According to Salesforce’s Connected Shopper Report “Across the board, the ‘visual search’ response was more popular than personalised recommendations and notifications, mobile wallets, chatbots or drone delivery”. That echoes 2017 research from Accenture that found 69 percent of young consumers are interested in making purchases based on visual-oriented searches alone. So, customers like visual search and, for retailers, it’s worth investing in it due to the much higher purchase rates and Average Order Values (AOV).
It seems inherently obvious that products such as fashion, furniture, furnishings, jewellery, art and, somewhat contentiously, dating are inherently visually based rather than textual. Quite aside from specific language – few know what ‘appliqué’ actually means – there’s the fundamental issue that consumers are using a different part of the brain when processing images as opposed to when we’re dealing with language. So, a technology that doesn’t require customers to consciously try and explain a feeling or an impression is a significant step forward. They like a piece of art … it’s hard to rationalise and explain why in word that make sense to someone else or to a product search engine...
For retailers the benefit is considerable. A customer finding the right product at the right price is much more likely to purchase. Average sales uplifts of 30% are regularly achieved and, in some cases, much more.
However, it’s worth pointing out that it’s not as simple as implementing basic visual search technology which only displays items which look alike.
Customers have an intent; they may be searching for a particular type of clothing, or for a particular function. If a customer is searching for formal shoes, for example, they may well not care whether they are black or brown.
And the right price aspect is important … given a car catalogue customers will end up selecting a McLaren 720S … that, unfortunately, they can’t afford to buy. So, visual search systems that don’t take into account intent, budget and need won’t deliver for either the user or the retailer.
This is obvious but not so easy to solve. The user’s intent needs to be derived from their interaction (or lack thereof) with the content shown. It also means that more sophisticated classifications of items are needed. It’s not enough to simply display items that look alike; retailers need to display items that fulfil the same function, have the same style, are intended as gifts for a given demographic. Clearly, if a customer is searching for a GPS watch then they may well not be visually similar at all.
So, to actually make visual search work retailers need to go way beyond simple image similarity. They need to know what attributes and, ideally, what purposes the item represented addresses for their customers. They need to know what products are typically purchased (“other customers bought …” messages are very effective too) by which customers and then they need the image delivery sophistication that allows them to quickly infer their customer’s intent. It requires a mix of visual processing, AI, big data, deep learning and, still, heuristics as well as an understanding of the products.
Now, that’s not to say that it has to be complicated for customers. On the contrary, the user experience must be simple and clear … a presentation of a sequence of images of likely candidate purchases.
But, to get this working right and to then display the right images takes expertise.
Visii is the leading provider of visual, semantic, data and cohort optimisation solutions that deliver real revenue uplift and our experts can work with you to create a seamless user experience. Talk to us if you need that expertise …
.png)