Visual Inspiration that pushes your intent to purchase fashion is everywhere – From a beautiful summer dress spotted on your favourite blogger on Instagram to the perfect pair of leather boots on the street. The process from inspiration to sale – inspiration, search and then finally purchase – has traditionally been broken and not smooth.
The Future of E-commerce
Global smartphone users are estimated to cross 6 billion mark by 2020, says Baidu. This implies that over 6 billion people will have access to the camera of their phone on a daily basis. It also estimates that m-commerce purchases will make up 49 percent of total online retail commerce in 2020, a significant increase from 29 percent in 2015.
Given the growth in smartphone users and m-commerce purchases, e-commerce players ought to make the purchase journey of a customer easier. This has led to e-commerce players such as ASOS to build their Reverse Image Search tool. It allows customers to take a picture of something they like and use it to search for similar products in ASOS’s product catalog.
ASOS (est. 2000) is a fashion retailer, which offers fashion products for men and women. The company sells more than 85,000 branded and own-brand products for men and women through its localised mobile and web platforms in the UK, the US, Australia, France, Germany, Spain, Russia and Italy. The company had 13.4 million active customers (customers who shopped at ASOS in the last 12 months) as of December 2016.
ASOS launched the ‘Style Match’ feature in the US in March 2018 for iOS and Android App versions. Users can search for their favourite items using keywords or upload a picture of an item they really loved in the Search Bar. ASOS returns search results with the exact items visitors are searching for or akin to something they’re looking for.
How does Visual Search Work?
In Image search – a text-based query is taken and the best image/visual match is returned. This is quite different than the modern visual search method. Visual search – takes an image as its ‘query’, instead of text like in Image Search.
Visual Search employs a method called ‘Computer Vision’ – a subfield of machine learning – that helps computers to locate, identify and analyse images the same way a human being would and provide appropriate outputs.
Steps in Visual Search –
1. Query Understanding: Here the system derives information about the product from the image. Information such as colours, visual features and category.
2. Object Detection: An image may contain multiple objects. Object detection uses Machine Learning to detect the relevant objects using previous search history trends. The machine learns what objects customers have been interested in. The system uses positive feedback loops to determine if the searches are correct — if users engaged with results for a dress, then it was probably a dress. With that, the system has lots of ways to train these machine learning algorithms.
3. Object Search and Output Matching: Given an input image, the system finds the most visually similar objects from its product catalogue with similar attributes and characteristics. The system then maps those items to the original image and returns search results most relevant and similar to the query image using AI.
Why is Visual Search Important to ASOS?
Fashion is visual in nature and as a user when you’re searching for something you might use text which differs from person to person. Google Search or website may give different search results for 2 people looking for the same item. Visual Search is hassle free and requires fewer clicks to arrive at a product. Ease of product discovery and fewer number of steps will result in a better product funnel. Users don’t have to page through the retailer’s 85000+ inventory list anymore, this feature does the ‘fashion legwork’.
In summary, if customers use these tools – we will witness a huge shift in the way purchases are made. Customers when in retail stores can click pictures and search for products on various shopping apps to find the most economic option. This will have a huge impact on ‘brick and mortar’ and e-commerce retailers both, forcing them to either offer products that can’t be found online or better prices. Either way – The customer wins!
Challenges and Next Steps
- Items such as clothing and footwear can have very minute and subtle differences that can become very difficult for an algorithm to detect.
- These hurdles can become more challenging since the style matches are dependent on the image quality from the user. Poor image quality could lead to incorrect style matches and a bad user experience eventually.
- Popular item searches could also navigate customers to items which are out of stock and could eventually lead the customer to get dropped off.
For companies such as ASOS who have invested in building a Reverse Image Search tool, the questions above merit some thought as consumers continue to consume content and purchase more and more digitally.
 FRPT. 2018. “Software Industry Snapshot”. FRPT Research. P41-42 [Accessed 13 Nov. 2018]
 Samuely, Alex. 2018. “Mcommerce To Capture 49Pc Of Online Retail Sales By 2020: Report | Retail Dive”. Retaildive.Com. https://www.retaildive.com/ex/mobilecommercedaily/mcommerce-purchases-set-to-jump-160pc-by-2020-report.
MarketLine. 2018. “ASOS Plc Marketline Company Profile”. MarketLine. [Accessed 13 Nov. 2018]
 HAWKSWORTH, HARRIET. 2018. “Visual Search Set To Make World Of Imagery Instantly Shopable”. The Business Of Fashion. https://www.businessoffashion.com/articles/fashion-tech/visual-search-set-make-world-imagery-shopable [Accessed 13 Nov. 2018]
 Boyd, Clark. 2018. “The Past, Present, And Future Of Visual Search – The Startup – Medium”. Medium. https://medium.com/swlh/the-past-present-and-future-of-visual-search-9178f006a985. [Accessed 13 Nov. 2018]
 Lynley, Matthew. 2018. “How Pinterest’S Visual Search Went From A Moonlight Project To A Real-World Search Engine”. Techcrunch. https://techcrunch.com/2017/02/22/how-pinterests-visual-search-went-from-a-moonlight-project-to-a-real-world-search-engine/.[Accessed 13 Nov. 2018]
 Le, James. 2018. “Pinterest’S Visual Lens: How Computer Vision Explores Your Taste”. Medium. https://medium.com/cracking-the-data-science-interview/pinterests-visual-lens-how-computer-vision-explores-your-taste-47d591b42d7c.[Accessed 13 Nov. 2018]
 Lomas, N. (2018). Asos adds search-by-photo to its fashion ecommerce app. [online] TechCrunch. Available at: https://techcrunch.com/2017/08/10/asos-adds-search-by-photo-to-its-fashion-ecommerce-app/ [Accessed 13 Nov. 2018].