Why Average Metrics Lie: A D2C Case Study

Why Aggregate Metrics Hide What Matters

Laurence Debroux, former CFO of Heineken, famously stated that “revenue is vanity, profit is sanity, and cash is king.” While revenue impresses and profitability matters more, what ultimately determines survival is cash flow. The same concept applies beyond finance. When businesses fail to identify the metric that actually reveals their health, they end up celebrating numbers that look good while the fundamentals quietly deteriorate.

Many companies track performance using aggregate indicators such as Monthly-Recurring-Revenues (MRR), daily customer counts, or growth in sign-ups. These metrics can appear healthy and reassuring. Yet they frequently hide how customer behavior evolves beneath the surface. A business can grow its revenue metrics while the quality of its customers quietly deteriorates, only to discover later that profitability and cash generation are under pressure.

Consider a simple bakery, where rising daily revenue may suggest strong demand. However, if that growth is driven by overproduction that increases waste, higher revenue may be followed by a lower gross profit. Without understanding who is buying and when, management risks focusing on the wrong objective: maximizing daily sales rather than minimizing waste.

Aggregate metrics such as those discussed above fall short because they combine customers with very different buying histories. What ultimately determines a healthy business is not growth in isolation, but how customer behavior evolves over time.

How Analyzing Customer Behavior Reveals What Aggregates Hide

The objective of any business is to generate profit, but profits are maximized more effectively when customer behavior is well understood to support efficient decision-making. While qualitative insights remain valuable, businesses operating at scale must rely on quantitative methods to observe how customers behave over time.

Moving beyond aggregate metrics requires reframing business problems through a customer-behavior lens. Consider a bakery struggling with waste. Daily revenue may be growing, yet margins are shrinking due to overproduction. The typical response is operational: adjust baking schedules or reduce batch sizes. The analytical solution, however, is behavioral. Customers can be segmented by purchase patterns. Morning commuters buying coffee and a croissant behave differently from afternoon families purchasing multiple pastries. Weekday demand also differs from weekend demand. By tracking sell-through rates by customer segment and time window, the bakery discovers that waste is not a production problem but a demand-forecasting problem. The solution is therefore not reactive operational adjustments, but customer-centric production planning that matches output to predictable behavioral patterns.

The same logic applies to e-commerce. Instead of celebrating aggregate revenue growth or total customer counts, businesses need to ask customer-centric questions: Are customers returning, or are we constantly replacing them? Are newer customers more profitable than earlier ones? Is growth building compounding value or merely masking churn? Answering these questions requires moving beyond aggregate totals to analyzing customer-level patterns, whether through cohort analysis, experimental methods, or econometric modeling that isolates the behavioral drivers of retention, referral, and long-term value.

The remainder of this article applies this perspective to a simulated D2C e-commerce brand to illustrate how standard executive dashboards can show healthy growth while obscuring the customer behavior patterns that determine whether that growth is sustainable. Such dashboards lead management to make strategic decisions based on signals that appear data-driven but are actually incomplete.

The Simulated Business Scenario: Inside a D2C Pet Brand's First Year

To illustrate how aggregate metrics mislead, I simulated 12 months of daily transaction data for a D2C brand selling premium coffee wood dog chews (€12.79 per unit) across Europe from a German base. The simulation reflects realistic D2C dynamics: 25-40% of orders receive discounts, with seasonal spikes during events like Black Friday. Customers who reorder typically do so 30-60 days after their first purchase. Delivery times follow standard fulfillment patterns (1-3 days), with a small share of orders experiencing delays or requiring refunds.

Unit Economics: Variable cost of €11.59 per order yields a contribution margin of approximately €1.20 per order under a cost-plus pricing strategy.

Note: This simulation is designed to demonstrate analytical pitfalls in aggregate metrics, not to represent any specific company. Full methodology and code available on GitHub.

From Data to Insight: How This Analysis Is Structured

The analysis examines performance across three dimensions that such businesses naturally track:

• Acquisition Channel: Organic (unpaid search/direct traffic), Paid Search (search engine advertising), Paid Social (social media advertising), and Referral (third-party referrals)
• Geography: 24 European countries
• Customer Type: First-time buyers vs. repeat customers

The dashboards are organized into three views that reflect how most businesses assess performance:

1. Static Views: P&L and Acquisition

The first dashboard presents a simplified profit and loss summary segmented by channel, country, and customer type, showing total revenue, costs, and profit contribution. The objective is to assess overall financial performance: which segments appear profitable and whether the business is growing healthily.

The second dashboard breaks down order volumes and customer acquisition across the same dimensions. The objective is to identify which channels, countries, and customer types are driving growth, providing a basis for resource allocation and marketing decisions.

These views reflect how performance is typically assessed through averages and totals. While they suggest what is happening within the business, such as which channels appear profitable or which markets generate the most orders, they rarely explain why it’s happening, which underlying drivers require action, or whether the patterns are sustainable.

2. Growth Metrics: MoM Growth Rate

This dashboard introduces a time dimension, focusing on how performance evolves month over month. Specifically, it examines the development of monthly revenue across acquisition channels, countries, and customer types. This layer highlights growth dynamics that are not visible in static summaries, including divergence between segments and changes in momentum over time.

3. Delivery Performance by Country & Customer Type: Delays, Failures & Refunds

This dashboard summarizes operational outcomes, such as delayed deliveries, failed deliveries, and refunds, by country and customer type over the 12-month period. These metrics are commonly reviewed when businesses evaluate service quality and customer retention.

The intent is not to diagnose operational problems, but to illustrate how such indicators are often interpreted in isolation, without considering their interaction with customer acquisition patterns and growth composition.

1. Static Views: P&L and Acquisition

a. P&L Charts: Revenue, Cost and Profit

Examining the simplified income statement of the e-commerce startup, social media advertising appears to generate the highest gross profit, followed by the organic channel. In contrast, Paid Search appears comparatively less attractive due to its higher costs relative to revenue.

From a geographic perspective, excluding Germany, which represents over 60% of total demand, countries such as Greece, Latvia, and Spain appear to contribute higher profits. This could lead the founder to conclude that these markets deserve increased attention and investment.

Breaking results down by customer type further reinforces this optimistic picture. The majority of total revenue and profit comes from first-time customers, suggesting that paid social campaigns are effective at attracting new demand and driving initial purchases. Overall profitability is positive, reinforcing the impression that the business model is healthy and scaling successfully.

However, this interpretation relies entirely on aggregate outcomes. What remains unclear is how sustainable this performance is. The dashboard does not reveal whether customers return, how demand is distributed over time, or whether growth is driven by repeat behavior or continuous acquisition of new customers. As a result, the dashboard appears informative while quietly omitting critical drivers of long-term business health.

b. Acquisition Metrics: Customers, Orders & Channel Breakdown

Turning to order inflows and volume, a similar pattern emerges. Countries such as Greece, Latvia, and Spain again show relatively high order activity, consistent with their strong contribution to profit observed in the P&L view. At the acquisition-channel level, Paid Social accounts for approximately 36% of total orders, while referrals represent less than 10%, reinforcing the perception that paid channels are the primary growth engine of the business.

Looking more closely at customer behavior, an interesting pattern appears. First-time customers account for roughly 6,234 customers, yet they purchase a total of 8,105 products. This implies that many first-time customers place multiple-unit orders, suggesting early engagement and initial product acceptance.

At first glance, this could be interpreted positively: customers seem willing to buy multiple units per order, and acquisition efforts appear effective. However, this observation immediately raises a deeper question. If first-time customers are purchasing multiple units, why does repeat customer volume remain comparatively low? Are customers returning later, or is demand being driven almost entirely by the continuous acquisition of new buyers?

At this stage, the question remains unresolved. Order volume alone cannot distinguish between healthy repeat behavior and short-term demand spikes driven by promotions or acquisition intensity. To understand whether the business is building durable customer relationships or merely replacing churn with new customers, a time-based view of performance becomes necessary.

2. Growth Metrics: MoM Growth Rate

Looking at the Month-over-Month (MoM) revenue growth, performance appears highly volatile across countries, acquisition channels, and customer types. This volatility is visible both in the country-level map, which dynamically displays MoM growth by month, and in the bar charts segmented by acquisition channel and customer type.

A particularly notable pattern emerges toward the end of the observation period. Across all dimensions, the final months show a sharp and synchronized decline in revenue. This raises two immediate questions. First, why does a relatively stable consumer product, such as a dog chew, exhibit such pronounced month-to-month volatility? Second, what explains the abrupt drop in sales during the final months?

Examining acquisition channels provides partial insight. Large positive growth spikes are concentrated around Paid Social and Paid Search, particularly in March 2025 and July 2025 for social media advertising, and March 2025 and October 2025 for search advertising. Between these peaks, however, growth in paid social is frequently negative, indicating that demand quickly weakens once advertising intensity declines.

While several explanations are possible, such as temporary reallocation of marketing budgets or campaign experimentation, the broader pattern suggests that revenue growth is strongly dependent on paid acquisition rather than sustained customer demand. This interpretation is reinforced by the behavior of the referral channel, which shows consistently low contribution to growth and increasing volatility toward the end of the period.

Taken together, these dynamics suggest that the business is effective at generating short-term revenue spikes through advertising but struggles to maintain momentum once acquisition slows. Growth appears driven by repeated inflows of new customers rather than stable repeat purchasing. This raises a critical question for long-term sustainability: why are customers not returning, and what is preventing revenue from compounding organically over time?

3. Delivery Performance by Country & Customer Type: Delays, Failures & Refunds

The first hypothesis is that customers may not be returning due to issues with fulfillment and delivery. To explore this possibility, it is intuitive to summarize delayed deliveries, failed deliveries, and refunds across countries and customer types over the 12-month period.

From a country-level perspective, Hungary, Spain, and Denmark exhibit the highest share of delayed deliveries relative to total deliveries within those countries. Portugal shows the highest share of failed deliveries, while Italy and Croatia record the highest refund rates. At face value, these figures may suggest operational weaknesses concentrated in specific markets.

Looking at customer types, first-time customers account for a disproportionately large share of observed issues: approximately 78% of all refunds, 84% of failed deliveries, and 79% of delayed deliveries. When viewed in isolation, this might lead to the conclusion that first-time customers experience systematically worse service, potentially explaining the weak repeat purchase behavior observed earlier.

However, this interpretation changes once the metrics are properly normalized. When delayed, failed, and refunded deliveries are calculated as a share of total deliveries per customer type, the differences between first-time and repeat customers largely disappear. Both groups experience similar delivery issue rates relative to their total order volume. This indicates that the higher absolute number of issues among first-time customers is not by worse fulfillment outcomes but by volume effects.

In other words, while delivery issues exist and may contribute marginally to customer dissatisfaction, they do not sufficiently explain the low referral rates, weak repeat behavior, or the high dependency on paid acquisition channels observed earlier. The dashboard appears informative, yet it still fails to answer the most important question: why customers do not stay, refer, or return, despite purchasing a durable product with clear subscription-model potential.

At this point, a founder relying solely on such dashboards may be tempted to draw qualitative conclusions, conduct a handful of customer interviews, or make intuitive operational adjustments. But without a clearer understanding of how customer behavior evolves and how different cohorts behave after acquisition, even the customer interviews become generic, which still doesn’t answer the underlying issues of customer retention.

This is precisely where more rigorous, product, and behavior-driven analysis becomes necessary to uncover the underlying drivers of retention, referral, and long-term value.

Conclusion

The dashboards examined here reflect how most businesses track performance: through aggregate totals, growth rates, and operational summaries. They provide reassurance, revenue is growing, margins are positive, acquisition is scaling, while obscuring the behavioral dynamics that determine whether that growth compounds or collapses.

To understand why customers aren’t returning in this simulated business would require shifting from aggregate metrics to customer-level analysis and behavioral analytics.

These analyses often reveal uncomfortable truths. A business celebrating 30% month-over-month revenue growth might discover that 90% of customers never make a second purchase, that paid acquisition costs exceed customer lifetime value by month three, or that the most “profitable” channels are simply the fastest at burning through addressable demand.

The gap between what dashboards show and what customer behavior reveals is where strategic decisions either create durable value or quietly destroy it. For businesses serious about building sustainable growth, the question isn’t whether to track aggregate metrics; it’s whether to trust them as the foundation for decisions.