Learn · Should you run a loyalty program at all
The frequency and margin tests
Two calculations decide whether a program is worth building. The frequency test asks whether a median customer reaches the first meaningful reward within about six months. The margin test asks what share of gross margin the planned give-back consumes. A program that fails either test at the design stage fails faster in production.
Two calculations decide whether a program design is viable before anything gets built. The frequency test measures how long a median customer takes to reach the first reward worth wanting; much past six months, and engagement never starts. The margin test measures what share of gross margin the planned give-back consumes; more than about a tenth, before any evidence of incremental behaviour exists, and the design is betting the P&L on a hope.
Both bars are working rules. No published cross-industry benchmark establishes a safe give-back rate or a maximum time to first reward, so treat the numbers here as starting positions your own data will replace.
## The frequency test: time to first reward
Compute it directly: reward threshold, divided by points earned per visit, divided by visits per month. A coffee shop makes the mechanics visible. The average ticket is 6 dollars, the median customer visits twice a week, and the program earns 1 point per dollar, so a median member accumulates about 48 points a month. Set the first reward at 500 points and it sits ten and a half months away, which is too far; the member has stopped noticing the program by month four. Two honest levers exist, and raising the earn rate is neither of them. Lower the threshold and the reward together, 250 points for a 2.50 dollar reward instead of 500 for 5 dollars, and the wait halves while the economics stay identical. Or seed new members with a welcome bonus that moves them partway there on day one.
## The margin test: give-back as a share of margin
Divide the give-back rate by the gross margin percentage. The same coffee program returns 5 dollars per 500 points earned on 500 dollars of spend, a give-back of 1 percent. At a 65 percent gross margin, rewards consume about 1.5 percent of gross margin, comfortably inside the bar. Move the identical design to a retailer with 10 percent margins and the same 1 percent of spend now eats a tenth of gross margin, the entire working limit, before a single operating cost has been counted.
## The two tests pull against each other
The obvious fix for a failed frequency test is a higher earn rate, and every extra point of earn rate raises the give-back the margin test just capped. The free variables are elsewhere. Threshold and reward can shrink together without moving give-back at all, as the coffee example showed. Reward cost can also fall below face value when redemptions come from your own inventory, letting perceived generosity rise while the margin cost does not.
## Breakage cannot rescue a failed test
A design that clears the margin bar only by assuming a third of points are never redeemed is assuming, in the same spreadsheet, that members are engaged enough to change their behaviour and disengaged enough to abandon their balances. At the design stage you get one assumption. Model the margin test at full redemption. Breakage that shows up later is cushion; breakage that was planned for is a contradiction.
Leave the lesson with both numbers computed for your own business: months to first reward at median frequency, and give-back as a share of gross margin. If either fails, redesign now, because module 2 assumes the design in your hands passes.