Short answer. A loyalty programme in Algeria in 2027 rests on two things: identification at the till (phone number or QR, five seconds, without slowing the queue, points computed on the real receipt) and the win-back (the list of those who have not come back for twenty-one days, a dated message, a measured return). The first produces the data; the second produces the revenue. Between the two, rules that prevent cheating, a card in the phone and a dashboard per outlet and per cashier. The stamp card does half the job and learns nothing; the tool does the other half.
This guide is for the chain of cafés or tea rooms, the fast-food, pizzeria or chicken brand run directly or as a franchise, the bakery-pastry with several outlets, the minimarket or chain of shops, the perfumery or clothing store, and the diaspora founder who used the McDonald’s or Burger King app in France and wants to launch a brand in Algeria with a programme from day one. It does not sell points: it describes what a brand builds to know who comes back, and to act when he stops.
What the figures say
Loyalty has become the first channel of the big chains. In its results of 11 February 2026, McDonald’s reports that systemwide sales to loyalty members reached nearly 37 billion dollars in 2025, up 20%, across 70 markets, with nearly 210 million 90-day active users, up 19%; the target set in December 2023 is 250 million users and 45 billion dollars of sales by the end of 2027, from 150 million and a little over 20 billion in 2023. Burger King France reports in its 2025 report 7.8 million active Kingdom accounts for 617 restaurants. Starbucks counts about 34 million 90-day active members in the United States over its fiscal 2025. On the customer side, an Ifop study for Captain Wallet of May 2024 gives 87% of French people enrolled in at least one programme, 40% holding a restaurant chain card, 65% who find it more practical to group their cards in the phone wallet, and 54% of 18-49s who say notifications influence their purchases. A summary of the Fevad 2025 figures adds, on our reading, about six programmes per person and 43% of French people who have already abandoned one: loyalty that brings nothing gets dropped.
In Algeria, the outlets are many and almost none knows its customers. The trade register counted 2,419,913 operators on 10 December 2025 according to the Minister of Domestic Trade, 42.11% of them in retail, i.e. 1,092,176, and 37.6% in services; there is, to our knowledge, no recent official count of cafés and fast-foods, the last figure quoted by the press being about 7,000 fast-foods in 2012, to be taken as an old order of magnitude. The chains are arriving: according to L’Express Franchise of 18 August 2026, KFC opened its first Algerian restaurant on 14 April 2024 in Dely Ibrahim and has four in Algiers two years later, with Oran planned; Chicken Street is in Algiers, Oran and Mostaganem; about a hundred franchise networks are present, and neither McDonald’s nor Burger King has ever opened, Quick having stayed from 2007 to 2011. Coffee itself is a mass market: the press puts national consumption at around 126,600 tonnes a year and ranks Algeria first in the Arab world in the rankings it quotes, estimates without a published method, but which tell the frequency: one coffee a day, often in the same place.
Two more figures. In December 2023, Jumia stopped meal delivery in seven countries including Algeria, a business that weighed about 11% of its volume and had never been profitable; the restaurants that depended on it lost at once the customers the platform held for them, and that is the lesson of this guide: the file must belong to the brand. And the phone is in every pocket: the regulator ARPCE counts 51.63 million 4G and 5G subscriptions in the second quarter of 2026, with 12.79 gigabytes consumed per month and per subscriber, while the GIE Monétique counts more than 126,000 payment terminals in August 2026. The till is where the data is born; we describe the 2027 till in the guide on the till and payment reconciliation.
A regular customer in six moments
A programme is not judged on its points but on six moments, from the till to the customer’s return.
Identification is the moment that decides everything. In Algeria, it goes through the phone number typed by the cashier or dictated by the customer, or through a QR shown from the app or the wallet pass; five seconds, no account to create, no password. If it slows the noon queue, cashiers stop offering it in the second week, and the programme dies quietly. Points are computed on the real receipt, by the till itself: one receipt, one credit, for the printed amount. A programme that has the cashier type the points by hand manufactures cheating and disputes. The reward is given in product, the tenth coffee, the menu, the pastry, not as a general discount: it costs the cost price, is consumed on site and brings the customer back; its rule is written, with an expiry date, and visible in the phone.
The card in the phone is a branded app for the brands that also have ordering and pick-up to put in it, or a simple pass in the phone wallet for the others, with the balance and the offers; the customer who installs nothing keeps his number at the till and receives an SMS. The win-back is the moment that pays: every morning, the platform lists those who used to come and have not been back for twenty-one days, sends them a message with a dated offer, and counts how many came back within fourteen days. The dashboard finally tells the owner what he could not see: frequency per customer, number of dormant customers, average basket of members against the others, identification rate per outlet and per cashier.
Why the win-back is the product, and the till comes first
The stamp card identifies the customer without learning anything about him. The platform learns his frequency and acts when it drops.
The till is the first layer to build, because without fast identification there is no data. The programme lives inside the till software or next to it on a tablet, with the number or QR entered before payment, points computed on the receipt, and an offline mode that keeps the credits when the network drops and sends them afterwards. The rules are the second: one receipt equals one credit, a cap per number and per day, a void that removes the points, a reward limited to off-peak hours or chosen products, expiry at six months. They are written before the first card, because they are impossible to tighten afterwards. The customer sees his balance, his history and his offers in the app or in the wallet pass, and can receive a gift from a relative.
The win-back is the layer that pays. It splits the file into segments, dormant at twenty-one days, big baskets, newcomers of the week, customers of a single outlet, and sends each a dated message, by notification when the app is installed, by SMS otherwise; then it measures the return over fourteen days before sending the next one. An AI agent can write these messages in Arabic and French, adapt them to the segment and the hour, and propose the least costly offer that brings people back; we describe where these agents are useful in the guide on AI agents in business. Management sees the identification rate, frequency before and after, dormant customers brought back, reward cost as a percentage of identified revenue, per site and per cashier. That last line says whether the programme is worth what it costs.
«The stamp card identifies the customer without learning anything about him. The platform learns his frequency and acts when it drops.»
A worked example
An example on assumptions, to redo with your figures: a chain of five cafés, two thousand receipts a day, a 350-dinar average basket.
Identification. With a number typed in five seconds and cashiers who offer it on every receipt, the assumption is 35% of receipts identified after three months, i.e. 700 visits a day and about 4,000 customers known by their number, their frequency and their usual outlet. The cost. A coffee with a 60-dinar cost price given every ten 350-dinar receipts comes to about 1.7% of identified revenue; a 10% discount on everything would cost six times more and bring back no more. The win-back. Out of 4,000 known customers, the assumption is 800 dormant customers a month, who used to come and have not been back for twenty-one days; one SMS to each, at 3 dinars per message, assumption, costs 2,400 dinars; if 10% come back and make four visits in the month, that is 80 customers, 320 visits and about 112,000 dinars, for a dated offer that costs a coffee.
Frequency. This is the figure that justifies everything else: if the 4,000 known customers make half a visit more per month at 350 dinars, that is 2,000 visits and about 700,000 dinars a month, assumption, for a chain that makes 21 million. The dashboard measures members’ frequency before and after; without it, the owner will never know whether the half-visit arrived, and the programme will be judged on an impression. The orders of magnitude, identification rate, basket, dormant customers, return, are to be replaced by yours from the first month.
Six silent leaks
A loyalty programme loses money quietly, through six leaks that the till rules close.
The first four are till leaks. The cashier who credits his own number, or a cousin’s, for every customer without a card, is seen as soon as points are tied to the receipt and to the cashier and the top numbers are shown to the manager; a cap per number and per day does the rest. The receipt scanned twice is fixed by a unique QR per receipt, valid once, twenty-four hours, for the printed amount. Points without a purchase disappear when a void removes the points and zero-amount sales do not credit. The reward claimed at noon, when the queue is longest, is fixed by a written rule: off-peak hours, chosen products, visible in the app before coming.
The last two decide whether the programme still exists in three years. The message sent to everyone, twice a week, produces opt-outs, a cost and customers who no longer read; one message per segment, with a return measured before the next, keeps attention. And the file at the vendor’s, in a tool nothing can be exported from, is the most expensive leak: three years of customers lost the day the tool closes or changes its price. Customers, points and history are in the brand’s name, exportable, on a server and accounts in its name; that is the lesson of the restaurants that depended on a delivery platform in December 2023.
Seen in France: what transfers, and what does not
The McDonald’s app in France does three things at once: it identifies, it takes the order and it takes the payment. Burger King’s converts every order into crowns exchanged for products. The founder who used them for ten years wants the same thing in Algiers or Oran, and he is right on the substance, not on the order.
What transfers: identification on every order, the reward in product rather than discount, the balance visible in the phone, dated offers, the dashboard per restaurant. What does not transfer as is: payment inside the app. In Algeria, some customers pay in cash and others by CIB or Edahabia card at the terminal or online; the programme must therefore identify the customer independently of how he pays, by the number or the QR at the till, otherwise it will only see the customers who pay online. Second gap: not everyone installs an app for a coffee; the pass in the phone wallet, depending on what the customer’s phone accepts, or a simple web link with a QR, and the number at the till for the others, give the same programme to everyone. Third gap: the notification only reaches those who have the app; the SMS, for a few dinars, reaches the others, and that is why the win-back is measured per segment and per channel rather than sent to everyone. Order-ahead and pick-up come next, for the brands where the noon queue justifies them, and that is when a branded app becomes the right answer; we compare the branded app and presence on a super-app in the guide on that choice.
«In Algeria, the programme must identify the customer independently of how he pays, otherwise it will only see those who pay online.»
From France: launching a brand, or following a chain that already exists
Two readers live in France. The first is launching a coffee, fast-food or pastry brand in Algeria; the second already owns a chain of outlets there, run by a brother or a partner, and wants to know who comes back without standing behind the till.
The founder who launches writes the programme before opening: the points, the reward, the guardrails, the off-peak hours, all on one page, and it is much harder to add once cashiers have their habits. He opens the accounts in his name, server, SMS gateway, store developer accounts, and receives the dashboard from France: identification rate per cashier, frequency, dormant customers, reward cost. The same programme, with the same rules, applies to the second outlet and the first franchisee; the central file stays with the brand. The remote owner adds the programme to the existing till and reads, every Monday, what the manager on site did not tell him: which customers come less, which outlet identifies less, which cashier credited nobody all week. We describe the till, the stock and the sales seen from France in the guide on running a business remotely.
One feature speaks to both: the gift from a relative. A member of the diaspora credits the account of a relative in Algiers, an amount or a number of coffees, and the relative consumes it at the till with his number. How that credit is collected depends on the payment methods the brand accepts and on its setup, to be checked with its bank; the mechanism on the programme side, a gifted, named balance consumable at the till, is simple and popular in July, when the diaspora is there.
Who you are, what to build
Six profiles, six different first moves on the same base.
The chain of cafés starts with the number at the till and the tenth coffee free, because the frequency is already there and the point is to see it; the list of dormant customers comes in the first month, the wallet pass and order-ahead afterwards. The fast-food, direct or franchised, starts with points on the receipt and the off-peak reward, with identical rules in every site; the branded app, with ordering and pick-up, comes when the noon queue justifies it, and the central file stays with the brand, not the franchisee. The bakery-pastry has a season: festive orders; its card lives in the wallet, its points go on Eid and wedding cakes, and its win-back reminds in April of last year’s orders, then pre-orders with a deposit.
The minimarket or chain of shops prefers a balance in dinars to points, with offers per aisle, then a dashboard per shop and an average basket per segment. The perfumery or clothing store has a bigger basket and a lower frequency: customer account with history, points on the receipt, birthday message, then remote selling by WhatsApp or app tied to the same account. The brand launched from France writes everything before opening, and keeps its accounts in its name. A subscription, for a gym or a recurring service, is another mechanism, which we describe in the guide on a gym’s app.
The 2027 calendar
A loyalty programme has two seasons in Algeria: Ramadan, which starts around 8 February 2027, and summer. Everything in place before wins its first thousands of known customers there.
January: the rules and the till. Points, reward, guardrails and off-peak hours on one page; the till of one site equipped; cashiers trained to offer the number on every receipt, in five seconds. February and March: Ramadan. Hours change, evenings are the strongest of the year for a café or a fast-food, and that is where the programme wins its first thousands of numbers: evening offers, reward after iftar, identification on every receipt. April and May: the win-back. There are now dormant customers to win back; first campaigns per segment, return measured over fourteen days, wallet pass for those who want it.
June to August: every site. The summer crowds, the diaspora, the gift from France; the same rules everywhere, and the identification rate per outlet compared every week. September and October: segments. Big baskets, newcomers, single-site customers; the dashboard per cashier; reward cost as a percentage of identified revenue. November and December: the review. Members’ frequency before and after, dormant customers brought back, the next sites, the first franchisee, and the 2028 programme written with figures instead of impressions. We detail the real timelines of a first version in the guide on development time, and what an app costs to keep alive in the guide on the first year.
What makes this kind of programme fail
- Identification too slow. An account to create, a password, a mandatory app: the noon queue grows, cashiers give up, the programme dies in two weeks.
- Points typed by hand. If the cashier types the number of points, there is cheating and disputes; the till computes on the real receipt, or there is no programme.
- A discount instead of a product. 10% on everything costs six times the free tenth coffee and creates no habit; the reward is consumed on site.
- Rules written afterwards. Caps, voids, off-peak hours, expiry: impossible to tighten once customers are used to the opposite.
- The message to everyone. Two SMS a week to the whole file, without segment or measurement: opt-outs, cost, and nobody reads after a month.
- The file at the vendor’s. Three years of customers in a closed tool; the day it changes its price or disappears, everything is redone. Customers, points, history: in the brand’s name, exportable.
- No measurement before. Without members’ frequency measured in the first month, nobody will ever know whether the programme brought anyone back.
Frequently asked questions
The questions we receive from owners of cafés, fast-foods, pastry shops and chains of shops.
- Do we need to change our till software to launch a loyalty programme?
- Not necessarily. If the till accepts a module or a connection, the programme plugs into it: number or QR entered before payment, points computed on the receipt. Otherwise, a tablet next to the till reads the printed receipt by its QR and credits the points, without touching the till. What matters is that the points come from the real receipt, not from manual entry.
- Will our customers install an app for a coffee?
- Only some of them, and that is fine. The phone number at the till gives the programme to everyone; the pass in the phone wallet or a web link with a QR serves those who want to see their balance; the branded app comes when there is also ordering and pick-up to put in it. The win-back goes by notification for some and by SMS for the others.
- How do we stop cashiers from cheating?
- By rules, not by surveillance: one receipt equals one credit, for the printed amount, tied to the cashier; a cap per number and per day; a void that removes the points; and the most credited numbers shown every week to the manager. A cashier who credits his cousin shows up on the first Monday.
- What happens when the network drops at the till?
- The programme continues offline: numbers and receipts are kept on the till or the tablet and sent when the network returns; balances update afterwards, and the morning win-back works on the previous day’s data. A programme that stops with the Wi-Fi is abandoned by cashiers within a week.
- How long for a first usable version?
- About eight to ten weeks for the till, the rules, the number at the till, SMS and the dashboard on a first site; the wallet pass and the win-back per segment follow, then the branded app with ordering if the brand justifies it. We detail real timelines in the guide on the development time of a platform.
- I am launching a brand from France, where do I start?
- With a page of rules written before opening, points, reward, guardrails, off-peak hours, and with accounts in your name, server, SMS gateway, stores. The programme is in the till from day one, the dashboard reaches you in France, and the same rules apply to the second site and the first franchisee. The gift from a relative is planned from the start, depending on the payment methods you accept.