How to start a loyalty program: set the goal, pick a model, budget the rewards, connect your data and run a measured first year. A vendor-neutral guide.
Start with one business goal, such as more visits or bigger baskets. Choose the program model that rewards that behavior, budget the rewards as a liability, connect the program to your point of sale and customer data, then launch small and measure what members do against customers who did not join.
How to write a loyalty program RFP: requirements to include, security and privacy questions, scoring, scripted demos and the answers vendors prefer to skip.
A good loyalty program RFP describes your program and the problems it must solve, lists requirements by priority, asks the same security, privacy and cost questions of every vendor, and scores answers with weights agreed before responses arrive. Scripted demos built on your own scenarios then test what the written answers claim.
Points, tiers, member pricing or paid membership? How each loyalty model works, what it costs, when it fits your margins and customers, and the common traps.
Choose the model that rewards the behavior you want at a cost your margins can carry. Points reward repeat purchases over time and create a liability; tiers reward your best customers with status and suit businesses with frequent, high-value purchases; member pricing rewards identification at every visit with an instant discount; a paid membership charges a fee for benefits and suits customers who buy often enough to earn it back.
How to measure loyalty program ROI honestly: the full cost, the revenue members bring compared with similar customers, how points are counted, and the metrics.
Loyalty program ROI is the extra profit the program creates, minus everything it costs, divided by that cost. Measure the extra profit by comparing members with similar customers who did not join, not by adding up what members spend. Count the full cost: rewards, the liability for unredeemed points, technology and the people who run the program.
How B2B loyalty programs differ from consumer programs: who earns the reward, which rewards businesses value, tiers, data, common mistakes and a launch plan.
A B2B loyalty program rewards a business account, or the people buying for it, rather than an individual consumer. Purchases are larger and less frequent, several people influence each one, and the buyer usually values money, margin and service more than personal perks. That changes who earns the reward, what the reward is, how tiers are set and how much depends on clean account data.
Build or buy a loyalty platform? When a spreadsheet is enough, what building really involves, how the costs compare over years, and how to move to a platform.
Most programs should buy a configurable platform. Building makes sense only when the program depends on a mechanic no vendor can support and you can fund a dedicated team to run and maintain the system for years. Very small programs may not need a platform yet: a well-kept manual ledger can work until the admin time costs more than a subscription.
How loyalty programs set direct sales, retailers, franchisees and partners against each other, and how to write earning, funding and credit rules that avoid it.
Channel conflict happens when a loyalty program rewards one route to market at the expense of another: a brand's own store against its retailers, corporate locations against franchisees, or one partner against another chasing the same customer. Avoid it by deciding up front which purchases earn, who funds the points and who gets credit for the sale, and by writing those rules down before members and partners find the gaps.
Should your loyalty program launch a co-brand credit card? How brand and issuer split the money, what the card costs you, fit tests and choosing an issuer.
Launch a co-brand credit card only if enough of your members would use it every day, an issuer will pay more for the points it buys than they cost you to honor, and the contract keeps control of your currency and your member data with you. A card can bring steady revenue and engagement, but it also makes your program depend on a bank and adds points you must redeem.
Use the fewest tiers that change behavior. Many programs need none at launch; most tiered programs work well with two or three levels, such as Sephora and Starbucks. More levels suit businesses with perishable capacity and very wide differences in customer value, like hotels and airlines. Add a tier only when you can say in one sentence what behavior it rewards and show that its benefits pay for themselves.
How to estimate loyalty breakage: the formula, the cohort data you need, what makes estimates wrong, how cautious to be, and how often to review the rate.
Breakage is the share of issued points that will never be redeemed. Estimate it from your own history: follow groups of points issued in the same period, measure how many were redeemed or expired over time, and project the share that will never be used. Treat the result as a cautious, reviewable estimate that reduces the expected cost of rewards, not as profit, and update it whenever rules or behavior change.
Should your loyalty program offer lifetime status? What it costs over time, how Marriott Bonvoy qualifies members, and the limits that keep it affordable.
Offer lifetime status only as a rare reward for members who have already brought many years of value, and only after you have costed the benefits for as long as those members may use them. Base it on both cumulative activity and years at a tier, as Marriott Bonvoy does, keep the lifetime tier below your top tier, and write clear terms. For most programs outside travel, annual requalification is the better tool.
How to measure loyalty incrementality: holdout groups, test size and length, calculating lift, alternatives when you cannot hold out, and the mistakes to avoid.
Incrementality is the extra spending or behavior a loyalty program causes, beyond what members would have done without it. You measure it by comparing members who get the program, or a change to it, with a similar group that does not: a holdout group. The gap between the two groups over the same period is the program's lift, and only that gap should count as its return.
How to allocate market development funds: who qualifies, what MDF should pay for, how to approve and pay claims, fairness rules, and how to measure the return.
Allocate market development funds to partners and activities that can create demand you can measure, not as a flat share of last year's purchases. Require a plan before the money is committed, pay only against proof that the activity ran, and offer the program to competing partners on proportionally equal terms. Then compare the sales of funded partners with similar partners who were not funded, and move money toward what worked.
Should loyalty members be able to transfer points? What transfers, pooling and gifting do to breakage, liability and fraud risk, and which controls to use.
Allow point transfers in a controlled form if members have a real reason to combine balances, such as households saving toward one reward, and if you can absorb lower breakage. Open, unlimited transfers between any accounts invite resale and fraud. Most programs that allow transfers cap the amount and number per year, require accounts to be established and in good standing, and verify both sides.
Points or cash back for a loyalty program? How each works, what each costs, how members judge the value, when each fits, and a test to choose between them.
Cash back works better when your category is price-driven and members want a simple, immediate saving they can compare. Points work better when you want rewards other than money, such as products, experiences or partner redemptions, or levers like tiers and bonus events. Points can cost less per dollar of perceived value, but only if members still find them clear and reachable; otherwise they read as a discount with extra steps.
Spiffs vs rebates for channel partners: how each works, when each fits, sell-in or sell-through triggers, the true cost, tax reporting, and a test to choose.
Use rebates as the main channel incentive when you want partners to plan, stock and sell your products steadily over a quarter or a year. Use spiffs for short, specific pushes, such as a product launch or clearing a model, with a fixed end date. In both cases, pay on verified sales to end customers where you can, rather than on the stock partners buy from you.
Run a status match when your own tier is at least as hard to earn as the competitor tier you are matching, when matched members can realistically move spending to you within the trial, and when the expected extra margin covers the cost of the benefits. Grant the status for a limited period, ask members to requalify on your own terms, and measure the result against members you did not match.
How to choose loyalty tier benefits members value and cost them before launch: cost cards, expected usage, perceived value, pilots, caps and yearly reviews.
Choose tier benefits from what your best members already buy and what frustrates them, then cost each one per use, including every direct cost of delivering it. Multiply that unit cost by the number of eligible members and the share expected to use it, and test the total against an upper bound. If you have no usage data yet, run a small pilot before you commit the benefit to the whole tier.
Do unclaimed property laws apply to loyalty points? How escheatment works, why most points fall outside it, which rewards carry real risk and how to manage it.
Usually not for points given free and redeemable only for goods, services or discounts: the model unclaimed property act excludes these loyalty rewards from the property businesses must report. The risk rises when a reward can be redeemed for cash, when members paid for points, or when the reward is a stored-value card. State laws differ, so map your reward types and have counsel confirm your position.
How to name a loyalty program: what a good name does, the main naming styles, naming points and tiers, trademark checks and testing candidates with customers.
Choose a short name that is easy to say and spell, links to your brand and hints at what members get. Name the points currency and the tiers at the same time, so the whole system reads as one. Before you commit, search the USPTO database and the web for similar names in related businesses, then test a shortlist with real customers and staff.
Punch card or points program? How each loyalty model works, what a reward really costs, which businesses suit each one, and when to move from stamps to points.
A punch card gives a set reward after a set number of purchases, so it is easy to explain and suits frequent, low-ticket habits such as coffee or car washes. A points program turns spending into a balance members can save and redeem for a range of rewards, which suits varied baskets and higher spend. Choose a punch card for simplicity and points when you need flexibility, spend-based rewards and better data.
Who should own a loyalty program: what marketing, finance and product each optimize for, the decisions one owner must control, and how other teams fit in.
One person should own the program's economics: the earn rates, the reward budget and the points liability, with the authority to stop a promotion that costs more than it earns. That owner can sit in marketing, finance or a commercial team, as long as they answer for both the revenue and the future cost. Other teams run their parts, but one person makes the trade-off.
How to evaluate loyalty platform vendors: the brief to write first, scripted tests in place of demos, questions to ask, pricing comparisons and the data exit.
Evaluate vendors against the program you will actually run, not against their feature lists. Write a one-page brief with your rules, give every vendor the same test data, and ask each to show the same workflows: an earn, a return, a redemption, a balance correction and a full data export. Compare itemized pricing over several years, check references from programs like yours, and put the results you saw into the contract.
Should loyalty points expire? Inactivity vs fixed-date expiry, how long the window should be, what counts as activity, member notices and what to check first.
Often yes, but on an inactivity trigger rather than a fixed date. An inactivity rule lets points lapse only after a member has gone a set period without earning or redeeming, so active members keep their balance while dormant balances stop adding to what the program owes. Whatever you choose, state it plainly in the terms, warn members before points lapse and check local law before launch.
How to price a loyalty reward catalog: fixed charts or dynamic pricing, who carries the cost risk, pricing rewards from real cost, and checking value per point.
Use a fixed price in points for most rewards, set from what each reward really costs you, so members can trust what their points buy. Keep dynamic pricing for rewards whose cash cost genuinely moves, such as flights or hotel nights, and give it limits. Whichever model you use, calculate the value per point of every reward and make any differences deliberate.
How to validate channel incentive claims and find leakage: which claims to check, what evidence to ask for, automated checks and a reconciliation test.
Check channel incentive claims in proportion to their risk: run automated checks on every claim, require full evidence before paying claims above a value threshold, and audit a sample of the rest after payment. Then look for leakage across the whole program by reconciling what was issued, paid, redeemed and expired against what the rules say should have happened.
Emotional vs transactional loyalty: how each works, what it costs, how to measure attachment beyond spend, and how to decide where your program should invest.
Transactional loyalty is driven by economic value: points, discounts and cash back that make buying from you cheaper. Emotional loyalty is attachment: customers stay because of how you treat them, even when a competitor is cheaper. Most programs need a transactional base, and add emotional elements such as recognition and access where customers care about more than price and staff can deliver them consistently.
The main types of loyalty fraud, from account takeover to promotion abuse, the controls that stop each, what to do when points are stolen, and sizing defenses.
Loyalty fraud falls into a few main types: account takeover, promotion, referral and enrollment abuse, earning on purchases that were never made or were returned, employee abuse, and laundering or reselling points. Build controls where points change hands or turn into value: logins, balance changes, transfers and redemptions. When a member's points are stolen, restore them quickly, secure the account, then investigate.
Join a coalition loyalty program or run your own? How coalitions work, what you gain and give up, settlement and liability, operator risk and how to decide.
Run a standalone program when you have enough purchase frequency and customer data to build your own currency, and want full control of rules, data and costs. Consider a coalition when your customers buy from you too rarely to collect a meaningful balance on their own, or when a coalition already holds the customers you need. If you join, negotiate data access, settlement terms and a clean exit before you sign.
Loyalty program ideas for products bought once every few years: why points struggle, lifecycle memberships, ideas by category and when points still make sense.
When customers buy once every few years, a points balance rarely changes their next purchase, because it sits unused for too long. A better design ties benefits to the product they already own: service reminders, maintenance or warranty extras, accessory discounts and an annual check-in that gives you a reason to stay in touch. Points can work if there is a steady stream of smaller purchases, such as consumables or accessories.
How to design loyalty program enrollment: what to ask at sign-up, consent, channels, duplicate accounts, the welcome journey and the numbers that show it works.
Make joining as short as possible: one identifier, usually an email address or phone number, plus clear consent and an instant confirmation. Tell people what they get in one sentence before they join. Ask for more profile details later, after the member has earned or redeemed, and measure not just sign-ups but how many new members make a first qualifying purchase.
How to design a dealer incentive program: types and ideas, sell-in or sell-through triggers, proof of sale, reward size, fair treatment of dealers and a pilot.
Reward dealers for verified sales to end customers, not for the stock they order from you, and settle the reward against a proof-of-sale record you can check automatically, such as a point-of-sale feed, a serial number registration or a warranty activation. Keep the reward simple and paid on time. If you also reward dealer staff or technicians, tie the reward to quality outcomes they control, never to speed or parts sold alone.
Gamification in loyalty programs: which mechanics change behavior, which are decoration, where game features backfire, and how to test one before you scale it.
Gamification works in loyalty programs when it points members toward a specific behavior and is backed by value they already want: a challenge tied to a reward, a progress bar toward the next one, a season with a clear finish. It rarely works as a substitute for that value. Build the economics first, add game mechanics where they support a behavior you can measure, and test each one against a control group.
Loyalty program cost, line by line: setup, platform, people and reward cost per point, who funds the rewards, and how to defend the budget when finance asks.
Budget a loyalty program in three parts: setup, running costs and the cost of the rewards members earn, which is usually the largest line. Price rewards at the net cost per point, after the share you expect never to be redeemed, and book that cost when points are issued. Then name who pays for each reward: your own margin, a partner, a supplier or a member fee.
Where AI fits in a loyalty program: why balances need deterministic rules, where machine learning helps, safe uses for LLMs, and when to explain AI decisions.
Use deterministic rules for anything that changes what a member holds: points earned, redeemed or expired, and tier status. Use machine learning for predictions a person or a rule acts on, such as churn risk, fraud scores and offer ranking. Use large language models (LLMs) for drafting, summarizing and answering questions, but never to calculate or change a balance, and have people check anything a member could treat as a promise.
Promotion stacking and earning caps in loyalty programs: the margin test, types of cap, stacking rules for points and coupons, and how to add limits fairly.
Cap earning when the cost of the points a member earns can exceed the margin on the spending that earns them, usually in low-margin categories or for a few very heavy earners. Stop promotions stacking by allowing one promotional boost on top of the base earn rate and setting a maximum combined promotional value per member per period. Publish both rules before balances build up, and never apply them to points already earned.
The loyalty program metrics and KPIs to track first: enrollment, activation, active members, redemption, outstanding points, what comes next, and when to act.
Track participation before revenue. Start with enrollment completion, activation (the share of new members who make a first qualifying purchase within a set window), the active member rate over a fixed period, the redemption rate and the value of outstanding points. Once those are stable, add revenue per active member and incremental revenue measured against a control group.
How to set a minimum redemption threshold for loyalty points: why programs use one, what it costs members, how to size it by reward type, and rounding rules.
Set the minimum at the smallest reward that is genuinely worth delivering, not higher. For digital rewards and account credits, that can be a very low balance, because each redemption costs little to fulfill. For physical rewards, the minimum should cover the real cost of fulfillment. A threshold set higher than that does not reduce what you owe members; it only delays it and frustrates them.
Should you run a paid loyalty program or a free one? The break-even test for a fee, trials and billing, subscription loyalty, what to measure once you charge.
Run a paid loyalty program only when it buys a saving or benefit members can calculate and recover with their normal spending, and that a free program cannot offer. If you cannot show in one line how a typical member earns the fee back, keep the program free. A paid program then lives or dies on its renewal rate, so measure that first.
How to design points plus cash redemption: when it helps, how to price the points portion, the costs to check, refunds and split payments, and how to launch it.
Points plus cash lets members pay for a reward partly in points and partly in money. Offer it when many members hold balances too small for the rewards they want, and price the points portion at one fixed, published value per point. Set the cash portion so each redemption still covers its cost, decide whether the cash part earns points, and plan refunds and split payments before launch.
How to choose between a headless loyalty platform and an all-in-one suite: what each includes, what it costs, who owns the member experience, and what to check.
Choose a headless platform when the program has to live inside your own app, checkout and service tools, and you expect to change earn, tier and reward rules often. Choose a suite when standard points and rewards on the vendor's own member screens are enough and you have no engineering team to build and run the integration. Either way, check that you can export balances and history before you sign.
Loyalty program data privacy: who owns member data, what the platform contract should say, how to collect consent, the CCPA incentive notice, and how to audit.
Your member data, transaction history and program rules are the lasting asset of a loyalty program; the platform that stores them can be replaced. Secure that asset in the platform contract with clear ownership, export and exit terms. Use the data only on a lawful basis: what members need to run their account, plus separate, recorded consent for marketing, and honor members' privacy rights without taking away the benefits they are entitled to.
A loyalty program launch plan: checks before go-live, soft or full launch, staff training, launch messages, and what to fix and measure in the first 90 days.
Treat the launch as a controlled test. Before go-live, lock the rules, test every earn, redemption and reversal path, train staff and prepare support. Start with a soft launch for a limited group if you can, then announce widely with a short message sequence. In the first 90 days, reconcile the points ledger, watch activation and support issues, and change rules only to fix errors.
Which loyalty program rules change across borders: point expiry, marketing consent, changing terms, dormant balances, and running one program in many markets.
Loyalty program law changes by region on a few recurring questions: whether points can expire, how marketing consent must be collected, whether you can change the terms on your own, and what happens to dormant balances. The questions stay the same from market to market, but the answers do not. Map them for every market where members join, earn or redeem, and confirm the current rules with local counsel.
How to write loyalty program terms and conditions: what to cover, what to disclose before members join, expiry and change clauses, and readable wording.
Write loyalty terms that a new member can understand without a lawyer: how points are earned and redeemed, what they are worth, when they expire, what does not count, how data is used and how the program can change. Put a plain summary of the key numbers next to the join button, keep the full terms one click away, date every version and have counsel review them before launch.
Elite status inflation for program owners: what causes it, how to tell if your tiers are inflated, and how to restore scarcity without losing top members.
Tier inflation happens when so many members reach a tier that it stops signaling anything and its benefits become a general discount. It is caused by thresholds that stay fixed while prices rise, benefits added without higher thresholds, lowered thresholds and status handed out through promotions. Fix it by restoring scarcity in the existing tiers, with notice and a soft landing, before you consider adding a new level on top.
How to read a loyalty points liability report: the roll-forward, cost per point, breakage, the useful breakdowns and the signs liability is growing too fast.
A points liability report shows what the program owes members for points earned but not yet redeemed. Read it as a roll-forward: opening balance, plus points issued, minus points redeemed and expired, equals the closing balance. Then check the cost per point and the redemption assumption that turn that balance into dollars, and watch whether issuance is outpacing redemption.
How loyalty points are calculated and how to set earn rates: the share of spend you can return, points per dollar, category margins and testing before launch.
Start from the share of spending you can afford to give back, then divide it by the value of a point to get points per dollar. Keep one simple, whole-number rate for most purchases, lower it or exclude categories with thin margins, and raise it only where a test shows the extra points bring extra profit. Check the cost at your expected redemption rate before you publish anything.
How to set the value of a loyalty point: calculate cents per point, pick the reference reward, small vs large points, and keeping the value stable over time.
A point is worth the value of what it buys divided by the points needed. Pick one reference value based on the reward most members actually use, publish it in plain terms such as "500 points = $10", and set the earn rate from it. If your budget is tight, give fewer points rather than weaker ones, and keep the value stable once members rely on it.
How to set loyalty tier thresholds from your margins and member data: spend or activity, the break-even math, qualification periods, rollover and review.
Set each tier threshold at the level where the extra margin from a qualifying member pays for the extra benefits that tier gives. Work from contribution margin after variable costs, then check the result against your member spend distribution so the tier holds the share you intend. Most programs qualify on spend, measured over a calendar or membership year, and review thresholds every year.
Points devaluation explained: what counts as one, why programs devalue, when it is justified, how to protect existing balances, and how to give fair notice.
A devaluation is any change that makes points buy less: higher reward prices, lower earn rates, shorter expiry or fewer rewards. Do it only when the program's cost genuinely outgrows the margin that funds it, and first try changes that apply to future earning only. Protect balances members already hold where you can, give clear notice with the old rule, the new rule and the reason, and avoid repeated cuts.
How to handle loyalty tier downgrades: the difference between a soft landing and a hard reset, what each costs, which members to protect, and how to tell them.
Use a soft landing, dropping members one level at a time or keeping their tier for one more cycle, when the tier still exists and members earned their status in good faith. It costs some benefits for a year but protects members who are likely to requalify. Use a hard reset when the old tier no longer exists, when the benefit can no longer be offered, or for members far below the new requirement.
Are loyalty programs worth it for your business? Four tests on frequency, margin, identity and share of wallet, what loyalty cannot fix, and cheaper options.
A loyalty program is worth building when customers buy often enough to reach a reward, your margins can fund a reward they notice, you can link purchases to a person, and customers still spend part of their budget with competitors. If any of these fails, the program is unlikely to pay for itself. A simpler incentive, such as a stamp card or member prices, is often the better first step.
How far a personalized loyalty program should go: levels from segments to one-to-one rewards, when it pays, where it intrudes, privacy and how to test it.
Personalize as far as it pays and as far as members would find it reasonable, and no further. For most programs that means simple segments and event triggers for everyone, with one-to-one offers reserved for members whose spend and visit frequency can repay the cost. Every step up should beat the simpler version in a test with a holdout group, and every offer should be one you could explain to the member who received it.
Loyalty platform migration planning: what breaks when you switch platforms (balances, expiry dates, tiers, refunds) and how to run the cutover and notice.
A loyalty migration is a change to a promise, not just a data move. What breaks most often is the value of members' points and their tier status, followed by refunds and partner activity that straddle the cutover. Plan it by deciding first whether balances move at the same value or are settled before the switch, then run the old and new ledgers side by side until they reconcile, and tell members early.
Loyalty program profitability in practice: reward cost per dollar, margin after rewards, the sales lift you need to break even, and how to fix the leaks.
A loyalty program eats margin when the cost of the rewards members earn is larger than the extra profit the program creates. Convert your earn and reward rules into a cost per dollar spent, compare it with gross margin category by category, and test whether members would have bought anyway. Then fix the specific categories and rules that fail, rather than cutting the whole program.
Real-time loyalty or batch processing for points? How each posting method works, what members notice, what the ledger needs, and where batch still fits.
Post points in real time for any purchase where members can see, redeem or transfer their balance right away, because a missing balance after a purchase looks like a broken promise. Keep batch processing for partner settlement, internal adjustments and period-end calculations. Real-time posting only works if the ledger can accept each event safely once, so check that before you promise instant points.
Loyalty program reporting for the board: a one-page report template, growth and liability side by side, the cohort payback test, and when to change course.
Report the program as an investment with a cost the board can see: the incremental margin it creates next to the points liability and the expected cost of redeeming it. Show results by member cohort, so the board can judge whether points issued to a group of members earn back their cost. Agree in advance the signal that would turn the report from growth to cost control.
How to build loyalty program benchmarks against competitors: what you can see, how to compare reward values fairly, a template, and which numbers not to copy.
Benchmark what members can see: how fast they earn, what a point is worth, how long a first reward takes, tier thresholds and expiry rules. Collect these from each competitor's published terms, convert them into comparable figures such as value returned per $100 spent, and date every entry. Do not copy a competitor's earn rate into your own program; set yours from your own margins, because you cannot see how theirs is funded.
How employee loyalty fraud happens, from sweethearting to manual credits, the controls that prevent it, how to detect the rest, and how to keep service fast.
Treat points like cash and design the controls before launch. Separate the people who create points from the people who approve them, cap how many points any one employee can issue, require a reason code for every manual change, and keep an audit log no employee can edit. Then review exception reports regularly, because controls reduce fraud but do not catch all of it.
How to end a loyalty program: when closing is right, what it costs, a wind-down timeline, what the announcement email says, and the legal and accounting checks.
End a program when it no longer changes member behavior enough to cover its rewards and running costs, and a fix would not change that. Close it in stages: stop earning on a set date, give members a fair window to redeem, extend it for large balances, then close redemptions. Check your terms and local law first, agree the accounting with finance, and tell members early and plainly.
Loyalty program relaunch or fix? What each path costs, a test for which problems are fixable, recent relaunch examples, and how to change without losing trust.
Fix the program first in most cases. A fix changes rules inside the existing relationship and keeps members' balances; a relaunch closes that relationship and starts a new one, usually with a new name or converted balances. Relaunch only when the core promise can no longer be funded, the program's name has become a liability, or a legal change forces a reset. Old technology or a large liability alone are problems a fix can solve.
How loyalty points fraud works through laundering and reseller abuse, which features create the risk, the controls that close exit routes, and what to do next.
Points laundering turns stolen or illicit money into clean value by buying points, moving them between accounts and redeeming them for rewards that can be resold. Reseller abuse is a related leak: trade buyers earn consumer rewards on stock bought for resale. Both are controlled by mapping how value enters and leaves the program, limiting transfers and high-value redemptions, and deciding reward eligibility on each transaction rather than on each customer.
Loyalty partnerships that sell points to partners and card issuers: deal types, the floor price, breakage, what the earn agreement must fix, and when to refuse.
Sell points to a partner only at a price above what each point will cost you to honor, including servicing, and only when the partner brings members or spending you could not reach as cheaply yourself. Put the earn rate, price, settlement timing, volume and who keeps breakage in a written agreement before any points are issued. A sale below cost brings cash today and a larger bill later.
