Customer segmentation & lead scoring
On this page
A segment is just a rule: "high-value, active customers in these regions." Lead scoring is the same idea with a threshold. A rule engine lets marketing and growth teams define those rules and apply them to millions of people at once, or to one person, in real time.
The stakes🔗
Segmentation exists to power personalization, and personalization pays. McKinsey finds it lifts revenue by 5–15% and marketing ROI by 10–30%, and that the fastest-growing companies earn 40% more of their revenue from personalization than slower peers. It is also now expected: 71% of consumers expect personalized interactions and 76% are frustrated without them (McKinsey).
How it works🔗
A segment definition is a rule over customer attributes: for example, "high-value (≥ $500 spent), active (4+ orders), recent (within 90 days) customers in three regions." Run it across the whole base to build the segment; or run it in reverse (given one customer, which of hundreds of segments do they fall into right now?) to drive next-best-action at page load.
🤖 With AI in the loop
Predictive marketing runs model scores through marketer-owned rules. Braze assigns every customer a churn-risk score (0–100) from a model rebuilt about every two weeks, but marketers act on it through segment filters they can change today. Salesforce Einstein retrains its lead-scoring models every ten days and refreshes scores every few hours, while routing rules on the score assign leads to reps in seconds. Adobe's Customer AI writes propensity scores onto each profile precisely so segment rules can threshold them. In Winnow a propensity score is just another numeric column (customers with a churn score of 80 or more who ordered within the last 30 days, say), evaluated across the whole base in one pass, with thresholds the team tunes live. More patterns: Deep Dive: AI + Rules Engines.
Why Winnow fits🔗
- Recency, frequency, monetary values are number checks; region, tier, and channel are category checks; interests and tags are text matches.
- The whole base in one pass: bucket millions of customers at once, and re-segment on demand as definitions change.
- Real-time classification: place a single customer against every segment in microseconds, fast enough for on-page personalization.
- Marketer-owned rules: segments are plain text, edited without engineering and applied live, with no release cycle.
🧪 In simulation — 5,000 segments, 1 million customers, both directions
To put numbers on the above, we defined 5,000 segments — category-affinity segments, RFM tiers, regional campaigns like the three-region example above, lifecycle cohorts, and win-back rules keyed on inactivity — each carrying a lead-score weight, then ran the book both ways: 1 million customers classified one at a time, the page-load direction, and the whole base swept forward to build every segment's membership list in a single pass. 71% of customers landed in at least one segment, four on average, and each classification also returned the customer's lead score — the sum of the weights of every segment that fired.
What this is: a test run on generated segments and a generated customer base, not a customer deployment — the segments and profiles are invented, the speeds and costs are measured. It ran on one ordinary 16-core desktop computer, priced at a standard cloud rate of $0.043 per processor-hour running around the clock, the same basis as the Deep Dive: Cost Savings. Spend thresholds are held as exact money amounts, and both directions were double-checked against a slow, exhaustive pass over all 5,000 segments — which, honestly, is only about twice as slow here: a 5,000-rule book is small, and what the engine buys at this scale is not a big multiple but a microsecond page-load path, a whole-base rebuild measured in seconds, and a cost that tracks the segments a customer actually matches rather than the size of the book as it grows. Treat the speeds as approximate, moving by about a fifth with machine load, and the counts as exact; the test ships with Winnow for your team to re-run on your own segments.