Points that nobody tracks by hand anymore:
and referrals that pay out on their own
A loyalty program that runs on a spreadsheet works for the first fifty customers and falls apart after that: points go uncredited, referral payouts get forgotten, and someone has to manually check that a reward was actually earned. We build an agent that tracks the ledger, applies your tier rules and pays out referrals with a fraud check in front of every reward.
The process today
A loyalty program is simple to describe and tedious to run by hand: credit points on a purchase, move a customer up a tier when they cross a spending threshold, and pay a referral reward when a new customer completes their first order. The problem is volume and timing. Teams running this manually usually fall behind within the first few hundred orders, because crediting points is not the kind of task anyone does the moment it is earned, it is the kind of task that gets batched into a Friday afternoon, by which point a customer has already noticed their balance is wrong and written in to complain.
Referral programs carry a second problem on top of the delay: fraud. A reward paid per signup attracts the same account creating ten “referrals” from different email addresses on the same device, and a program with no automated check either pays all of it out or a person has to manually eyeball every referral before it clears, which defeats the point of having a self-serve program at all. Industry fraud benchmarks for referral programs commonly put abusive signups at 5 to 15 percent of total referral volume when there is no automated screening in place.
What the agent does
Credits points automatically on qualifying orders, applying your rules for what counts (full price, discounted, excluding gift cards, whatever your program defines) the moment an order settles.
Applies tier upgrades and downgrades on schedule, moving a customer up when they cross a spending or points threshold and down when a rolling period rule says they should drop, with the change reflected everywhere the tier matters (checkout discounts, early access, support priority).
Tracks referral links and attributes them correctly, matching a new signup or order to the referring customer even when attribution has to survive a browser switch or a delayed first purchase.
Runs a fraud check before any referral payout, flagging the same device, the same shipping address, or a burst of signups from one source for a human to review before the reward clears.
Handles redemption at checkout or through a messenger bot, so a customer can see and spend their balance without opening a support ticket.
Applies expiry rules to points and unused referral credit, and reports monthly on what the program cost against the revenue it drove, so the economics stay visible instead of assumed.
What stays with humans
The agent runs the ledger, it does not set the program’s economics. Any change to the points-to-currency ratio, the tier thresholds or the referral reward amount is a business decision that stays with you. Every payout above a value threshold and every referral flagged by the fraud check waits for a person, and a one-off exception, like honoring a loyalty balance after a policy change, is a judgment call a human makes, not the agent.
Guards
A dry-run period where the agent calculates every credit and payout but holds them in a review queue, so you can confirm the rules match your actual program before anything moves automatically. After that, every ledger change and payout logs the rule and the order behind it, a value threshold and fraud-pattern check route the larger or riskier payouts to a human, rate limits stop a sudden spike of signups from draining the referral budget before anyone notices, and a kill switch pauses crediting or payouts on the whole program instantly.
Price and timeline
| Package | Price | Best for |
|---|---|---|
| Single automation | from $1,200 | One points structure and one referral mechanic, wired into your store and payment system |
| Department package | from $2,500 | Loyalty and referrals plus customer segmentation and lifecycle messaging |
6 to 12 days covers encoding your tier and referral rules, wiring attribution correctly, and the dry-run review period before payouts run unattended.
Related
Works well alongside customer segmentation, churn alerts and push and email lifecycle messaging, since all three touch the same customer ledger. Part of automation of everything digital and built the way we build AI agents for our own products. Reward balances and crypto-adjacent payouts are exactly the kind of thing our own ProBay marketplace runs a margin guard against, and the referral-adjacent promo engine we built for a D2C store in Thailand covers eight separate promo mechanics on one catalogue; our marketplace engine handles promo codes the same way, with 223 automated tests behind it.
Tell us your tier structure and referral mechanic and we will send back a fixed price and a plan for the first week: get in touch.
Tired of doing this by hand? We can take the whole routine off your team, not just this step: Routine takeover, from $400 →
FAQ
How much does a loyalty and referral automation cost?
A single program, one points structure and one referral mechanic, starts from $1,200. A department package covering loyalty plus segmentation and lifecycle messaging starts from $2,500, priced once we know your tier structure and expected volume.
How long does it take to go live?
6 to 12 days: encoding your points and tier rules, wiring referral attribution to the right order and customer, and a dry-run period where payouts are calculated but held for review before the agent pays them automatically.
Which tools does it connect to?
Your store backend (Shopify, WooCommerce or similar) for order data, your CRM for customer records, your messenger or app for redemption, and your payment or store-credit system for the actual payout. Referral links are tracked through your existing attribution setup or a dedicated link layer we add.
What if the agent credits a fraudulent referral?
Every payout above a threshold you set, and every referral that matches a fraud pattern (same device, same address, a burst of signups from one source) is held for a human to check before it pays out. The agent explains which signal triggered the hold, so a reviewer can clear a false positive in seconds.
Is customer and payout data safe?
The agent reads only the order and referral data needed to calculate a balance, runs under your existing store and payment permissions, and we do not retain customer data outside the ledger pipeline. Access can be revoked at any time.