How Farfetch used AI for personalisation and marketplace operations
Farfetch built an in-house recommendation engine and partnered with AI vendors and a styling service. A look at what was published, from 2017 to 2023, and the limits of the evidence.
KEY TAKEAWAYS Summary by the editors
- PYMNTS reported in May 2023 that Farfetch's recommendation engine was built in-house, with a ranking algorithm that ranks products one to one for each shopper.
- In August 2017, Farfetch Black & White, its e-commerce services division, partnered with AI personalisation vendor Certona so luxury brand clients could offer real-time product suggestions.
- In February 2021, Farfetch announced a pilot with Wishi, combining stylists and technology, for Access loyalty members in the US and UK, and a partnership with Aurora Mobile for AI-powered operational analytics and push notifications.
- In May 2023 Farfetch CEO José Neves said the company was developing three proofs of concept in generative AI, which he described as early but promising.
- Published sources give no measured effect of Farfetch's personalisation on conversion or sales, so the benefit cannot be quantified from public material.
Farfetch used AI mainly for personalised recommendations and operational analytics, combining an in-house recommendation engine with vendor tools and partnerships. Public sources, drawn from 2017 to 2023, describe what was built and announced, but not how much it improved results. This case study covers those facts and flags the gaps.
What is the context for Farfetch's AI work?
Farfetch operated a marketplace connecting boutiques and brands with luxury customers, plus services for brands such as Farfetch Black & White. In 2023, PYMNTS reported that after two consecutive quarters of sales declines Farfetch reported a turnaround in its first-quarter 2023 results, with revenue up 8 percent year on year, but a loss after tax of $174 million. The same article notes that gross merchandise value rose by only 0.1 percent to $931.7 million, so the headline revenue growth should not be read as a clean recovery.
The article was titled as AI fuelling the return to growth, but the figures it contains do not isolate AI's contribution, and this article does not make that claim.
How did Farfetch build personalisation?
PYMNTS reported that Farfetch's recommendation engine was built in-house, with a ranking algorithm that orders products on a one-to-one basis, and that about three years before the May 2023 earnings call Farfetch had expanded its machine learning data science teams to build personalisation. The article says personalised recommendations helped convert shoppers into buyers, but it provides no conversion figures.
Earlier, in August 2017, Farfetch Black & White partnered with Certona, an AI personalisation software company. This let brand clients make what the announcement called sophisticated product suggestions in real time across customer touchpoints. At that time the division managed e-commerce platforms for eight luxury brands, including Manolo Blahnik, Christopher Kane and Thom Browne.
| Date | Initiative | What was reported | Results published |
|---|---|---|---|
| August 2017 | Certona partnership (Farfetch Black & White) | AI personalisation tools for eight brand clients | None |
| February 2021 | Wishi styling pilot | Digital personal styling for Access members in the US and UK | Company claims only, no data |
| February 2021 | Aurora Mobile partnership | AI tools, operational analytics, machine learning push notifications | None |
| May 2023 | In-house recommendations | One-to-one ranking algorithm | None |
| May 2023 | Generative AI | Three proofs of concept, described as early | None |
How did Farfetch use AI for styling and customer operations?
On 22 February 2021, FashionUnited reported a pilot of a digital personal styling experience built by Wishi and offered through Farfetch.com, matching shoppers with inventory using Wishi's technology and stylist team. It was for Access loyalty members in the US and UK and expanded Farfetch's Private Client personal styling programme. Farfetch said the approach was proven to help customers make purchases they are satisfied with; the article presents this as a company claim, not independent evidence.
Days later, on 26 February 2021, Farfetch announced a partnership with Aurora Mobile, a Chinese mobile developer services company, to supply AI tools, operational analytics and machine learning push notifications to help personalise shopping and deliver more effective services.
What did Farfetch say about generative AI?
On the May 2023 earnings call, CEO José Neves said Farfetch was developing three proofs of concept, three prototypes, and described the work as early but promising. He also said Farfetch avoids investing in technology for its own sake and argued that fashion's focus on individuality and creativity makes AI less ethically fraught than in broader search. Public material reviewed does not describe how those prototypes ended up.
What can fashion teams learn from the Farfetch example?
- Own the ranking logic where it matters. Farfetch built recommendations in-house while buying other tools, a mix that many marketplaces face.
- Mix people and software for high-value clients. The Wishi pilot paired stylists with technology for loyalty members.
- Demand evidence beyond announcements. Most sources here are partnership announcements, which describe intentions rather than outcomes.
- Keep financial context in view. AI may support conversion, but cannot by itself fix a cost base or market conditions.
What are the lessons for marketplaces specifically?
A marketplace has particular personalisation challenges. Inventory comes from many boutiques and brands, product data varies in quality, and availability changes quickly. A one-to-one ranking model must therefore cope with sparse data and short product lifecycles. The public sources do not say how Farfetch handled these, but they do show that the company invested in its own data science teams about three years before 2023, which suggests it saw ranking as a core capability.
The sources also show a pragmatic mix of build and buy: in-house ranking, vendor tools for brand clients in 2017 and partnerships for analytics and push notifications in 2021. For retailers weighing build against buy, the question is which capabilities differentiate the business and which are commodity services.
How reliable is the evidence on Farfetch's AI?
Treat it as moderate for facts about what was built and weak for impact. The announcements are from the companies involved, the Wishi claims are explicitly described as company statements, and the 2023 financial figures show a business under strain regardless of AI. Anyone using Farfetch as a benchmark should read it as an example of approach, not as proof of return on investment.
A practical test for any marketplace is whether personalisation also serves the brands and boutiques that supply it. The 2017 Certona arrangement was aimed at brand clients, not just consumers, and the Wishi pilot targeted loyalty members. Both illustrate that AI on a marketplace has several audiences, each with its own measures of success, which public sources do not report.
Finally, the corporate context has changed since these sources were written. Farfetch's ownership and operations may have changed since 2023, so this article should be read as a record of past practice, and readers seeking current information about the platform should consult up-to-date company disclosures.
Frequently asked questions
How did Farfetch use AI?
Farfetch used an in-house recommendation engine that ranks products one to one, vendor tools such as Certona and Aurora Mobile, and a styling pilot with Wishi. It also reported developing generative AI prototypes in 2023.
Did Farfetch build its own recommendation engine?
PYMNTS reported in May 2023 that Farfetch's recommendation engine was built in-house, with a ranking algorithm that ranks products for individual shoppers. Farfetch had also used outside AI vendors, including Certona for brand clients in 2017.
What was the Farfetch and Wishi pilot?
In February 2021 Farfetch launched a pilot of digital personal styling by Wishi for Access loyalty members in the US and UK, combining stylists with technology. Farfetch described the approach as proven, but no independent results were published.
Did AI cause Farfetch's 2023 return to growth?
The PYMNTS article links the two in its headline, but its figures show revenue up 8 percent and gross merchandise value up just 0.1 percent, and it does not isolate the effect of AI. The causal claim is unproven.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
