By Diosh — Founder, AHAeCommerce | eCommerce decision intelligence for $50K–$5M GMV operators
Your competitor launched a points program, so now you're staring at Smile.io and LoyaltyLion pricing pages wondering if you're falling behind. The pitch is clean: points increase retention, retention increases LTV, and LTV is the whole game. The problem is that the pitch quietly assumes every redeemed point bought you a repurchase that wouldn't have happened otherwise — and that assumption is false for the majority of your redemptions. This is a trade-off piece for operators who are about to spend real margin on a loyalty program: the trade you're actually making is between a measurable, near-certain cost (discount liability plus app fees plus ops) and an unmeasured, mostly imaginary gain (incremental retention you assume but rarely isolate). By the end you'll be able to model whether your category's purchase frequency can carry the program — or whether you're about to subsidize behavior you already had.
The Subsidy Problem: Who Actually Earns Your Points
Start with the uncomfortable accounting. The customers who earn and redeem the most points are, by definition, your most active repeat buyers — the people most likely to have bought again regardless. A points program is structurally regressive: it routes the largest discounts to the customers who needed the least convincing.
Picture a $2M GMV supplements brand. Roughly 40% of its revenue comes from customers who reorder every 45–60 days because they ran out, not because of points. When this brand layers on a 5% points-back program, that 40% segment captures the bulk of the points liability while contributing almost none of the incremental behavior the program is supposed to create. The brand is paying a 5% tax on revenue it already owned.
The honest question is never "does the program increase retention among members?" Members will always retain better than non-members — but that's selection bias, not causation. Your best customers self-select into the program. The question is the counterfactual: of the repurchases attributed to points, how many would have happened anyway? McKinsey's loyalty research has repeatedly found that membership correlates with spend without establishing that the program caused it, and that many programs fail to generate returns above their cost (McKinsey & Company, https://www.mckinsey.com). That gap between correlation and causation is where your margin disappears.
This is the same trap that distorts most retention analysis. If you haven't separated natural repurchase behavior from program-driven behavior, you're measuring loyalty you already had. The deeper version of this accounting lives in the real economics of a loyalty program — read it before you sign an annual app contract.
Incremental Retention Is Smaller Than the Vendor Decks Show
Vendor case studies love a headline like "members spend 2.3x more than non-members." Treat that number as marketing, not measurement, because it almost never isolates incrementality. The figure that should drive your decision is incremental retention — the lift attributable to the program, net of what would have occurred without it.
Across categories, well-run programs tend to produce incremental retention in the low single digits — roughly 2–5% in my experience modeling these for operators, and that range is consistent with the modest causal lifts academic and consulting analyses report once selection bias is stripped out. Harvard Business Review's long-running work on loyalty economics has shown that the assumed link between loyalty-program membership and profitability is weaker and more conditional than managers expect, with many "loyal" customers being no more profitable to serve (Harvard Business Review, https://hbr.org). A 2–5% incremental lift is real, but it's not the 30%+ retention transformation the pricing page implies.
Run the arithmetic on that supplements brand. Say the program costs 5% of member revenue in points liability plus a $599/month app tier plus a few hours of ops weekly. If members represent $800K of annual revenue, the points liability alone is roughly $40K, before fees and labor. To break even, the program must generate enough incremental gross margin to cover $40K+ — and at a 3% incremental retention lift on a 40% gross margin, the incremental contribution may not clear the liability. The program can run for a year looking "successful" on a membership dashboard while quietly running negative on a contribution basis.
This is why the LTV number on your dashboard usually lies — it blends subsidized and organic revenue into a single flattering figure. And it's why you have to set the loyalty decision inside the broader retention-versus-acquisition math: a dollar of points spend competes with a dollar of acquisition spend, and points often have the worse incremental return.
Purchase Frequency Is the Make-or-Break Variable
Whether a points program pays back is dominated by one input: how often your category is naturally repurchased. Points need repeat visits to compound — a customer has to return often enough to accumulate a meaningful balance and feel the pull of redemption. If natural frequency is low, the program never gets the at-bats it needs.
Here's the operator heuristic worth internalizing. If your category produces fewer than roughly two purchases per customer per year on its own, a points program rarely pays back, because most customers never reach a redemption threshold and the rewards you do pay out go to the minority who were already frequent. Consumables, pet supplies, coffee, and cosmetics replenishment can clear this bar. Furniture, mattresses, luggage, and most considered durable goods do not — a customer buying a mattress every seven years will not be moved by points.
- High natural frequency (3+ orders/year): Points can amplify an existing rhythm. A coffee subscription brand at 8–12 orders/year per active customer has enough redemption velocity to make tiered rewards feel earned and to nudge the at-risk middle.
- Borderline (1.5–2.5 orders/year): The program lives or dies on segmentation. Blanket points-back bleeds margin; targeted offers to the at-risk cohort might clear. Model before launching.
- Low natural frequency (<1.5 orders/year): A points program is almost always a margin leak. The customers won't return often enough to redeem, and the few who do were your loyalists anyway.
A $1.5M GMV home-fragrance brand learned this by running the numbers after a year on a points app: natural repurchase sat at 1.3 orders/year, redemption clustered entirely in the top decile, and the program's contribution was negative once the app fee and the founder's weekly ops time were costed. The fix wasn't a better program — it was killing it and redirecting the spend toward winback flows. The frequency question is also a churn question, which is why it belongs next to how to actually price the cost of a lost customer.
The Accounting Trap Nobody Books: Points as a Liability
Now the part most operators never see coming. Unredeemed points are not a marketing perk — they are a liability that accrues on your balance sheet. Every point you issue is a future claim on your margin, and until it's redeemed or expires (breakage), it sits there growing. Most $1M–$5M operators never book this liability, run the program off the app's dashboard, and discover the exposure only when redemption spikes.
Breakage — the share of points that expire unredeemed — is your friend here, and it's real: a meaningful fraction of issued points are never redeemed, which is part of why programs can pencil out at all. But breakage rates are not a constant you control, and leaning on them is a bet, not a plan. Bond Brand Loyalty's annual loyalty research documents how wide the gap is between points issued and points that actually drive satisfaction or redemption, and how often members can't even recall their balances (Bond Brand Loyalty, https://bondbrandloyalty.com). You can't budget a program on the assumption that customers will keep forgetting their points.
The trap has a second, sharper edge: when redemption happens. Points get redeemed during your highest-intent, lowest-margin windows — Black Friday, end-of-season clearance, your own promotional emails — and they stack on top of SKUs you've already discounted. So the liability doesn't just sit; it comes due precisely when your margin is thinnest, on the exact products where you can least afford it. An operator who never modeled stacked redemption can watch a "5% program" eat 15–20% of margin on a promotional weekend because customers combine points with the sale price. Deloitte's customer-loyalty analyses make the same point about hidden program economics — that the visible reward cost understates the true margin impact once redemption timing and stacking are included (Deloitte, https://www.deloitte.com).
How to Model It Before You Sign Anything
Don't launch and measure later. Build the model first, because the costs are knowable in advance and the gains are not. Here's the sequence I run with operators deciding this.
Step 1 — Segment by natural repurchase rate
Pull your last 18–24 months of order data and bucket customers by how many times they bought without any program. This gives you the organic baseline — the repurchase that points cannot take credit for. If 60% of repeat revenue comes from a high-frequency core, you already know the bulk of any points liability lands on customers who didn't need it.
Step 2 — Estimate incremental retention conservatively, not optimistically
Use a 2–4% incremental lift as your working assumption, not the vendor's headline multiple. If the program only pencils at a 10%+ lift, it doesn't pencil — that lift is rare and you shouldn't bet the budget on it.
Step 3 — Book the full cost, including the liability and stacking
Sum the four real costs: points liability at full issuance (don't pre-credit breakage), the app subscription, ops labor at a real hourly rate, and a stacked-redemption margin hit during promo periods. Compare that total against the incremental gross margin from Step 2. If the incremental margin doesn't clearly exceed the cost, the program is a subsidy.
Step 4 — Check it against your discount posture
A points program is a structural discount. If your brand already leans on promotions, you're compounding a dependency, and you should read how discount dependency quietly hollows out a brand before adding another always-on discount channel.
The Decision, Stated Plainly
The trade-off is not "loyalty program: yes or no." It's "a certain, ongoing margin cost in exchange for an uncertain, usually small incremental gain — gated almost entirely by your purchase frequency." Run a points program when your category is naturally high-frequency, your customers reach redemption thresholds on their own behavior, and you've modeled the liability including stacked promotional redemption. Skip it when natural frequency is below roughly two orders per year, because you'll be paying your existing loyalists to keep doing what they were already going to do.
If you take one action from this: segment your base by natural repurchase rate this week, model the discount liability specifically on the customers who would have returned without it, and only then open the pricing page. The math, not the competitor's launch announcement, decides whether points cost you margin.
