speed of output.

• Real data wherever possible, synthetic where unavailable, always labeled. Customer survey data didn’t exist. Rather than skip the segmentation inputs or fake them quietly, I generated synthetic attitudinal data and marked every instance. This lets the reader know exactly where the strategy rests on solid ground and where it rests on assumptions.

• Executable audience CSVs, not just recommendations. Most strategy documents end with “target high-value repeat purchasers.” This one ends with a CSV file you can upload to Google Ads today. The gap between strategy and execution is usually a handoff. I wanted to close it.

Scale

• 15 strategy documents, approximately 50 pages
• 4 production APIs connected and queried
• Behavioral segmentation from real purchase data
• Audience CSVs ready for immediate Google Ads upload

What I Learned

Following a structured framework with AI forced me to do every step in order, which is something I’ll admit I’ve been lazy about in the past. It’s easy to skip orientation and jump to tactics when you think you already know your market. The framework won’t let you, and the data often disagreed with what I thought I knew.

The biggest surprise was using synthetic data for the first time. We didn’t have attitudinal survey data, so I generated it and labeled it honestly. What struck me was how closely the synthetic patterns correlated with our existing assumptions about the market. That’s either validating or terrifying, depending on how you look at it. It forced a real conversation about which of our beliefs were grounded in evidence and which were just comfortable stories we’d been telling ourselves.

That’s the thing about dashboards. When you only look at them, you trust what you see. When you pull the raw data and follow a framework that forces you to check acquisition quality before discussing tactics, you learn to distrust it first. The strategy engine didn’t just produce a document. It produced a different way of looking at our own performance data.

---

Built for Hewyn, a DTC wellness brand. Framework and methodology shown. Financial targets and customer data anonymized.