AI in fashion loyalty programmes: segmentation, rewards and incentives

KEY TAKEAWAYS Summary by the editors

1. In April 2026 M&S relaunched its Sparks scheme with cash-value rewards in a digital wallet and said it was powered by a step-up in AI and data use, with machine learning in use and generative AI models due later.
2. Olymp launched its Insider programme in eight European markets on 28 September 2026, with bonus and status points, three tiers, and a planned later phase in which members voluntarily share preferences in exchange for tailored content and benefits.
3. McKinsey and Business of Fashion's State of Fashion 2026 report said more than half of executives cite retention strategies as a key 2026 theme and that fostering customer loyalty is emerging as an important front line.
4. Loyalty membership can be used as an incentive linked to returns: when H&M introduced a £1.99 returns fee in the UK in 2023, returns stayed free for members, and a GlobalData analyst saw the fee as a prompt to join.
5. AI in loyalty is mainly segmentation, offer selection and prediction of lapse or return behaviour; it depends on clean identity resolution across stores and online and on customer consent for the data used.

AI in fashion loyalty programmes is used mainly to segment members, choose which offers and rewards each person sees, and predict behaviour such as lapse, so that incentives go where they change outcomes. Published examples from 2026 show retailers rebuilding schemes around digital wallets, tiers and consented preference data. The limits are data quality, customer consent and the risk of spending on rewards for purchases that would have happened anyway.

Why are fashion retailers rebuilding loyalty schemes now?

Loyalty has moved up the agenda. The McKinsey and Business of Fashion State of Fashion report published on 17 November 2025 states that fostering customer loyalty is emerging as an important front line, with more than half of executives citing retention strategies as a key 2026 theme. The same report says more than 35% of executives already use generative AI in areas such as online customer service, image creation, copywriting, consumer search and product discovery, and that customers increasingly use large language models to search and compare products.

Olymp's managing director for sales, Peter Hoever, framed the company's new scheme in this context: the brand wants to keep direct customer relationships as AI agents automate more of shopping, saying that personal connection, direct engagement and trust are becoming more important. Owning the relationship and the consented first party data is the strategic reason for the investment.

What do current examples look like?

These are company announcements, so they describe intent rather than measured results. None of the sources reports uplift figures for AI features, and you should be cautious about vendor claims that do.

Read also

How adidas uses AI: from design archives to generative content

What does AI actually do inside a loyalty scheme?

• Segmentation: clustering members by behaviour (categories, price sensitivity, channel, return habits) rather than only by spend tier.
• Offer and reward selection: ranking which offer, category prompt or bundle to show each member. M&S describes bundle rewards and rewards for trying categories customers do not usually shop.
• Prediction: estimating the likelihood that a member lapses, responds to an offer or returns an item.
• Content and communication: generating or tailoring messages, with human approval for tone and claims.
• Service: answering member questions about points and tiers, subject to the safety principles that apply to any customer-facing assistant.

How do rewards and incentives need to be designed?

The central economic question is incrementality: does the reward change behaviour, or does it subsidise purchases that would have happened anyway? A model that targets heavy buyers with discounts may raise measured redemption while lowering margin. The remedy is a holdout group, a share of eligible members who do not receive the AI-selected incentive, against which you compare spend, margin and retention over a sensible period.

Reward design also needs to account for returns. A scheme that pays on gross spend can reward high return behaviour. Linking rewards to net spend after returns, or using membership as the condition for free returns as H&M did in 2023, aligns loyalty with margin. Appriss Retail's survey of North American consumers reported that 55% avoided retailers with restrictive returns policies and 70% spent more after a positive return experience, so returns terms for members are part of the loyalty proposition, not a separate topic.

Measurement should extend beyond redemption rates. Useful indicators are incremental net margin per member, repeat purchase rate, share of members active in more than one channel, return rate by tier and the cost of rewards as a percentage of net sales. Report them by segment so that gains among new or lapsed members are not masked by already loyal customers who would have bought anyway.

What data and consent does it need?

1. Identity resolution across stores, online, app and customer service, so a member is recognised wherever they buy.
2. Clean transaction and returns data at item level, with net sales after returns.
3. Consent and preference records that govern which data can be used for personalisation and for which channels.
4. A clear value exchange: Olymp plans members to share interests and preferences voluntarily in exchange for tailored content, services and benefits.
5. Governance for automated decisions: explainable segment logic, human approval for sensitive actions and an appeal route for members.

Read also

Is fashion AI biased? Sizes, skin tones and fairness explained

What are the risks and limits?

Over-personalisation can feel intrusive, and thin data produces poor recommendations: an occasional shopper provides little signal. Models can also learn to favour those who already spend most, leaving other members under-served. Fairness checks matter where offers differ materially between customers, for example in who receives early access or higher rewards. Finally, programme complexity can confuse members; M&S noted that its changes followed feedback from thousands of customers who wanted a better scheme, a reminder that clarity is part of the design.

Brands that sell mainly through wholesale partners face an additional constraint: they often hold little first party purchase data, so loyalty ambitions depend on partnerships or on direct channels such as an own online shop. A sober approach is to start with one decision, such as which of three category incentives to show, prove its incremental value with a holdout, and only then extend to more decisions. Generative features for content can follow once the underlying data and consent foundation is reliable.

Frequently asked questions

How is AI used in fashion loyalty programmes?

Mainly to segment members, select the most relevant offers or rewards, predict lapse or returns, and tailor communications. M&S said its relaunched Sparks scheme uses machine learning now, with generative AI models planned for later.

Do AI-driven loyalty programmes increase sales?

Published sources describe plans and intent rather than measured uplift. The only reliable way to know is a holdout test measuring incremental spend, margin and retention against members who do not receive the AI-selected incentive.

What data do fashion loyalty programmes need for personalisation?

Linked identity across channels, item level transaction and returns data, and consent and preference records. Olymp, for instance, plans a later phase in which members voluntarily share preferences in return for tailored benefits.

Should returns be part of a loyalty scheme?

They can be. H&M kept returns free for members when it introduced a £1.99 fee for non-members in the UK in 2023. Rewarding net spend after returns also stops a scheme from paying for high return behaviour.

GuideThe complete guide to AI in fashion merchandising and buyingRead the complete guide