The fashion industry could do with a facelift - so how can Pomelo's machine learning and AI solution help? (IMG/Pomelo Fashion)

The fashion industry could do with a facelift – so how can Pomelo’s machine learning and AI solution help? (IMG/Pomelo Fashion)

Thailand’s Pomelo is revolutionizing SEA’s fashion industry with machine learning and AI

Machine learning and AI have a multitude of applications in various industries, but fashion might sound like a curious segment for their use.

E-Commerce is big business in Southeast Asia (SEA). With movement restrictions coming and going so often in the region, consumers have increasingly moved towards digital means of retail, from fast-moving consumer goods (FMCG) to electronics, to fashion.

Whilst segments like electronics are often perceived to be popular, statistics show that the fashion segment of e-commerce, surprisingly, gets the lion’s share of the multi-billion e-commerce pie. 

The SEA digital fashion industry reported a 22% GMV growth in 2020, with a projected market volume of US$ 18 billion for 2021. These impressive stats suggest a massive opportunity in this e-retail sector for brick and mortar SMEs to digitalize their services to tap on this growing market.

E-Commerce in Southeast Asia

SEA is a hotbed of online retail, with multiple e-commerce platforms serving different markets. Prominent players include Shopee, Lazada and Tokopedia.

All are or intending to serve the SEA market at large, with Tokopedia recently merging with delivery giant Gojek to form an e-commerce, delivery, and financial services giant.

Elsewhere, smaller players, usually brands, have taken to selling their products online due to easier access to customers, and cheaper operating costs.

Popular Thai fashion brand Pomelo entered the fashion e-Commerce scene in 2013, offering a hybrid brick and mortar and online marketplace fashion and lifestyle service to consumers in Thailand. They eventually expanded their digital offerings to the rest of Southeast Asia. 

Leveraging the power of big data

When it comes to goods that rely on unstable trends, such as fashion, demand forecasting is a vital component to drive efficiency in operations and management. 

The fashion industry is notorious for suffering from demand uncertainty, especially with a lack of historical data and ever-changing trends. These often result in overstocked apparel, which retailers may be forced to sell off at loss, in order to make way for more current trends on the shelves.

Pomelo CEO David Jou, who co-founded Lazada Thailand, realized the opportunity to leverage the massive customer, sales, and historical performance marketing data collected since 2013.

Recently, Pomelo launched PRISM, “an end to end brand solutions platform that aims to provide total solutions for brands to scale their business”. The platform offers a comprehensive suite of tools and services from merchandising, to analytics, to marketing, to even logistics to help smaller fashion retailers achieve better economies of scale.

They also employ a team of data scientists who can help brands identify the best potential customers through their various technological solutions including big data.

Machine learning and AI drives powerful insights

In an exclusive interview with CEO David Jou, Tech Wire Asia sought to better understand the technology powering a crucial part of their newest platform — its demand forecasting service.

According to Jou, their demand forecasting service is based on a proprietary machine-learning (ML) engine that took 18 months to develop. Machine learning is a subset of artificial intelligence (AI) that enables machines to process information and execute automated decisions in a more accurate manner. 

The engine plays heavily into supporting both the customer and supply chain sides. On the customer front, it offers personalized services based on the user’s behavior on their website and mobile application, taking into account “hundreds of millions” of data points which are then fed back into the engine. 

On the supply chain side, the ML engine uses predictive analytics to analyze data points from products and matches them to the movements of similar products elsewhere in the world in order to accurately understand and predict their popularity. 

Prism feeds its ML engine with data from both their online e-commerce platform Pomelo, as well as from their brick and mortar stores around Thailand. This information will then provide demand forecasting, which will inform inventory and merchandize planning, as well as product development.

“We have done a variety of tests, with very accurate results north of 85%. We’re all about getting the best quality data and figuring out which data points are the biggest driver of sales.”, shared Jou.

Most of the platform uses cloud computing from Amazon Web Services (AWS). However, they have yet to introduce edge computing into their digital infrastructure.

On data privacy and cybersecurity

When queried about their cybersecurity and data privacy practices, Jou shared that they have a separate team that works on data, with a focus on ensuring the right permissions are granted. 

He asserts that customer data is strictly obtained from their mobile apps and online store, and does not track user activity elsewhere, such as on social media or external websites.

Additionally, they anonymize customer data before feeding them into their ML systems. Since Pomelo’s inception in 2013, they have yet to face any data leaks or issues. When we asked who their existing clients are, Pomelo declined to answer. However, Jou did state that, down the road, some authorized case studies of clients will be released.

Machine learning and AI also powers the supply chain

On the supply chain end, Pomelo doesn’t own nor operate factories, and instead, purchases from third-party manufacturers.

Currently, Pomelo does not plan to incorporate any internet-of-things (IoT) solutions in their supply chain nor brick and mortar stores.

“We utilize a proprietary stack called Henry to automate processes. It manages everything from forecasting, prototyping, material sourcing, and even automated bidding with factory partners. We try to digitize our operations as much as we can, with our approach based on software”, he added.

At time of writing, Jou shared that the company intends to shortly launch across all markets in SEA, with a focus on onboarding fashion and lifestyle brands, with a long-term focus on selling sustainably produced fashion.

He also indicated that there will be opportunities to bring on board makeup, cosmetics, and skincare brands to the platform as well.