Customer data and the metrics that matter
Structure transaction and event data, compute the core marketing metrics, and avoid the traps of averages, vanity metrics and dirty data.
Marketing analytics starts with behavioural data: who bought what, when, for how much, and which marketing they saw before. Surveys tell you what customers think; behavioural data tells you what they do. Before any model, you need the data in the right shape and a clear view of the metrics that matter.
The three shapes of marketing data
- Transaction (order) data: one row per order, with customer ID, date and amount (and optionally product lines). This is the basis for RFM, cohorts, CLV and basket analysis.
- Event or journey data: one row per touchpoint (visit, email click, ad view) with customer or session ID, channel and time, plus conversions. This is the basis for attribution and funnel analysis.
- Aggregated time series: one row per week or month, with sales, media spend, price and other drivers. This is the basis for forecasting, marketing mix models and causal impact analysis.
Most of the tools in this course accept a simple pasted table: for RFM, for example, three columns (customer ID, order date, order value). Export them from your shop system, CRM or analytics platform.
Core metrics
| Metric | Definition | Example question |
|---|---|---|
| Conversion rate | Conversions ÷ visitors (or sessions) | Did the new checkout convert more visitors? |
| Average order value (AOV) | Revenue ÷ orders | Are bundles raising basket size? |
| Purchase frequency | Orders ÷ customers in a period | Do loyalty members buy more often? |
| Retention rate | Share of customers who buy again in a later period | How many first-time buyers come back? |
| Customer acquisition cost (CAC) | Marketing spend ÷ new customers | Which channel brings customers most cheaply? |
| Customer lifetime value (CLV) | Expected future profit from a customer | How much can we spend to acquire one? |
| Return on ad spend (ROAS) / ROI | Revenue (or profit) attributed ÷ spend | Which channel deserves more budget? |
The single most useful ratio in customer analytics is CLV ÷ CAC: a healthy business acquires customers for well under their lifetime value (a ratio of 3 is a common rule of thumb).
Averages lie
Customer data is extremely skewed. In the online shop example used in the next lessons, 520 customers placed 3,775 orders worth 272,231 in total, an average of 524 per customer. But just over a quarter of the customers generate more than half of the revenue, and many bought only once or twice. The “average customer” describes almost nobody. That is why the following lessons segment customers (RFM), follow them over time (cohorts) and model them individually (CLV).
Vanity metrics versus decision metrics
Page views, followers and impressions are easy to count and easy to grow, but they rarely drive decisions. Prefer metrics that connect to revenue and that you can act on: conversion, retention, CLV, incremental sales. For each metric, ask: if this number changed, what would we do differently?
Data quality checks
- Customer IDs: guest checkouts and duplicate accounts split one customer into several; merge on e-mail or a stable ID where possible.
- Returns and cancellations: remove or net them out, or they inflate revenue and frequency.
- Test orders and staff accounts: exclude them.
- Dates and time zones: use one consistent format; the tools read the common formats.
- Outliers: a few wholesale customers can dominate averages; analyse them separately.
The data cleaning tool helps with duplicates, text-to-number conversion, dates and outliers before you paste data into the analytics tools.
Correlation is not incrementality
The hardest question in marketing analytics is causal: did the campaign cause the extra sales, or would they have happened anyway? Customers who see more ads are often customers who were going to buy anyway. Keep this question in mind throughout the course; Lessons 6 and 9 show the methods that answer it (experiments and quasi-experiments).
Key takeaways
- Know the three data shapes: transactions, events or journeys, and aggregated time series.
- Track decision metrics: conversion, AOV, frequency, retention, CAC, CLV and ROI.
- Customer data is skewed; segment and model instead of relying on averages.
- Clean IDs, returns and test orders first, and separate correlation from incremental impact.
Check your understanding
Answer each question to see the explanation. Finishing the quiz marks this lesson as complete.