Cashback App Case Study: Desconto Premiado Loyalty Build

Most consumer fintech founders who pitch a cashback app are building the wrong half of the product. The consumer app is the easy half. The merchant side is the graveyard. Desconto Premiado is a case study in how to ship a cashback and loyalty app that actually closes the merchant-side math for local commerce, not just for e-commerce checkout.

If you are a consumer fintech PM, a loyalty program lead at a local-commerce platform, or a founder evaluating a regional cashback build, this post walks through the architecture, the engineering tradeoffs, and the one section most loyalty posts skip entirely: why merchants renew or churn.

The US analogs (and where they leave a gap)

Every US reader has a mental map for cashback apps. Rakuten pays you for shopping at online retailers. Ibotta and Fetch reward grocery receipts. Drop and Dosh link a card and credit back on eligible purchases. Upside handles fuel and restaurants with location-based offers. RetailMeNot runs coupon-plus-cashback overlays.

That lineup looks saturated until you look at what they actually serve. Four of them are e-commerce heavy. The rest optimize for national chains with centralized loyalty APIs. The independent coffee shop, the neighborhood hardware store, the family-owned pharmacy, the three-location regional apparel brand - those merchants are functionally invisible on Rakuten. They are the entire addressable market for a regional cashback and loyalty app.

Desconto Premiado targets that gap. It is a consumer mobile app plus a merchant admin panel that rewards in-store and online purchases at independent local merchants, surfaces them by geolocation, and lets the merchant control their own cashback percentage, campaigns, and sweepstakes.

The problem - savings plus retention, at the same time

Two jobs on one platform:

• Consumer job: save passively, without hunting coupons. Link the app to a local merchant network and let the app surface where your money already goes.
• Merchant job: retain customers without buying an enterprise loyalty stack. Most independent merchants cannot afford a six-figure annual license for a CRM and loyalty suite, and they do not have the staff to operate it.

Global cashback apps collapse both jobs into one e-commerce-centric flow: shop online, click through, earn. That flow is broken for a coffee shop on Main Street. No affiliate link, no centralized API, no analytics team. Regional operators who ignore the online-first playbook and build for physical, independent merchants carve out the category that Rakuten will never reach.

Product concept - Desconto Premiado

The product is intentionally dual-sided from day one. Consumer and merchant were designed in parallel, not as phase 1 and phase 2. Core surfaces:

Consumer mobile app

• Automatic cashback on in-store and online purchases at partner merchants
• Points-based loyalty with clear redemption (discounts, prizes, experiences)
• Geolocation map of partner merchants, radius search, category filters
• Digital wallet with full transaction and points history
• Sweepstakes and rewards mechanics to drive re-engagement
• Push notifications for earned points, nearby offers, and campaign activations

Merchant admin panel

• Merchant self-service for cashback percentage, offers, and campaign calendars
• Transaction confirmation flow for in-store purchases
• Reporting on new customer attribution, repeat visits, and redemption rate
• Sweepstakes creation and prize management
• Settlement view tied to the payment gateway

The core insight: a merchant who cannot see ROI in their panel within the first 90 days will churn. The panel is not an afterthought, it is the retention product.

Architecture overview

Lean stack, chosen for speed to ship and small-team operability:

Data flow at the simplest level: customer transacts at a partner merchant, merchant confirms the transaction in the panel or through an integration, backend posts a credit to the loyalty ledger, push notification fires, user redeems against a campaign or a direct discount at the merchant.

Engineering tradeoffs that actually mattered

1. Loyalty ledger correctness

The most common failure mode in loyalty software is a mutable balance field. Someone double-credits a purchase, a refund reverses incorrectly, a disputed charge does not reflect in points, and the customer sees their balance change with no explanation. Trust dies there.

We modeled points as an append-only ledger. Every credit and every debit is an immutable entry with an idempotency key tied to the source event. Double-posting a purchase becomes a no-op. Reversals are new entries with explicit reason codes, not edits to prior rows. The balance is a projection, not a stored value.

This is the same discipline fintech teams use for fiat ledgers. If you are building loyalty, steal it. Our fintech app build guide covers the ledger pattern in more depth for US regulatory contexts like PCI-DSS, SOC 2, and state money transmitter considerations.

2. Payment gateway split - who actually funds the cashback

This is where most loyalty apps quietly break. Cashback has to be funded by someone. Three models:

• Merchant-funded: the merchant absorbs the cashback as a marketing expense. Platform takes a take rate on the transaction or a flat subscription. Cleanest economics, hardest sell to merchants without proven ROI.
• Platform-funded: the platform eats the cashback and recoups on advertising, sponsored placements, or data. Requires scale and a separate revenue engine. Fragile at launch.
• Shared: split between merchant and platform. Commonly used with a tiered structure (new customer acquisition cashback funded partially by platform, repeat customer cashback fully by merchant).

Gateway-side splits push the accounting to the payment processor, which is cleaner for reconciliation but locks you into that gateway. Platform-side reconciliation gives you flexibility to swap gateways but means you own the correctness problem. We chose a gateway-aware split with platform-side verification - the gateway performs the split, the platform ledger verifies every entry against the expected split ratio within 24 hours. Any mismatch flags for manual review before settlement closes.

3. Anti-abuse rules

The abuse vectors on a cashback app are predictable:

• Sock-puppet accounts - one consumer creates five accounts to multiply new-user bonuses
• Merchant collusion - a merchant rings up fake transactions to hit volume tiers or to trigger sweepstakes
• Refund fraud - purchase, claim cashback, refund the purchase outside the app so the cashback is never clawed back

We did not build an ML fraud engine at MVP. We built rules: device fingerprinting, velocity caps on new-user bonuses per device and per payment instrument, refund-event webhooks from the gateway that trigger automatic ledger reversals, and a merchant transaction-pattern dashboard that flags outliers for human review. Rules scale to hundreds of thousands of transactions. Move to ML only once you have labeled fraud data.

4. Geolocation discovery performance

Merchant map queries are the single hottest path in the consumer app. A naive PostGIS radius query on every pan and zoom kills the database under even modest load. We precompute nearby merchant lists at the neighborhood grid level on a nightly job and cache them. Live geospatial queries run only for custom radius filters and for new merchant onboarding. Map render targets sub-300ms at the p95 on mid-tier Android devices.

Merchant-side economics - the section every other cashback post skips

The app is the easy part. The take-rate model is the hard part. If a merchant does not see payback within roughly three months, they churn. That rule holds across every loyalty SaaS benchmark we have seen.

Four levers drive merchant ROI. Every cashback and loyalty platform lives or dies on tuning these:

• Take rate: what the platform charges the merchant per transaction or per month. Too high, payback is impossible. Too low, platform unit economics break. The defensible range for local-commerce cashback sits below enterprise loyalty take rates but above pure ad-network rates.
• Redemption rate: the percentage of earned points that users actually redeem. High redemption increases customer spend per visit but erodes merchant margin. Redemption caps, expiration policies, and category-level multipliers are the controls.
• New-customer attribution: the percentage of transactions from customers acquired through the platform versus existing customers earning cashback on visits they would have made anyway. Without attribution, the merchant is subsidizing loyal customers. Geofenced first-visit tracking and new-account attribution windows are how you prove the number.
• Repeat purchase lift: the delta in visit frequency among app users versus non-app customers. This is the number merchants care about most. It is also the hardest to measure without matched cohorts.

Desconto Premiado's partner merchants saw an average 25% increase in customer return. That number is the output of those four levers working together, not a headline bought with discount inflation. Replicating it in another market means tuning take rate, redemption, attribution, and lift for that specific merchant base - not copying the percentage.

Contrast this with the three-sided marketplace math in our last-mile delivery case study, where platform economics are dominated by courier supply and order density. Loyalty platforms live or die on merchant payback math instead.

Results

Using only the verified metrics from the shipped product:

• Thousands of active users in the launch region
• Hundreds of partner merchants registered
• Average 25% increase in customer return for partner merchants
• 4.5+ star rating on the App Store and Google Play

Those are the numbers, and they are the only numbers. They are enough to validate the pattern.

What this means for US consumer-fintech builders

Rakuten-for-local-commerce is a crowded category on the surface and a nearly empty one in practice. The path to winning is not a better consumer UX - the existing players already have that. It is winning on merchant-side economics: a lower take rate, a better merchant panel, faster payback visibility, and attribution that holds up when the merchant scrutinizes the numbers.

A dual-surface consumer fintech with a cashback engine, a loyalty ledger, basic geolocation discovery, and a merchant panel fits inside a 60 to 120 day MVP envelope. Our 2026 app development cost guide for US companies walks through how that range translates into fully loaded team cost, and our mobile app development playbook covers the sequencing. If the product wraps a broader ecommerce build, see custom ecommerce vs marketplace for the make-or-buy call. And if the go-to-market leans heavily on push and re-engagement (it should), push notifications best practices is a short, useful read.

For adjacent patterns built on similar dual-sided or multi-surface foundations, read our laundry management app case study (vertical SaaS for service SMBs) and our B2B procurement AI case study (marketplace orchestration with an AI layer). Different verticals, same discipline of designing for the paying side from day one.

Nearshore execution note

FWC delivered Desconto Premiado as a small senior team with Brazil-hour overlap with every US time zone, which matters when the merchant panel needs same-day iteration with a US product owner. If you are evaluating a build for a regional cashback or loyalty app, a nearshore team with fintech depth is the shape of team that ships this well. We cover our approach in the Brazilian nearshore app development guide and the US fintech app build guide.

Thinking about your own cashback app build?

If you are scoping a regional cashback or loyalty product and you want a pragmatic conversation about architecture, merchant economics, and cost, we are easy to reach.

• Request an app quote - scope, timeline, and team shape tailored to your product.
• Contact us - short-form message if you are still in discovery.

A cashback app is not a winner because the consumer side is polished. It is a winner because the merchants renew. Build for that from week one and the 25% customer return lift we saw at Desconto Premiado is a template you can take to your own regional market.