Last updated: September 2026
Many tier structures are expensive in the same two ways: they hand status to members who would have spent the money anyway, and they take it back at the moment those members are most likely to leave. Loyalty program tier design is the set of decisions that determines which behavior earns status, how many levels a program has, how far apart the thresholds sit and what happens when a member stops qualifying, calibrated so that status changes behavior rather than rewarding behavior that would have happened anyway.
Those decisions fall into two phases with different psychology. While a member climbs toward a tier, the goal-gradient effect does most of the work. Once status is held, loss aversion takes over, and a downgrade policy written as an afterthought can undo the loyalty the climb created. This article covers both phases for loyalty directors and heads of CRM, from the qualification metric to measuring whether each tier pays for itself.
A tier structure is a pricing decision presented as a recognition scheme. Every benefit attached to a level, whether a points multiplier, free shipping or early access, has a cost that is justified only by behavior the tier causes. Some share of the members in any upper tier would have spent at that level anyway. For them, the tier is a rebate.
The incremental value sits in two groups. The first is members within reach of a threshold, whose purchase timing and share of wallet can shift because a goal is close. The second is members who already hold status and might otherwise drift to a competitor, for whom tier benefits raise the cost of switching. A structure that concentrates benefit spend on those two groups can create margin. One that spreads rich benefits evenly across everyone above a static line is more likely to erode it.
Consider an illustrative specialty retailer whose top tier starts at $1,000 in annual spend. If most members above that line spend $3,000 or more, they cleared it without trying. The members spending $700 to $900, where the tier could change behavior, may never be told how close they are. The design question is where the members are, not where a round number sits.
The first practical step is therefore a map, not a benefit list: plot the distribution of annual value across active members and mark where the current thresholds fall. Every later decision uses that map.
Loyalty tiers should qualify on the behavior the business most needs to change: spend rewards revenue, frequency builds habit, and hybrid gates suit top tiers.
Members respond to the qualification metric literally. Qualify on visits or transactions, and some members will split baskets to accumulate credit faster. Qualify on spend, and a frequent, low-ticket customer may never see progress worth pursuing.
Spend is also a revenue measure, not a margin measure. Where margin varies widely by category or discount depth, a member can reach status on the least profitable baskets. Programs in that position can qualify on margin-weighted spend, or exclude deeply discounted purchases from qualifying credit.
North American travel programs show the two poles and a hybrid between them. Delta Air Lines qualifies Medallion Status on Medallion Qualification Dollars, a spend measure, with thresholds of $5,000, $10,000, $15,000 and $28,000 for Silver through Diamond (Delta Air Lines, 2026). Marriott Bonvoy qualifies its lower levels on nights, from 10 nights for Silver Elite to 75 for Titanium Elite, and adds a spend requirement only at the top: Ambassador Elite requires 100 nights plus $23,000 in annual qualifying spend (Marriott International, 2026). Nights alone cannot unlock the most service-intensive tier.
The table compares five qualification bases by the behavior each rewards and what it costs when members optimize for the metric rather than the intent.
|
Qualification basis |
Behavior it rewards |
Fits when |
Main risk |
What to monitor |
|
Spend (annual dollars) |
Basket size and share of wallet |
High-value customers buy infrequently; margin varies little by purchase |
Frequent, low-ticket customers see little progress and disengage; where margin varies, status is earned on low-margin baskets |
Share of active members within reach of each threshold |
|
Frequency (visits, orders, nights) |
Habit and visit cadence |
Traffic itself has value, as in QSR, convenience or hotel occupancy |
Basket splitting; status earned on low-margin behavior |
Average transaction value by tier over time |
|
Hybrid (frequency, plus a spend gate at the top) |
Habit at entry, value at the top |
Top-tier benefits are costly to serve |
Members misread two-part rules |
Near-miss rates on each criterion; rule-related service contacts |
|
Engagement actions (reviews, referrals, profile data) |
Data and advocacy |
First-party data or referrals have a measurable value |
Status earned without revenue |
Revenue per member by tier |
|
Paid membership tier |
Upfront commitment |
Benefits are used often enough to show payback |
Members who pay but underuse benefits cancel at renewal |
Benefit usage and renewal rate |
Every basis depends on identity resolution. A member whose store and online purchases are not matched to one profile sees less progress than they have earned, which overstates the distance to the threshold and weakens the goal gradient. Hybrid rules add a second dependency: the rules engine, member dashboard and proximity messages must all evaluate and explain both criteria consistently, and a program that can't yet do that is better served by a single metric.
In practice, a loyalty program usually works best with three or four tiers: enough for a reachable next goal, with a top tier small enough to signal status.
The clearest evidence on tier count comes from Drèze and Nunes (Journal of Consumer Research, 2009). Increasing the number of members in the top tier diluted its status, while adding a subordinate tier beneath it enhanced status. Members who did not qualify for status preferred programs with multiple tiers.
Two design rules follow. First, the top tier's value depends on its size, so a promotion that fast-tracks thousands of members spends the scarcity the top tier is selling. Second, a middle tier does two jobs: it gives non-elite members a reachable goal, and it makes the top tier feel further away. The research does not set a number. In practice, three or four levels usually cover both jobs; a fifth earns its place only if a distinct group of members sits between thresholds with nothing within reach.
Spacing matters as much as count. Delta's thresholds rise in equal $5,000 steps to Platinum, then jump $13,000 to Diamond (Delta Air Lines, 2026). One reading, consistent with Drèze and Nunes, is that equal steps keep lower tiers reachable while the wide final gap protects top-tier scarcity. A workable method for any program is to set each threshold from the member value map: place it where a meaningful group of members sits just below the line, so the goal is close enough to change behavior, then check that the benefits above the line cost less than the incremental margin those members are likely to generate. Smaller or newer programs without enough member history can start from all-customer transaction data, launch with two or three tiers, and add a level once the distribution shows members with nothing within reach.
The common failure is setting thresholds once, at launch. Price changes, inflation and mix shifts move the value distribution every year, so review thresholds against it annually, with finance, before the next qualification year opens. Raising a threshold is also a partial demotion for members who would have qualified under the old rule, so announce increases a full qualification year ahead.
The attainment phase runs on the goal-gradient effect: effort accelerates as a goal gets closer. Kivetz, Urminsky and Zheng documented it in a café reward program, where members bought coffee more often as they approached a free drink; inter-purchase times fell by an average of 20% (Journal of Marketing Research, 2006). The same research found that purchase rates reset to a lower level after the first reward, then re-accelerated toward the next, and that members who accelerated more strongly were more likely to be retained.
For tiers, the reset is the risk. A newly promoted member has closed the gap that was motivating them, and without a visible next goal nothing pulls the next purchase forward. Show the next milestone at the moment of attainment, whether the next tier, the requalification target or a benefit unlocked within the tier, rather than waiting for a scheduled campaign.
The endowed progress effect applies at the other end of the climb. Nunes and Drèze found that people given artificial advancement toward a goal were more likely to complete it, and to complete it faster: a card requiring 10 stamps with two already filled outperformed an equivalent card requiring eight (Journal of Consumer Research, 2006). Tier programs can apply this by crediting new members with initial progress toward the first tier, or by carrying excess qualifying activity into the next year. The same research found that the reason offered for the head start, if any, moderated the effect, so its framing deserves as much testing as its size.
Progress visibility is the dependency. Goal-gradient behavior requires members to know how close they are, so qualifying activity has to post quickly and consistently across channels. If credit posts in a monthly batch, the member closest to a threshold is working toward a number they cannot see; where batching can't be avoided, time proximity messages to each posting cycle.
Distance to the next threshold is also a segmentation variable the program already holds, and it should change treatment. Members within reach get proximity messages timed to their usual purchase cycle. Members far above a threshold get recognition rather than incentives, which would subsidize behavior already happening. Members far below get the nearest goal, not the top tier. Attainment is also a natural moment to ask for data: a newly promoted member has a reason to say which tier benefits they would use, and those answers let the program move benefit spend toward what members value. Ask with clear consent, and say how the answers will be used.
A tier downgrade should be gradual, predictable and communicated early, because demoted members can become less loyal than members who never held status at all.
Once status is held, loss aversion changes the member's reference point: tier benefits start registering as something the member owns, so losing them is felt more sharply than gaining them was. Wagner, Hennig-Thurau and Rudolph tested this directly (Journal of Marketing, 2009). In a scenario experiment, demoted customers showed lower loyalty intentions than customers never awarded preferred status, and a field study using company sales data confirmed the negative effect. The same research tested how firms can soften a demotion. Clear information about membership conditions and the customer's own spending helped, and an apology reduced negative feelings and raised loyalty intentions. Monetary compensation had no significant effect. Even with every measure applied, demoted customers stayed less loyal, on average, than those whose status never changed.
Downgrade rules are therefore a retention decision with a customer-equity cost. Four levers change that cost:
Lenient rules carry their own cost: members who keep status without the qualifying behavior still consume benefits, and a growing top tier dilutes its status. The balance depends on how expensive top-tier benefits are to serve and how much of a lapsing member's value is recoverable.
Changes to thresholds and downgrade rules also carry regulatory exposure. In September 2024 the US Department of Transportation opened an inquiry into the four largest US airlines' rewards programs. The practices it examined included adding new hurdles to qualify for status, and it required each airline to describe every program change over the previous six years and how it affected existing points and status (US Department of Transportation, 2024). The inquiry covered airlines, but it showed which tier changes a federal regulator treated as potential devaluation. Outside air travel, the Federal Trade Commission has authority to prevent unfair or deceptive acts or practices in commerce (Federal Trade Commission, 2026). Threshold and downgrade rules belong in the program terms, which should state how the rules can change and with how much notice. Write those terms carefully, including how progress members have already earned is treated.
Measure loyalty tiers by comparing members just above and just below each threshold, then weighing their incremental margin against the full cost of tier benefits.
Comparing top-tier members with base members answers the wrong question, because top-tier members were selected for high value. A better comparison is local: members who finished a qualification year just above a threshold against those just below it. The groups behaved similarly before the line, so next-year differences come closer to the effect of status itself. For the climb, compare members' purchase timing near a threshold with their own timing earlier in the year.
In an illustrative grocery program with a $2,500 mid-tier threshold, members who finished the year at $2,300 to $2,499 are compared with those at $2,500 to $2,700. If the promoted group's next-year margin runs ahead by more than the cost of the tier's benefits, the tier is paying for itself. Two limits apply. Members who pull purchases forward to cross the line are more motivated than those who fall short, so the comparison can overstate the effect of status; checking both groups' prior-year behavior reduces that bias. And in a smaller program the bands may hold too few members for a stable result, so widen them, pool several years and treat the finding as directional.
|
Measure |
What it shows |
Warning sign |
|
Near-threshold acceleration |
Whether thresholds change behavior during the climb |
No change in purchase timing as members approach the line |
|
Post-attainment spend and retention |
Whether status holds value after the climb |
Spend falls back to pre-climb levels after qualification |
|
Downgrade lapse rate |
The customer-equity cost of the retention rules |
Demoted members lapse faster than never-elevated members of similar value |
|
Benefit cost vs. incremental margin, by tier |
Whether each tier pays for itself |
Benefit cost rising faster than incremental margin in the top tier |
Tier points multipliers need finance input because they change liability as well as cost. Under ASC 606, expected breakage on amounts held in a contract liability is recognized as revenue in proportion to the pattern of rights members exercise (FASB, ASU 2014-09, 2014). If top-tier members redeem at a higher rate than the base, expected breakage for that tier is lower than the program average, and a multiplier costs more than a blended estimate suggests. High breakage within the top tier signals the opposite: members may value the status and service benefits but not the points, which argues for moving benefit spend away from multipliers.
Measurement depends on at least one full qualification cycle of member-level history and on agreement with finance about how benefit costs are allocated; settle both before the first review. Track near-threshold acceleration monthly so you can adjust proximity treatment in-year, and run the cost-versus-margin review annually, before the next cycle's thresholds are set.
Tier design is usually discussed as a question of benefits. The commercial question sits underneath. A tier structure pays for itself when its thresholds sit where members can be moved, when the qualification metric rewards the behavior the business needs, and when the rules for losing status do not undo the loyalty the rules for gaining it created.
The two phases need separate designs. During the climb, the goal-gradient effect and endowed progress reward goals that are visible and reachable, and the post-attainment reset argues for showing the next goal the moment a member arrives. Once status is held, loss aversion moves the member's reference point, and the evidence on demotion suggests a hard downgrade can leave a member less loyal than if they had never been promoted. Count and spacing decisions cut across both phases: a small top tier carries status, and a well-placed middle tier makes that status visible to everyone below it.
None of this is settled at launch. Tier design holds its economics when it is treated as an annual decision made with finance and measured locally at each threshold rather than across the whole base.
If you mapped your member value distribution against your current thresholds today, how many members would be within reach of the next tier, and how many would be sitting well above it without being asked to do anything?