The iceberg model — what we see vs. what drives behavior
Above the waterline: what customers do (purchases, clicks, returns). Below: why they do it (motivation, mood, trust, loyalty, situational context). Most analytics work happens above the waterline because that data is plentiful and structured. The real insight comes from explaining the behavior with what is below — and that requires qualitative work, not just dashboards.
ABC analysis — Pareto for customers
Roughly 20% of your customers generate roughly 80% of your revenue. ABC analysis sorts customers into A (top revenue generators), B (middle), C (long tail). The insight is not the sorting — it's the differentiated treatment that follows. A customers get personal attention; C customers get cost-efficient automation. Without ABC, you treat every customer the same — which means you over-invest in C and under-invest in A.
Customer Lifetime Value — past, present, future
CLV is the total expected revenue from a customer relationship over its full duration. Three flavors: historic CLV (what has this customer brought in so far?), current CLV (what's the rolling 12-month value?), and predictive CLV (what's the modeled future value?). Each serves a different purpose. Acquisition decisions need predictive; retention decisions need current; reactivation decisions need historic.
Segmentation and clustering
Segmentation defines groups by rules (e.g., "premium customers in DACH with 3+ purchases"). Clustering discovers groups in data without prior rules (k-means, hierarchical, density-based). Both have their place. Segmentation is for activation; clustering is for discovery. Confusing them produces either rigid models or incomprehensible groups.
Descriptive, predictive, prescriptive
Three analytics stages with very different organizational implications:
- Descriptive: What happened. Reports, dashboards. Necessary, not sufficient.
- Predictive: What will happen. Models for churn, CLV, propensity. Requires data engineering and validation.
- Prescriptive: What should we do. Decision logic on top of predictions. Requires explicit policy choices, not just models.
Most organizations stop at descriptive and call it data-driven. Real maturity starts at predictive and reaches its potential at prescriptive — but prescriptive is also where most stakeholder discussions break down, because the policy choices become visible.
Where to go from here
The German-language pillar at /wissen/customer-insights/ contains the full depth (cluster articles on each topic plus usability research and consumer behavior models). For consulting on a Customer Insights audit: consulting page.