A coupon for everyone pays the regulars who would have bought anyway. It does little for the member who has gone quiet. Behavioral segmentation fixes that by grouping customers by what they do, such as how often they buy, what they buy and how they respond. It is the base of most loyalty programs, because behavior is the one thing a program sees in full.
What is behavioral segmentation?
Behavioral segmentation sorts customers by how they act with a product or service. The OpenStax textbook calls it the how, next to the who of demographics and the why of psychographics (OpenStax). Inputs are acts a business can log. Buys, visits, channel use and offer response all count.
What are the types of behavioral segmentation?
OpenStax names three main bases: benefits sought, occasion and usage. Loyalty teams add status, recency and journey stage. The table shows each with a typical signal and a use in a loyalty program.
| Type | Groups customers by | Typical signal | Loyalty use |
|---|---|---|---|
| Benefits sought | What they want from the product | Category and feature choices | Choose which reward to feature |
| Occasion | When they buy | Holiday, birthday or trip purchases | Time offers to the occasion |
| Usage rate | How much they use it | Light, medium and heavy buyers | Protect heavy users; grow light ones |
| Loyalty status | How committed they are | Tier, points balance, redemptions | Tier benefits and recognition |
| Recency, frequency, monetary (RFM) | How recently, how often and how much | Transaction history | Win-back, upsell, reward the best |
| Journey stage | Where they are in the relationship | New, active, lapsing, lapsed | Onboarding, nurture, reactivation |
On usage, OpenStax notes that heavy users are a small share of the market but account for a high share of buying. It cites the Pareto rule of thumb that 80% of revenue comes from the top 20% of repeat or loyal customers (OpenStax). Treat that as a prompt to check your own split. It is a rule of thumb and not a law.
Why does behavior predict behavior?
Past behavior guides future behavior in two ways. Ouellette and Wood showed that well-practiced acts in stable settings recur because they become automatic. In that case, frequency reflects habit strength. When an act is new or the setting is unstable, people decide with thought, and intentions drive it (Ouellette and Wood, 1998).
That split tells you where behavioral segments are strong. They predict well for habitual categories, like coffee or groceries. They predict less well for rare, high-stakes buys, where each purchase is a fresh decision. Russell Haley made the wider point in 1968: descriptive facts about buyers do not predict future purchases well, so he started from benefits sought (Haley, 1968).
How does RFM segmentation work?
RFM scores each customer on how recently they bought, how often they buy and how much they spend. Most teams use 1 to 5 scales. High scores on all three mark the customers to protect. High frequency and spend with a fading recency score marks a valuable customer who is drifting. For illustration, a member who last bought 20 days ago, bought 14 times in a year and spent $620 scores 5, 4 and 4 on a scale where 5 is best.
RFM needs only transaction history, and it explains itself to any manager. Its limits are real. It is blind between purchases and it looks backward. It also punishes customers who are mid-cycle in long-cycle categories. Our segmentation guide says to add engagement and predictive scores on top. Do not drop RFM.
Fader, Hardie and Lee gave RFM a formal link to customer lifetime value. Their iso-value curves group customers who have different histories but similar future value (Fader et al., 2005).
Why is behavior a better basis than demographics?
Do brands attract different kinds of people? Hammond, Ehrenberg and Goodhardt checked buyers in over 20 grocery categories. They found little brand segmentation. Rival brands had buyers with similar ages, incomes and the like (Hammond et al., 1996). So who someone is rarely tells you which brand they pick. What they do does.
How do loyalty programs use behavioral segments?
Tiers are behavioral segments with names. Starbucks sets a member’s level by the Stars earned over 12 months. A member reaches Gold at 500 Stars, and Reserve at 2,500 new Stars earned after Gold (Starbucks Rewards Terms). Delta sets Medallion status by spend, with 2027 thresholds from $5,000 to $28,000 in Medallion Qualification Dollars (Delta Air Lines).
Programs also segment by activity. At Starbucks, Green Level Stars expire six months after the month earned. The member can extend them with a purchase, a reward redemption or a reload of $30 or more in the month before expiry. Gold and Reserve Stars do not expire while the status holds (Starbucks terms). The program splits active members from lapsing ones by rule.
What are the limits of behavioral segmentation?
Behavioral segments have limits. A new member has no history, so there is nothing to score. A history shows what a member did and not why. And a habit can break when the setting changes, as the habit loop research shows. Pair behavior with declared preferences, and rescore often. Pick the window with care too. A 90-day window will miss members who buy once a year, so match the window to the buying cycle.
How do you build and test behavioral segments?
Start with the decision. Each segment should name the treatment it triggers. OpenStax’s ADAMS test is a good filter: accessible, differentiable, actionable, measurable and substantial (OpenStax). Then pick windows and thresholds, such as active in the last 90 days.
Name each segment by the action it triggers. Our segmentation guide gives two examples. High value with falling frequency exists because a win-back offer will target it. New with a second purchase pending exists because onboarding will. If nothing hangs off a segment, drop it.
Then time the loop. The guide has a test for this. Count how long it takes to go from deciding to treat a group differently to treating the first member differently. Mature programs answer in hours. If it takes weeks, the limit is activation and not analysis.
For illustration, a program scores 12,000 members on RFM. Recency scores 5 for a purchase in the last 30 days, and 1 for none in 150 days or more. It finds 1,500 members in a drifting valuable group. They have high frequency and spend but low recency. It sends a win-back offer to 750 at random and holds out 750. If 18% of the treated group buy in the next 30 days and 11% of the holdout do, the offer added 7 points. That is about 50 extra buyers. Check whether the holdout also bought less over the following month, since an offer can pull purchases forward.
How GRAVTY handles behavioral segmentation
The technology page says each action and sale becomes first-party data. Our segmentation guide says segments sit in patented Visual Rules. The rules read live member data and model scores. Each offer, reward and message shares the same rules. A member who crosses a line mid-purchase sees the change in that session.
Frequently asked questions
What is an example of behavioral segmentation?
Loyalty tiers are the classic example. Starbucks sets a member’s level by Stars earned over 12 months, with Gold at 500 Stars (Starbucks Rewards Terms). Each level is a behavioral segment with its own benefits.
What are the main types of behavioral segmentation?
The textbook types are benefits sought, occasion and usage rate (OpenStax). Loyalty teams add status, RFM scores and journey stage. Most programs cross two or three of them.
What is the difference between behavioral and demographic segmentation?
One groups customers by what they do. The other groups them by who they are. Haley said in 1968 that descriptive facts predict buying poorly (Haley, 1968). Actions are logged directly, so they are usually the stronger signal.
What is RFM segmentation?
RFM scores customers on recency, frequency and money spent. Fader, Hardie and Lee tied it to customer lifetime value (Fader et al., 2005). It works from sales history alone. It cannot see what happens between purchases.
Why is behavioral segmentation useful for loyalty programs?
Because past behavior predicts future behavior, most of all for habits in stable settings (Ouellette and Wood, 1998). A program records every earn and burn on a member identity. That history is the raw material for tiers, win-back offers and next-best actions.
