How Mytheresa uses data and personalisation to serve its top luxury clients
Mytheresa has used predictive models since 2017 to find future high-value customers, while keeping buying human-led. What is published, and what AI is not doing there.
KEY TAKEAWAYS Summary by the editors
- According to Glossy (June 2026), Mytheresa has used predictive models since 2017 to judge whether marketing spend brings in valuable customers, instead of relying only on cost per click.
- The models described weigh a first purchase, search behaviour, payment method, products and brands bought, addresses and time of day; the product bought is described as the clearest signal.
- When a first-time shopper is flagged as likely to be valuable, Mytheresa may offer preferred service, free shipping and returns, and possibly personal shopping before the customer has spent much.
- Mytheresa says new-season buying remains human-led, because data cannot show what will sell nine months ahead, and it reportedly does not use AI as a shopping adviser.
- A February 2026 summary of WWD coverage reports that Mytheresa's top 4 percent of clients generate 40 percent of its business.
Mytheresa uses predictive models to identify which new customers are likely to become top clients, then treats them accordingly early. It has done so since 2017. According to published interviews, buying and much of clienteling remain human work, and AI is being explored mainly to reduce manual preparation for personal shoppers.
Why does Mytheresa focus on top clients?
A February 2026 summary by the International Association of Department Stores, drawing on WWD, reports that Mytheresa's top 4 percent of clients generate 40 percent of its business, and that the company focuses on this group with tailored experiences, exclusive capsule collections and high-touch service. The same page says the platform curates 250 brands and that technology and AI are used to personalise service while keeping a human element, without naming methods or metrics.
That concentration explains the logic of the data work. If a small group drives much of the revenue, finding the next members of that group early is worth more than optimising average conversion.
How does Mytheresa predict which customers will become valuable?
Glossy reported in June 2026 that Mytheresa has used predictive models since 2017 to judge whether marketing spend brings in valuable customers, rather than relying only on cost per click or marketing spend as a share of revenue. The models weigh the first purchase, search behaviour, payment method, the products and brands bought, shipping and billing addresses and the time of day of a transaction.
According to Michael Kliger, CEO of LuxExperience (Mytheresa's parent), as reported, the product is the clearest signal. An expensive dress or jacket may suggest wardrobe building, while an expensive handbag may be a one-off purchase. When a first-time shopper is flagged as high future value, Mytheresa may offer preferred service, complimentary shipping and returns and possibly personal shopping before the customer has spent much, treating the person, in Kliger's words, as if they were already a good customer.
| Use | What is reported | Reported limit or caveat |
|---|---|---|
| Marketing spend decisions | Predictive models since 2017 judge customer value | Article gives no accuracy or return figures |
| Early VIP treatment | Service, shipping and returns benefits for flagged first-time shoppers | Offered selectively, 'may offer' |
| Search | Semantic search handles more natural language | Customers still mostly search by brand, category and size |
| Recommendations | Models prioritise recent activity and similar customers' behaviour | Older history and repeat-purchase logic weighted less |
| New-season buying | Human-led under the chief buying officer | Data cannot show what will sell nine months out |
| Personal shopper support | AI explored to cut manual preparation | Not used as a shopping adviser |
Where does Mytheresa use AI in search and recommendations?
The Glossy report says Mytheresa has adopted semantic search that handles more natural language, though customers still mostly search by brand, category and size. Recommendation models now prioritise recent activity, such as what a shopper viewed or bought in the past few weeks and what similar customers are engaging with, over older history. A separate Glossy report on LuxExperience says many customer-facing AI features run on a Google Vertex AI partnership that began in 2021, using product data and user events such as clicks, purchases and add-to-cart actions to return ranked results for individual users.
The same report says the group has widened its use of algorithms to on-site and newsletter recommendations, on-site search, on-site merchandising, and product copy and imagery. It does not attribute specific revenue gains to these tools.
What does Mytheresa deliberately not automate?
Two boundaries are stated. New-season buying remains human-led, driven by the chief buying officer and her team, with Kliger noting that data cannot show what will sell nine months out. And Mytheresa is reportedly not using AI as a shopping adviser in any capacity, though it is exploring AI to reduce manual personal-shopper work such as pulling customer information and preparing recommendations. At events, conversation insights feed back into the company, and personal shoppers use customer background information for seating decisions.
Do the results prove AI works for Mytheresa?
Not by themselves. In LuxExperience's fiscal third quarter of 2026, reported on 22 May 2026, Mytheresa was the group's strongest business, with net sales up 9.9 percent at constant currency and US net sales up 33.8 percent. Glossy notes that the article does not attribute the results to AI. Bernstein analyst Luca Solca was also quoted warning that search could shift to AI platforms, reducing the relevance of multi-brand retailers like Mytheresa.
- Lesson 1: define value before predicting it. Mytheresa's models target future customer value, not clicks.
- Lesson 2: act on the prediction in service terms. The reported benefit is treatment (shipping, returns, personal shopping), not only a different ad.
- Lesson 3: keep human judgement where data is weak, notably forward-looking buying.
- Lesson 4: watch for platform risk, as AI assistants may intermediate the customer relationship.
What are the risks of predicting customer value early?
Predicting future value from a first purchase involves trade-offs that the sources only touch on. A model that favours certain signals, such as payment method or address, can misclassify customers and treat some better than others for reasons they cannot see. Preferential service given before spending can be wasted on customers who never return. The reports describe offers as selective rather than automatic, which limits cost, but they publish no accuracy figures.
There are also data protection considerations. Using addresses, time of day and payment method for profiling is subject to privacy law in the EU and elsewhere, and companies need a lawful basis and transparency. The sources do not describe Mytheresa's compliance approach, so none is claimed here.
What does the Mytheresa model suggest for smaller retailers?
The principle scales down better than the tooling. A smaller retailer can define what a high-value customer looks like in its own data, look at which first purchases predict repeat buying, and decide in advance what extra service it will offer. Even simple rules, tested against actual repeat behaviour, apply the same idea. What the Mytheresa reports add is the reminder that models should serve a service decision, and that human judgement remains in buying and personal service.
Frequently asked questions
How does Mytheresa use AI?
Mytheresa uses predictive models, reportedly since 2017, to estimate which first-time customers will become valuable and to guide marketing spend. It also uses semantic search and recommendation models, and is exploring AI to reduce manual work for personal shoppers.
Does Mytheresa use AI to decide what to buy?
No. According to Glossy, new-season buying remains human-led, and Mytheresa's CEO noted that data cannot show what will sell nine months out.
Does Mytheresa have an AI shopping assistant?
Glossy reported in June 2026 that Mytheresa is not using AI as a shopping adviser in any capacity. It is exploring AI to help personal shoppers prepare, such as gathering customer information and drafting recommendations.
What share of Mytheresa's business comes from top clients?
A February 2026 International Association of Department Stores summary of WWD coverage reports that the top 4 percent of clients generate 40 percent of business. The figure is relayed by a secondary source.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
