A marketing director at a retail chain with a few hundred stores receives three proposals for a loyalty app. The first is EUR 65,000, the second EUR 210,000, the third EUR 560,000. All three have attractive mock-ups: a points screen, coupons to activate, a birthday prize wheel. None of them answers the question the operations director will ask in the first meeting: what happens at the till at 11:00 on a Saturday when a customer shows a code and the store’s connection is down.

The spread is not because one vendor is greedy and another naive. It is because each one priced a different thing. The cheapest priced the app, meaning the screens. The middle one priced the app plus a points backend. The most expensive priced a loyalty programme, which also includes POS integration, price history, consent management and a back office for marketing. Only the third proposal describes something you could actually run in stores.

This article is for marketing and e-commerce directors in retail facing that decision. The argument is simple and may sound contrarian: the loyalty app is the cheapest and least risky part of a loyalty programme. The cost and the risk sit in integration with point-of-sale (POS) and ERP systems, in the rules engine, in promotional pricing that complies with the EU Omnibus rules, and in marketing consent. So budget from the till outwards to the phone, not the other way round, and build your own app only when there is a feature a digital wallet pass cannot deliver.

Does a retailer actually need its own loyalty app?

Not always, and that is the first thing to settle before you issue an RFP. A loyalty programme needs three things: a way to identify the customer at the point of purchase, a place where customers see their benefits, and a channel for sending them offers. An app is one possible way of delivering the second and third. The first, identification, comes from POS integration, and you need that whatever the customer shows at the till: an app code, a plastic card, a pass in Apple Wallet or Google Wallet, or a phone number.

A loyalty pass in a phone wallet already does a lot. Google’s Wallet API documentation describes loyalty cards that show points and rewards directly on the phone, can be issued from an app, a website, an email or an SMS, and are updated through a REST API. Apple Wallet supports an equivalent pass type. For the customer, it is a card that is always with them, without installing yet another app. For the retailer, it is a channel with no app store publishing overhead and no second codebase to maintain.

Your own app wins in three situations. When customers are meant to do something in it, not just show a code: activate coupons, browse offers, order for in-store pickup, scan products. When the retailer wants a direct notification channel it controls, rather than one mediated by the wallet provider. And when the app is meant to replace the paper receipt and the weekly flyer, becoming an everyday tool rather than a card in a drawer. If none of these applies to your chain, start with a wallet pass and spend the money on POS integration.

Apple’s rules add another argument. Section 4.2 of the App Store Review Guidelines requires apps to include features, content and UI that elevate them beyond a repackaged website, and 4.2.2 says apps should not primarily be marketing materials, advertisements or a collection of links, with catalogues as the exception. An app that only shows a barcode and a PDF flyer risks rejection and, more to the point, deletion after the first week.

Which layers make up a loyalty programme, and where does the money go?

A loyalty programme that works across physical stores and online has six layers. Each has a different cost, a different owner and a different risk. Proposals that disagree usually differ in how many of these layers they cover.

Identification at the till. An app code, a wallet pass, a plastic card or a phone number, read by the POS and passed to the loyalty engine. This is where you decide whether a benefit is applied before the receipt closes or after the fact.

Rules engine. Earning points, tiers, coupons, redemption, expiry, returns and adjustments. This is the heart of the programme and the place where marketing wants the most freedom.

Price and product data. Regular prices, promotional prices, the price history Omnibus requires, product groupings. The source is usually the ERP or a pricing system.

Customer profile and consent. Personal data, purchase history, marketing consents with date and wording, channel preferences. This is where GDPR and ePrivacy rules apply, and in the US the TCPA, CAN-SPAM and state privacy laws.

Customer channels. Mobile app, wallet pass, website, email, SMS, push notifications.

Back office for marketing and customer service. Campaign configuration, segments, reporting, handling points disputes.

Compare proposals against this list and you will often find the cheapest covers only the customer channels and part of the rules engine, assuming the remaining layers are “on the client side”. That is fair if it is stated plainly. The problem is that on the client side nobody usually has the time or skills to build integration with several POS systems in a quarter. It helps to describe the till process before you order software; we show how in our piece on modelling a process in BPMN before ordering software.

Which loyalty app features earn their keep, and which are decoration?

A feature earns its keep if it changes customer behaviour in a way you can see in sales data: more frequent visits, larger baskets, purchases in a new category, lapsed customers coming back. A feature is decoration if it lifts downloads or time in app but does not show up in receipts. Seen this way, features fall into four groups.

The core, without which the programme does not work. Identification at the till, points balance and transaction history, coupons with activation, redemption, sign-up and login, consent management and account deletion. Account deletion is not a courtesy: App Store guideline 5.1.1(v) requires any app that supports account creation to offer account deletion within the app.

Features that usually pay off. Personalised coupons based on purchase history; a digital receipt in the app (customers get a reason to scan at every purchase and the retailer gets a fuller history); a shopping list linked to offers; reminders about expiring points; in-store pickup of orders placed in the app. If the app takes payment for those orders, you also need a payment provider and integration, which we cover in our guide to choosing and integrating a payment gateway.

Features that pay off conditionally. A digital flyer (useful when tied to coupons, not when it is a PDF); geolocation and proximity notifications (pay off less often than promised, need extra consent and annoy customers); a product scanner with price and reviews; a referral scheme.

Decoration. Games, prize wheels, badges, leaderboards, elaborate profiles, brand chat. That does not make them worthless. It means they should not compete with POS integration for the first release’s budget.

A simple test for every item on the requirements list: which sales metric should change, by how much, over what period, and how will we measure it. If the answer is “engagement”, the feature goes to release two. The same test protects you from over-engineering the visual layer; we cover when a custom design system pays back and when it does not in our article on design system cost and payback.

How do you integrate the app with POS, and why is it the most expensive part?

POS integration is the most expensive part because it sits at the junction of two worlds with different requirements. The app and the loyalty engine live in the cloud, ship every two weeks and tolerate a second of latency. The till lives in the store, is updated rarely, often by a third-party vendor, and has to serve a customer in a dozen seconds, including when the network is down. On top of that, a retailer with history usually runs more than one POS: one in company-owned stores, another in franchises, a third in acquired chains.

The first decision is when benefits are applied. In the online, before the receipt closes model, the till sends the basket to the engine, the engine returns discounts and points, and the till prints a receipt with the discount. This gives the best customer experience and is the only correct route for coupons that reduce price, but it needs an engine response measured in hundreds of milliseconds and a fallback mode. In the after the fact model, the till sends the receipt once the transaction closes and the engine awards points asynchronously. It is simpler and resilient to outages, but rules out discounts at the till. Many retailers combine the two: coupons online, points after the fact.

The second decision is offline behaviour. What does the till do when the engine does not answer? The common answers are: accept the customer ID, skip coupons and send the receipt for later settlement; or keep a local copy of the customer’s active coupons and apply them itself, accepting the risk of double redemption. Each is a business decision, not a technical one, and should be made before pricing.

The third decision is returns and adjustments. A customer spends EUR 120, earns points and a coupon, then returns half the goods. The engine must reverse points proportionally and either restore or forfeit the coupon used on the returned item. Returns in a different store, returns without a receipt and returns after points have expired take one line in a specification and weeks in testing.

The fourth decision is the integration model. If the POS software vendor provides an interface for loyalty systems, integration means using and testing it. If not, the POS software needs a change that its vendor will scope, price and deliver on its own release schedule. That item is the one most likely to push go-live back by a quarter. So the first question in the project is not “which app technology” but “what can our POS do and who changes it”. General principles for pricing such connections are in our glossary entry on systems integration; in a loyalty programme, integration typically covers POS, ERP for prices and products, the online store, email and SMS tools, and analytics. Every external call also goes through a defined API, and the contract of that API is worth agreeing before any app screen is drawn.

The fifth issue is fiscal and receipt rules. In many EU countries, including Poland, Italy and Germany, tills are subject to fiscal or anti-tampering requirements that govern what the receipt shows and how transactions are recorded. A discount applied at the till must appear correctly on the receipt. The details depend on the country and the POS model, so the integration team needs someone who knows your actual POS software, and testing has to happen on real devices, not just a simulator.

What do the Omnibus price rules change for member promotions?

More than most specifications assume. Article 6a of Directive 98/6/EC on price indication, added by the Omnibus Directive (EU) 2019/2161, requires that any announcement of a price reduction state the prior price, meaning the lowest price applied during a period of not less than 30 days before the reduction. Member states have applied it since 28 May 2022. How it applies to loyalty schemes is a matter of national interpretation, so check guidance in each market you operate in.

One of the most detailed national documents comes from Poland. The President of UOKiK, the Polish competition and consumer protection authority, published guidance in 2023 titled “Information on price reductions” with a separate section on loyalty programmes. Its general rule: member discounts on specific products or product groups, whether announced generally (in the app, on the website) or individually (for example by email to selected members), are price reductions and must show the lowest price of the previous 30 days. The obligation does not apply to discounts not tied to specific products: a discount on the next order after a number of purchases, redeeming accumulated points for a product, or cashback for members with a certain tenure. The number of members and the number of products covered are irrelevant.

Three details from that guidance matter most for system design, and they are useful even outside Poland as an example of how strictly a regulator can read the rule. First, in physical stores the prior price must be on the shelf label. UOKiK considers it insufficient to show it only in the app if a discount activated in the app is used in a physical store. A coupon in the app therefore implies a change on the shelf, which means integration with label printing or electronic shelf labels.

Second, the member price becomes a reference point for later reductions. In UOKiK’s examples, if a store offered a product at 20% off with a loyalty card, that price is the reference point for subsequent reductions of the product in that store over the next 30 days, including for non-members, and the retailer may state that the lowest price applied within the loyalty programme. The pricing system must therefore store price history that distinguishes general and member prices, and the loyalty engine must feed it.

Third, channel matters. UOKiK treats an app through which customers order and collect in store as a separate sales channel: the promotional price in the app is not the reference point for reductions in that physical store. Price history should therefore be kept per channel, and for a chain, national flyers require the lowest price across the whole network.

In practice, a loyalty programme budget needs a line item called “price history and 30-day lowest price per product, channel and price type”, owned by the ERP or pricing team. If you already calculate this for flyers and the online store, it has to be extended to member prices. If you do not, it is probably the largest line after POS integration.

A loyalty app is first and foremost a personal data machine: a named person’s purchase history, location, preferences, sometimes data about children (birthdays) or health (pharmacy or drugstore purchases). So design consents and legal bases at the start, together with your data protection officer, not at the end when the screens are finished.

In the EU, electronic marketing by email, SMS or push is governed by national laws implementing Article 13 of the ePrivacy Directive 2002/58/EC, which generally require prior consent, with a narrower “soft opt-in” for existing customers by email in many countries. Article 5(3) requires consent for storing or accessing information on the user’s device unless strictly necessary for the service. That matters for analytics and advertising SDKs that development teams like to add to apps “just in case”. Member states implement these rules differently; Poland, for example, moved them into its Electronic Communications Law, in force since November 2024.

GDPR adds its own requirements. Consent must be as easy to withdraw as to give (Article 7(3)), so the app needs a consent screen, not just a link to a form. Customers have the right to object at any time to processing for direct marketing, including profiling (Article 21(2) and (3)). You also need a legal basis for running the programme itself (usually contract, through the programme terms) and a separate one for profiling offers. Personalised coupons based on purchase history are profiling and must be explained in the privacy notice.

If you operate in the US, three sets of rules shape the design. The Telephone Consumer Protection Act requires prior express written consent for marketing text messages sent with automated systems, so SMS opt-in needs its own clear checkbox and record. CAN-SPAM requires a working opt-out in every commercial email and honouring it promptly. In California, the CCPA regulations treat loyalty programmes that offer benefits in exchange for personal information as financial incentives and require a notice of financial incentive explaining the terms and how the value of the data was estimated. Other states with comprehensive privacy laws have their own loyalty programme provisions, so check each market.

What does this mean for the system? A consent register storing wording version, date, channel and source (app, till, website), available to the messaging system before every campaign; a consent management screen in the app; account deletion with a defined fate for transaction history (anonymisation rather than deletion if receipts must stay in the books); and a list of third-party SDKs in the app with an assessment of which need consent. These are features to be priced, not legal paperwork.

Wallet pass, PWA or native app: which channel should you choose?

Choose the channel after you have decided which features should earn money, not before. The four options differ in build cost, running cost and what customers can do with them.

ChannelWhat the customer can doBuild cost (relative)Running costEnough when
Apple / Google wallet passShow a code, see points and rewards, receive pass updatesLowestLow, no app store publishingPoints programme without coupons to activate or in-app purchases
Progressive web app (PWA)Account, coupons, history, consents; limited notifications and device accessLow to mediumMedium, one codebaseRetailers with a strong online store that already has customer accounts
Cross-platform app (e.g. Flutter, React Native)Full feature set, push notifications, scanner, ordersMediumMedium, one codebase, two app storesMost retailers that need an app
Two native apps (iOS and Android)Full feature set, best OS integration, widgets, hardware featuresHighestHighest, two codebasesVery large customer base and OS-dependent features

A wallet pass does not compete with an app; it complements it. Even a retailer with its own app benefits when a customer who will not install it can add a pass to their wallet from a link texted at the till. It is a cheap way to raise the share of identified transactions, and that share, not download numbers, determines the value of programme data.

Between cross-platform and native, the deciding factors are usually the team that will maintain it and OS-dependent features. For a typical loyalty app, cross-platform technologies such as Flutter or React Native are sufficient. For benchmarking proposals, we have collected mobile developer rates in Poland for 2026. Whatever you choose, plan device and OS-version test coverage early, because a till queue is a bad place to discover that the code does not render on an older Android phone.

Off-the-shelf loyalty platform or your own rules engine?

This decision matters more than the choice of app technology, because the rules engine is what marketing will change every week for years. The market offers SaaS loyalty platforms with ready-made points, tiers, coupons, segments and a campaign console, usually priced per member, per transaction or by subscription. A custom engine is built from scratch or from components.

An off-the-shelf platform is a good choice when the programme mechanics are standard and your advantage lies elsewhere: range, price, store locations. It gives a fast start, a proven marketing console and a vendor responsible for uptime. It does not remove the need for POS and ERP integration or for Omnibus price history, because those are specific to your chain. Its cost grows with members and transactions, so for a large chain model fees over five years, not just year one.

A custom engine makes sense in three cases. When programme rules are part of your pricing strategy and must be tied to the pricing system and margin, not only to points. When transaction volumes make platform fees exceed the cost of running your own team. And when the programme has to handle unusual cases, such as several brands with points exchange between them, or external partners with their own settlement.

The deciding criterion is who changes the rules and how often. If marketing must launch campaigns on its own, without a software release, a custom engine needs a good configuration console, and that costs more than the points calculation itself. If rules change quarterly and campaigns are simple, a custom engine can be simpler than platform vendors suggest. We apply the same buy-or-build logic in our article on e-commerce development: SaaS, headless or custom; a loyalty programme inherits many constraints from that choice, for example whether the store account and the programme account are the same entity.

How much does it cost to build a retail loyalty app?

The table below is an estimate, not a price list or market data. It shows three scenarios for a chain with a few hundred stores and an online store, priced parametrically: team composition times duration times rate. Rates are midpoints of the senior daily ranges on ARDURA’s pricing page: backend EUR 335–485 (midpoint 410), mobile EUR 370–520 (445), QA EUR 335–465 (400), frontend EUR 315–445 (380), project manager EUR 355–485 (420), data engineer EUR 370–520 (445). We assume 21 working days per month. The estimate excludes platform licences, POS software changes made by the POS vendor, electronic shelf labels, marketing campaigns and the retailer’s own staff time.

ScenarioScopeTeam (FTE)DurationLabour cost (estimate)
A. Wallet pass on an off-the-shelf platformApple/Google pass, browser sign-up, platform integrated with POS after the fact, basic consent register1 backend, 0.5 frontend, 0.5 QA2–3 monthsapprox. EUR 34,000–50,000
B. Custom engine, POS integration, cross-platform appPoints and coupon engine, coupons applied at the till with offline mode, member price history, iOS and Android app, campaign console, consent register2 backend, 1 mobile, 0.5 QA, 0.5 PM5–7 monthsapprox. EUR 176,000–246,000
C. Extended programmeScenario B plus two native apps, personalisation on purchase data, click-and-collect orders, electronic shelf label integration2 backend, 2 mobile, 1 QA, 0.5 PM, 0.5 data9–12 monthsapprox. EUR 480,000–640,000

The arithmetic for scenario B: two senior backend developers come to EUR 17,220 a month, one senior mobile developer EUR 9,345, half a QA engineer EUR 4,200 and half a project manager EUR 4,410, for a total of EUR 35,175 a month. Five to seven months gives EUR 176,000–246,000. Scenario A runs at about EUR 16,800 a month, scenario C at about EUR 53,400.

Three things stand out. First, in scenario B the mobile app is less than a third of labour cost. The rest is the engine, integration and quality, exactly what the mock-ups do not show. Second, moving from B to C more than doubles cost, and the new features (native apps, personalisation, shelf labels) should be justified with data from scenario B rather than forecasts. Third, the most likely cause of a budget overrun is none of these lines; it is the wait for a POS software change and the number of POS types in the estate.

For comparison with proposals: ARDURA’s software development services page states that a typical three-person team (two developers and a QA engineer) costs about EUR 14,500–28,700 a month depending on seniority, and that simple MVPs take two to three months and mid-size applications four to eight. For a broader view of cost and engagement models in custom software, see our custom software development guide.

What costs appear after launch?

Building is the beginning. A loyalty programme carries running costs that belong in the board pack alongside build cost, because they hit the P&L every year.

Maintenance and development. Apple and Google release new OS versions every year and app stores change their requirements. An app nobody updates starts having publishing problems within a year. Add the feature requests marketing will make. The size of this line depends on the pace of change; as a starting point, plan for a small standing team rather than “support by the hour”.

The points liability. Unredeemed points are a promise of future benefit. Under IFRS 15, and similarly under ASC 606 in US GAAP, an option that gives the customer a material right is a separate performance obligation, so part of the sales revenue is deferred until points are redeemed or expire. Agree the treatment with your finance team, but the system must supply the data: points issued, redeemed and expired, and an estimate of breakage, the share of points that will never be redeemed.

Fraud. A programme where points have value attracts abuse: account takeover, claiming points from other people’s receipts, return abuse. You need limits, monitoring of unusual patterns and a process for customer reports. Price it from the start rather than after the first incident.

Accessibility. Since 28 June 2025, the European Accessibility Act (Directive 2019/882) has applied to e-commerce services in the EU, including those delivered through mobile apps. If the app lets customers buy or reserve products, accessibility requirements apply to it directly. Build accessibility testing into every release. In the US, the same work reduces exposure to ADA claims.

Customer service and operations. Points disputes, adjustments, questions about programme terms. A customer service console with transaction history and manual adjustments that record who changed what and why is part of scenario B, not an add-on.

How do you measure whether a loyalty programme pays back?

App downloads and member counts are activity metrics, not returns. A programme can have a million members and change nothing in customer behaviour if it rewards purchases that would have happened anyway. So ask the payback question before building, because the answer determines what data the system must collect.

The most important measure is incremental lift: how much more, or more often, members buy than they would without the programme. Comparing members with non-members does not measure this, because the people who join are mostly those who already shop often. Control groups do: some members do not receive a given campaign, and the difference in purchases between groups is its effect. The loyalty engine and campaign system must support randomly holding members out of campaigns and record who was in which group.

The second measure is the share of transactions with an identified customer. It determines the value of the data: if only one transaction in five in store is linked to a member, personalisation works on a fragment of history. This share rises with easy identification at the till (wallet pass, phone number, digital receipt), not with new screens in the app.

The third is programme cost per identified transaction: the cost of customer benefits (discounts, redeemed points) plus operating and maintenance cost, divided by the number of such transactions. It lets you compare the programme with other forms of promotion, such as the general flyer.

Tie every feature on the roadmap to one of these metrics and to a test campaign that will measure it. That discipline protects the budget from wish lists and gives the board arguments for, or against, the next stage.

How should you plan the first months, and what should you require from a vendor?

Budgeting from the till outwards translates into a concrete plan. Stage one is a short discovery phase that settles the questions in this article; at ARDURA, discovery on software development projects takes one to two weeks and defines scope and architecture. It should produce a description of the till process (including offline mode and returns), an inventory of POS types and their integration capabilities, the buy-or-build decision on the engine, the price history model for Omnibus, a consent design agreed with your data protection officer, and a first-release feature list with metrics. If the scope is still uncertain after that, frame release one as an MVP with explicit learning goals rather than a reduced version of the full programme.

Stage two is a pilot in a handful of stores on one POS type, with a wallet pass and, if justified, the first version of the app. The pilot is meant to prove integration under real conditions, not to impress the board. Stage three extends to other POS types and the whole estate, and only after that come the scenario C features.

Ask vendors for a few things that will immediately show whether they priced a programme or a set of screens:

  • a quote broken down by the six layers in this article, stating explicitly what they assume the retailer provides;
  • a description of system behaviour when the store is offline and when goods are returned;
  • a test plan that includes real POS devices, not only a simulator;
  • how the 30-day lowest price is calculated for member prices and the app channel;
  • the list of third-party SDKs in the app, with an assessment of which require consent;
  • post-launch maintenance terms: who updates the app after a new OS release and within what timeframe.

The contract model should match the uncertainty. POS integration whose capabilities you do not yet know is a poor candidate for fixed price: the vendor will add a risk premium or defend itself with change requests. We look at who really pays for risk transfer in fixed-price software contracts. A sensible split: discovery as a separate engagement, the pilot on time and materials with a cap, and later stages on whatever model fits what the pilot teaches you.

What are the key takeaways before choosing a loyalty app vendor?

The loyalty app is the visible part of the programme, but not the most important one. A programme earns money when customers are recognised at the till in most transactions, benefits are applied quickly and correctly, promotional prices comply with the Omnibus rules, and consents allow you to come back to customers with offers. None of that happens on the phone screen.

Remember six things. First, start by asking what your POS can do and who changes it, because that sets the timeline and the budget. Second, consider an Apple and Google wallet pass as the first channel or as a complement to the app; build your own app for features a pass cannot deliver. Third, put member price history, the 30-day lowest price per channel and shelf-label information into scope, following the strictest guidance in your markets. Fourth, design consent under GDPR, ePrivacy and, where relevant, US rules at the start, with a consent register and account deletion. Fifth, compare proposals by the six layers of the programme, not by their mock-ups. Sixth, measure incremental lift with control groups and the share of identified transactions, not downloads.

If you are preparing an RFP or want to check whether the proposals on your desk cover the whole programme, see how we approach custom software development or book a call. Discovery takes one to two weeks and ends with a scope, a POS integration architecture and an estimate you can take to the board.