86% of people aged 16 to 74 in Germany have bought online at least once (Federal Statistical Office), and 13.5% of retail revenue now comes from the online channel (HDE). Anyone running an online shop in this market buys reach at a high price and loses it again when nothing follows the first order. This article sets out what can be substantiated about customer retention in retail, which loyalty program mechanics follow from it, and which technical and legal requirements a shop has to meet.
How Large the Market Is in Which Retention Works
German online retail generated net revenue of 92.3 billion euros in the 2025 reporting year, an increase of 3.9% over the previous year (HDE). This is no longer a niche channel but a market with its own competitive logic: anyone growing through reach alone pays for every order again. Retention starts at the other end - with the customer already won. For implementation in the shop this is a question of data management and processes, not of campaign planning; how a shop system has to be set up for it is described in our work as a Shopware agency.
The industry association bevh puts the B2C goods volume of German online retail at 83.1 billion euros and growth at 3.2% (bevh). The Federal Statistical Office measures 2.7% more revenue in real terms across all retail for 2025 and 10.1% more in internet and mail order retail (Federal Statistical Office). The figures say the same thing from different directions: the online channel is growing faster than the sales floor, and it is growing among an audience that already knows it - 86% of people aged 16 to 74 have bought online, 70% within the past three months alone (Federal Statistical Office). Visibility remains necessary, but it only carries the first contact; the basics are covered in our SEO work.
| Metric | Value | Source |
|---|---|---|
| Online revenue Germany 2025 | 92.3 billion euros | HDE |
| B2C goods volume 2025 | 83.1 billion euros | bevh |
| Online share of retail | 13.5% | HDE |
| Growth in internet and mail order | 10.1% | Federal Statistical Office |
| Share with online purchase experience | 86% | Federal Statistical Office |
How strongly retention works economically is often claimed and rarely substantiated. One of the few quantified figures comes from a study by Bain & Company reproduced by Loopwork: 5% more customer retention raises profit by 25% to 95% (Loopwork citing Bain & Company). The range is so wide that it is useless as a planning figure - it works as an indication of direction. What counts for your own calculation is what is measurable in your own shop: repeat purchase rate, interval between two orders, share of orders from returning accounts. Anyone who does not record these values cannot substantiate the effect of a program later either. How the first order can be improved at all is described in the article on conversion optimization.
Tier Programs: Structure Instead of Promises
Before a tier is designed, it is worth looking at actual purchase frequency. In the EU, 34% of online shoppers bought once or twice and 33% three to five times (Eurostat). The typical customer in online retail is therefore not a frequent buyer but someone with a handful of orders. A program that only grants a benefit after twenty orders never reaches this core.
The interval is recorded as well: among online shoppers in the EU, the most recent purchase was within the past three months for 62%, three to twelve months ago for 11% and more than a year ago for 6% (Eurostat). A good third of the base is therefore not lost but simply has not been around for a while. This is exactly where tier logic works that does not let status expire immediately.
Entry Tier
Every account is included from the first order. The tier costs the shop nothing beyond a clean assignment of order and account and supplies the data basis for everything else.
Second Tier
Reachable within a few orders, matched to actual purchase frequency. The benefit should become visible during checkout, not only in the account area.
Third Tier
For the regular base. Here benefits are worthwhile that cost money in operations - shipping terms, returns, early access to new items.
Top Tier
Deliberately kept narrow. It is a signal inward and outward, not a volume discount, and it needs a level of service that operations can actually deliver.
Anyone planning tiers should know the opposite direction. In the Recurly network the median annual churn in e-commerce is 4.25%, of which 2.87% is voluntary and 1.38% involuntary cancellation (Recurly). The involuntary share is remarkable: almost a third of churn arises not from dissatisfaction but from expired payment methods and failed debits. That is a technical problem with a technical solution, not a case for a loyalty program.
Anyone setting the thresholds too high loses the middle of the base: it sees the tier, never reaches it and ignores the program. As an empirical value, Loopwork cites 5% to 10% churn per month for subscription brands in e-commerce (Loopwork). So set the second threshold at the purchase frequency your order data actually shows, and check after three months what share of the base has reached it.
Personalization: What Can Be Substantiated
Personalization is often equated with recommendation logic but starts earlier. 80% of people aged 16 to 74 search online for information about goods and services, 59% take part in social networks (Federal Statistical Office). Customers therefore arrive informed and keep comparing, including after the purchase. Anyone wanting to retain them has to offer something the comparison does not deliver: a fitting approach at the right moment instead of the same mailing to everyone.
What personalization delivers in the email channel is decided by the data: splitting by purchase history, status and last activity lets subject line and content refer to the recipient's situation instead of the average. That requires the data in the shop to come together and be queryable. How recommendations can be derived from it is described in the article on AI product recommendations; connecting it to the shop itself is a matter of programming.
- Purchase history per account, not per session - otherwise every visit remains the first.
- Cart and wishlist contents as a signal, not just completed orders.
- Carry returns and complaints along so the approach does not miss the actual experience.
- Preferences the customer has set themselves take precedence over inferred assumptions.
- Consent status per channel, stored separately and revocable at any time.
As a reference value for your own channel, median figures work better than best-in-class figures. In e-commerce the median open rate is 32.67% and the click rate 1.07% (MailerLite). Anyone well below that usually has a deliverability or list problem, not a content problem. Which flows are worthwhile and in which order they should be set up is covered in the article on email automation.
Non-Monetary Incentives
Not every incentive has to cost money. Early access to a new range, an extended return period, stock reserved ahead of general sale: these are benefits that barely show up in the contribution margin and are still perceived as preferential treatment. They also have the advantage of not turning into a permanent price reduction that the customer later treats as the normal price.
One reason not to tie incentives to the shop alone: 73% of shoppers use several channels on their shopping journey (Harvard Business Review). A status visible only in the web shop therefore disappears for a large share of contacts. How channels can be connected without a data break is described in the article on omnichannel strategy.
Who Buys Online and How Often
Anyone tailoring a program should know where purchase participation lies. In the EU it is highest in the age group of 25 to 34 year olds at 90%, followed by 35 to 44 year olds at 87% and 16 to 24 year olds at 84% (Eurostat). The differences are smaller than the usual narrative of the young online shopper suggests - a program aimed only at the youngest group leaves out the higher-spending part of the base. What follows from this for the direction of a shop is set out on our e-commerce page.
Among internet users in the EU, 78% bought or ordered goods or services online (Eurostat). What is notable is where retention fails: 63% of online shoppers reported no problems at all, and the most common complaint, at 20%, was slower delivery than expected (Eurostat). A loyalty program does not compensate for that shortcoming. It only works once delivery time, returns and reachability are right.
Market Size
German online retail generated net revenue of 92.3 billion euros in the 2025 reporting year, an increase of 3.9% (HDE).
Reach
86% of people aged 16 to 74 in Germany have bought online (Federal Statistical Office).
Digital Everyday Life
88% send or receive emails, 71% use online banking (Federal Statistical Office).
The channel through which a program runs day to day is usually email - and there automation separates clearly from campaigns. A triggered flow reaches the customer at a point they set themselves: after the first purchase, after an abandonment, after a longer pause. Anyone running a loyalty program without triggered flows gives away the more effective part of the channel. Which flows these are in detail is covered under email automation.
Technical Implementation in the Online Shop
Technically a loyalty program is not a module but a chain: the shop system holds account and order, a point balance or status is derived from it, and both have to be visible in inventory management, shipping and customer service. The connection runs through integrations, for inventory management and financial accounting often through SAP Business One. Triggered processes in the shop itself can be modelled with the Flow Builder. It is important not to add data protection afterwards: data protection by design and the security of processing belong in the design, not in the acceptance test.
- Clarify the account model: guest orders without an account cannot be assigned any status.
- Bring order and returns data together per account before a tier is calculated.
- Keep product data clean so recommendations and benefits hit the right items - the basis is a maintained PIM.
- Calculate status logic server side and version it, not in the front end.
- Show benefits during checkout, not only in the customer account.
- Plan the evaluation from the start, otherwise the effect cannot be substantiated later - how we set this up in e-commerce.
The effort of integration pays off where several channels come together: omnichannel customers spend 10% more online than single-channel customers and 4% more in store (Harvard Business Review). For a status to apply across channels, there has to be a place where the data comes together; the role a middleware plays in this is described there.
Data Integration as the Foundation
The data basis of a program is at the same time its greatest risk. Anyone bringing together purchase history, channel behaviour and status per person processes personal data at a depth that calls for purpose limitation, data minimisation and storage limitation. That includes a record of processing activities, a way to hand over stored data on request in a transferable format, and a level of protection appropriate to the risk. What a central customer data store looks like for this is described in the article on the customer data platform.
In practice, integration rarely fails at the shop and often at the existing system: customer numbers do not match, returns arrive late, address changes run in two directions. Anyone clarifying this in advance saves later correction runs across the whole base; the article on ERP integration describes the typical breaking points. Recurring reconciliations can then be modelled through automation.
A loyalty program needs a legal basis under Art. 6 GDPR for every processing operation and has to comply with the principles in Art. 5 GDPR. Data subjects can request access (Art. 15 GDPR), erasure (Art. 17 GDPR) and data portability (Art. 20 GDPR); they can object to processing for direct marketing at any time (Art. 21 GDPR). Anyone deriving a status from behaviour is carrying out profiling within the meaning of Art. 4 GDPR and has to observe the limits on automated decisions under Art. 22 GDPR. For email advertising Section 7 (2) UWG applies, and for approaching your own existing customers the exception in Section 7 (3) UWG; storing and reading information on terminal equipment is governed by Section 25 TDDDG. Added to this are the record under Art. 30 GDPR, the security of processing under Art. 32 GDPR and data protection by design under Art. 25 GDPR. The framework for an administrative fine extends to 20 million euros or 4% of global annual revenue. What we implement here is set out under privacy.
Metrics That Hold Up
For valuing customer relationships without a contract there is an established model: Fader, Hardie and Lee presented the BG/NBD model in Marketing Science. It describes purchases in a non-contractual setting, where churn is not announced but only becomes apparent because further purchases fail to appear. That is exactly the situation in online retail: nobody cancels a shop. A program that keeps status and points makes this silent churn visible for the first time - provided the model assumptions are not confused with contractual logic.
| Metric | Reference value | Source |
|---|---|---|
| Purchase frequency | 34% once or twice, 33% three to five times | Eurostat |
| Most recent purchase | 62% within the past three months | Eurostat |
| Median annual churn in e-commerce | 4.25% | Recurly |
| Email open rate (median) | 32.67% | MailerLite |
| Multichannel use per shopping journey | 73% | Harvard Business Review |
In the email channel the occasion decides the effect. Abandoned cart, welcome flow and browse abandonment act where purchase intent has already become visible, and the reminder about the abandoned cart picks up directly at the point of abandonment. For a loyalty program that means: status belongs inside exactly these flows, not in a separate mailing alongside them. Where the drop-off arises is described in the article on checkout optimization.
From Transaction to Relationship
A point balance alone does not create a relationship. It is an account the customer keeps, and it works exactly as long as the benefit is greater than the effort of redeeming it. Anyone wanting more from it has to start elsewhere: with the reliability of delivery, with the quality of advice, with the question of whether a complaint is handled without friction. These things are harder to build than a points table and last longer.
How widely the return of a channel varies is shown by a survey from Litmus: 35% of marketing leaders receive 10 to 36 US dollars for every dollar invested (Litmus). The range is an indication that it is not the channel that decides but its setup - list hygiene, segmentation and the question of whether the mailing is tied to a recognisable occasion.
- Occasion instead of schedule: a message about an order works differently than a message at the start of the month.
- Show the status where it takes effect - in the cart and during checkout.
- Describe benefits instead of claiming them as a discount amount.
- Do not penalise returns; anyone charging point deductions for returns trains customers not to order.
- Connect channels instead of multiplying them - the basics are in the article on social commerce.
Technical reliability is part of retention, even though it appears in no program description. A shop that slows down under load loses exactly the customers who order frequently, because they notice the deterioration first. What is needed for this is set out under hosting; where drop-offs arise during checkout is described in the article on checkout optimization.
A loyalty program is at the same time the most reliable source of your own data: what customers state themselves and what they actually order belongs to the shop and not to an external platform. Anyone deriving a status from it is nevertheless carrying out profiling and has to observe the limits on automated decisions - excluding someone from benefits solely on the basis of an automated evaluation is the case to be avoided. How a data strategy can be built without external cookies is described in the article on first-party data.
A loyalty program does not move any metric on its own. It makes visible what a shop already delivers - delivery, advice, returns - and rewards the customers who already use it. That is why the design belongs not in campaign planning but in operations planning: anyone unable to provide the benefits permanently should not promise them.
Outlook and Investment Decision
The online share of retail keeps growing, but slowly: it stands at 13.5% after 13.4% in the previous year (HDE). Anyone basing an investment decision on a jump in the channel is planning against the data. It is more realistic to assume that the distribution between providers shifts more than the channel as a whole - and that is exactly what retention acts on.
The everyday life of the target group is digital but unspectacular: 88% of people aged 16 to 74 send or receive emails, 79% make phone or video calls over the internet (Federal Statistical Office). Email therefore remains the channel with the broadest coverage, and it is the only one the shop owns itself. What can sensibly be automated within it is described on our page on automation.
Before deciding on a program there is a sober calculation: what does the benefit cost per tier redeemed, how many accounts reach it, and what share of revenue already comes from returning customers today? Anyone without these three figures is building a program into the unknown. We work them out in consulting using the existing order data.
First Steps
Getting started does not require a complete program. A first step is to segment the base at all: accounts with one order, accounts with several, accounts without an order in the current year. Even this three-way split shows where a benefit works and where it would be wasted - and it can be built with the data every shop already has.
The second step is data quality. Duplicate accounts, missing assignment of guest orders and outdated addresses distort every evaluation before the first tier is calculated. The fact that internet and mail order retail grew by 10.1% in real terms in 2025 while retail as a whole reached 2.7% (Federal Statistical Office) does not change that: growth covers up poor data, it does not fix it. How existing data can be enriched systematically is set out under data enrichment.
The figures in this article come from official statistics (Federal Statistical Office, Eurostat), from published market reports (HDE, bevh) and from provider analyses in the email and subscription business (MailerLite, Litmus, Recurly, Loopwork citing Bain & Company). The figure on multichannel use comes from the Harvard Business Review, and the model for valuing non-contractual customer relationships from Fader, Hardie and Lee in Marketing Science. The legal framework follows from the GDPR, the TDDDG and the UWG. Every figure in the article is backed by a source, a reference and a verbatim quotation.
Frequently Asked Questions About Loyalty Programs in E-Commerce
The size of the shop is not the yardstick, purchase frequency is. In the EU, 34% of online shoppers bought once or twice and 33% three to five times (Eurostat). Where your own base is similar, the second tier has to be reachable within a few orders, otherwise the program runs into nothing. We check this in consulting using the existing order data.
Both are mechanics, not effects. Points are easy for the customer to recalculate and easy for the shop to devalue; a status is slower moving but demands upkeep over time. What matters is less the mechanic than the question of whether the benefit becomes visible during checkout and whether operations can deliver it permanently.
Email advertising generally requires consent (Section 7 (2) UWG); for your own existing customers the exception in Section 7 (3) UWG applies under narrow conditions. Storing and reading information on terminal equipment is governed by Section 25 TDDDG. Every processing operation needs a legal basis under Art. 6 GDPR, and use for direct marketing can be objected to at any time (Art. 21 GDPR).
With the same metrics as before, only separated by participation. Reference values from the industry help with classification: the median annual churn in e-commerce is 4.25% (Recurly) and the median open rate in the email channel 32.67% (MailerLite). It is important to record the baseline before the start - it cannot be reconstructed afterwards.
An account model that reliably assigns orders to a customer, status logic calculated server side and a connection to inventory management and shipping. Triggered processes can be modelled in the shop system with the Flow Builder; the connection to the existing systems runs through integrations.
The reminder about the abandoned cart. Abandoned cart, welcome flow and browse abandonment act where purchase intent is already visible. Only after that is it worth building status and benefits into further flows.