Surviving The Perfect Member
By Ory Adler
In my last piece I argued that agents will push breakage toward zero and the most common pushback I've heard is about timing. Yes, agents complete a small share of purchases today, but the share of AI users in the USA who've let an agent complete a purchase went from 10% to 17% in a year, and agents don't have to transact to change how people shop. McKinsey found that 48% of US consumers used AI for shopping in the past three months, and nearly a third of AI shoppers said it was the biggest influence on their most recent purchase¹, so these customers are building their carts online, or walking into your store, already informed by an agent how to get the most out of the trip.
The exposure from a Loyalty standpoint comes in two flavours. Your base earn rate and publicly available hero offers are likely to be already being included in comparisons today. However, as 3rd party shopping agents will acquire access to points balances and personalized offers, and as your own AI assistant will start reminding members that they have $40 worth of points or offers available (ironic that it may turn out to be the most effective breakage killer) the erosion in points breakage will rapidly start to accelerate.
A simple stress test could reveal critical insights as you start developing your plans for the agentic world. Imagine that every member behaved like the perfect member, one that looks at every offer (mass or personalized), reads every rule and never lets a point expire, and model your program at maximum earn and zero breakage, this will be the most extreme, worst case scenario but not a realistic one as even at a heavily penetrated Agentic Shopping world, we will likely see at least 5-10% breakage. This said, this initial model will help you identify the most critical levers that you will need to manage in the long-term and inform decisions related to program features and economics in the short-term. Below are some thought starters on how to approach this:
Base & Bonus Offer Earn Economics
Start with the economics at 100% redemption and work back to the proposition. A 1.5% average annual earn rate that cost ~1.1% at 25% breakage will cost the full 1.5% at zero, and if the program’s incremental margin doesn’t cover the difference, members who forgot their points today have been subsidizing your earn rate investment.
From there, separate the everyday base earn for shopping (and customer identification) from the bonus offers that you pay to change behaviour (a second visit in the week, a first purchase in a new category etc.). Base earn is the number AI is already putting against your competitors', and most retailers struggle with it the most as it usually drives limited incremental spend. Changing your base earn is risky but may be necessary depending on its share of your total points issuance. Mass and Targeted Bonus offers should be measured to ensure that at a minimum they pay for themselves (independently or through vendor funding) when designed correctly, additional applications of these offers should also represent an opportunity for offsetting reductions in the base earn.
As you work through this, it’s important to keep in mind that agents will introduce other considerations, for example some customers may ignore or forget about a "shop four weeks in a row" bonus offer but an agent can track the requirement and remind the customer to complete it and suddenly, a promotion that was historically profitable due to partial participation will become considerably more expensive.
Reward Your Best Customers for Saving Their Points
Start by auditing your redemption rules (minimum thresholds, activation requirements, exclusions etc.) and assess whether each one protects a genuine operating need or mostly produces unused balances. The second type are going to be less likely to survive an agent and may represent an opportunity for removing friction and improving program engagement and satisfaction in the near term. Enabling easier burn has risks though, a growing balance drives retention as members come back because they're 200 points from their next reward, and if redemption becomes more frequent, the balance is likely to reset along with the reason to return.
Personalization also offers some interesting opportunities for managing breakage, for example: you could offer all members a simple everyday redemption scheme (e.g. 1000pts = $1) then offer your best customers an accelerated redemption table that rewards them for building a balance:
Points redeemed |
Standard value |
Accelerated value |
|---|---|---|
10,000 |
$10 |
$10 |
20,000 |
$20 |
$25 |
50,000 |
$50 |
$75 |
A customer that is earning ~5,000 points a month may choose to wait a few months to turn $20 into $25, and their agent may recommend exactly that, but you only offer the premium to the customers you've chosen. This said, you are deliberately increasing the reward cost for some members though, and an agent might just as easily recommend redeeming right away if the customer is unlikely to reach the next threshold or the future reward isn't worth the wait. The economics only work if the additional retention, incremental margin and breakage will justify the richer redemption value. Pointing those richer levels at own-brand or partner-funded rewards could also offer some relief, if members actually see the value in them.
Why Keep Earning Once the Points Are Spent?
Once agents make redemption frequent and predictable, the points balance stops doing much of the retention work on its own. Status and recognition give members another reason to keep earning, provided they unlock more value and better treatment based on what members have earned (not spent) over a year or a lifetime and remain intact when current balances are redeemed.
Starbucks relaunched its updated program in March 2026 with Green, Gold and Reserve tiers, based on Stars earned in a 12-month period². Gold members earn 20% more Stars and Reserve members 70% more, with perks like non-expiring Stars, exclusive events and paid trips at the top, but status must be re-earned every year.
Many airlines have historically offered benefits based on lifetime earn. Air Canada grants lifetime Aeroplan Super Elite status at 3M lifetime qualifying miles³, and British Airways asks for 550,000 lifetime tier points for Lifetime Gold⁴. A member who's close to lifetime status is unlikely to move their business for a slightly better offer down the street.
Airlines learned the hard way that activity and value aren't the same thing, for years many rewarded distance flown, which handed status to bargain long-haul travellers while full-fare business travellers on short routes fell behind, and many have since moved toward spend-based qualification. The same applies in retail, your base earn and promotional point investment need to be designed so most of it lands with the customers who matter most to your business.
I've seen the importance of this firsthand. When we designed Staples Easy Rewards, we focused on identifying the customers who generated disproportionate profit for the business (largely small business owners) and built a proposition that treated them differently, because their contributions mattered far more than their share of spend or transactions.
If you're designing tiers today, I'd push hard on the qualification metric. Spend, visits and units are convenient to measure as well as easy to explain to the customer, but none of them necessarily reflects profitability, and a customer buying frequently on promotion could qualify for your highest tier while contributing relatively little incremental margin. If agents make cherry-picking easier, that problem gets worse, so it's worth checking who is actually sitting in your top tier before you make it richer.
Where to Start?
Before you approve the next earn rate change, membership tiersstructure or bonus promotion, model the program at zero breakage and then at a realistic Agentic-era level. Look at whether your richest benefits are going to the customers who deserve them economically as well as which tactics generate incremental margin or work only because members overlook them. I suspect many programs will find that part of their economics depends on a large group of members never claiming the value they were promised, and AI is going to make that much harder to ignore. Proactively finding and addressing these risks now, will be easier and far more cost effective than when the Shopping Agents will start to reveal them.
Sources