Loyalty glossary · 7. Data and analytics (31)
Micro Segmentation
Micro segmentation is the practice of dividing a loyalty programme's membership into very small groups based on detailed behavioural and transactional data, then setting different earn rates, offers, or service levels for each group.
Most segmentation splits a loyalty file into a handful of groups. Micro segmentation splits it into dozens or hundreds, and the split is built on behaviour, not demographics. The operator then treats each micro segment as if its numbers were reliable enough to run a profit and loss on. Small samples produce noisy averages, and the noise gets promoted into insight.
The smaller the segment, the less stable its behaviour. A segment of a few hundred members can show a 30 percent redemption rate one year and 10 percent the next without any change in the programme, purely from random variation. Operators mistake that variation for a meaningful shift and change rules that confuse everyone else. An actuarial model built on those unstable rates will misprice the liability.
Work the arithmetic, because most apparent micro segment differences are just scale. A hotel programme has one segment averaging 2 stays per year and producing 4,000 dollars of annual margin. Another segment averages 5 stays per year and produces 10,000 dollars. The difference is 6,000 dollars, a 150 percent increase over the first segment, but the margin per stay is identical at 2,000 dollars. The segment with 5 stays is larger, not more valuable.
The actuarial model that sets breakage and liability assumptions must be fed at segment level, because redemption patterns differ across micro segments. Accrual rules then multiply, as each segment may earn points at a different rate. Activity-based qualification becomes a per segment threshold rather than a programme standard. None of this adds member value by itself.
The right use of micro segmentation is narrow. Find the handful of segments where a different rule changes economics materially, and leave the rest alone. Most programmes over-segment and end up managing hundreds of tiny rules that cost more than the margin they produce. The test is whether the segment specific action changes a recognised quantity like dollars or points by enough to justify the operating overhead. If not, cut the segmentation.