How to help the augmented consumer through the purchase process (episode 3 of 3)

In this third and final episode, we look into the future of online search, where this function becomes central to consumers' daily lives. As technologies continue to advance, search will remain an essential reflex, whether it’s finding a product, gathering information, or interacting with voice and visual assistants.

This article explores how voice and visual search are already shaping the user experience and why they present major opportunities for e-commerce businesses. The effectiveness, relevance, and adaptability of search will be key in turning visitors into buyers.

And tomorrow? Everything will be search!

The evolution of technology will always bring other ways to access products but it is certain that consumers will continue to use search because it has been the most natural and effective method since the arrival of Google at the end of the 90s.

In fact, when a person wants to look for a product, one of their first reactions is to use the most well-known search engine, Google, which owns more than 93% of the market share in France.*1 As already mentioned, 54% of consumers in the USA turn to Amazon to search directly for a product. We can therefore see that this trend is gradually shifting towards e-commerce sites, which need to offer an experience as intuitive, efficient and quick as Google.

Finally, the search function is at the heart of people’s daily lives, regardless of the means used and the type of search:

• Searching for technical information about a product or its price
• Also searching for information, news, a song, a film …
• Asking a vocal assistant what the weather will be like in Paris tomorrow or searching for Marvin Gaye songs for a playlist Taking photos of a product and searching for similar products (Google Images, Instagram, and Amazon already offer this)
• “Shazam-ing” a piece of music and searching for artists or equivalent titles
• etc.

These different activities, in some cases taking place daily, are all searches. They are simply interfaces that make the life of a consumer easier, and which the consumer may decide to use in one way or another depending on where they are, the system they use, the time etc.

Among these examples, two ways of searching are undoubtedly going to mark a turning point in the exploration of product catalogues: voice search and visual search. Let’s look at these a little more closely!

Voice search

Voice search is based on automatic speech recognition (ASR).

This relies on several algorithms capable of detecting phonemes, syllables and words, in order to transcribe them into text, including key words for the actions and commands that the person wishes to carry out:

Searching for a product, planning a meeting, adding a product to a shopping list, switching on the lights in a room, lowering the blinds, ordering a product etc.

Voice recognition is therefore the foundation of every voice search. This technology is applied in several stages:

1. Recording and digitisation of the phrase.

2. Voice recognition programme: voice synthesis, acoustic processing and analysis in order to transform the voice into text. Multiple technologies exist for each system / business that has developed an ASR programme (Google, Apple, Microsoft, etc).

3. The text is then sent to another programme for text analysis in order to detect intentions or actions that are linked to other words in a sentence.

4. These elements, extracted from the text analysis, are passed to the application in which the possible actions are coded to carry out the user’s request. It is important to understand that here, each application can have its own programme. In the context of voice search, the essential part of the programme takes place following the fourth step.

By its nature, the use of voice search is limited to specific situations. We know what we are looking for when we speak to our voice assistant.

In fact, today, a voice search via an assistant can only return a single result because there is no screen associated with it (or because the screen is too small). This requirement for precision and accuracy is a factor which means that it is unrealistic at this time to think that all purchases could be made in this manner.

Only purchases of recurring products (for example, 100 coffee capsules every two weeks), or that are extremely precise, with a reference number (for example, an ink cartridge for my Canon Pixma Pro MP 620 printer) could take place using this type of search.

Barely anybody today searches for a dress or a computer with a voice assistant, whether using this method directly or via a search result which would be sent via SMS, email, etc. We would lose the interest and speed that this technology offers.

Voice commerce is still in its infancy and voice search does has a bright future. However, to say the e-commerce revolution will happen through voice is to exaggerate. 

“Understanding” spoken language offers interesting possibilities for e-commerce in general but it is highly likely that consumers will adopt a pragmatic attitude and use the simplest and most effective method: voice for avoiding writing, but clicks to avoid speaking…

EXAMPLE OF NATURAL LANGUAGE VOICE SEARCH FUNCTION

1. Initiate vocal recognition

2. Express the need and the associated application, “I’m looking for a packet of 12KG Royal Canine biscuits for my adult beagle from “The Dog Biscuit King”.

3. The voice recording is digitised then analysed to deduce and to extract the meaning and words it consists of. Here, the search action is detected and launched on the application “The Dog Biscuit King”.

4. The cleaned query is simply run on the extracted words, in other words on a simplified version of the initial result (here: biscuits + 12kg + Royal Canine + beagle + adult)

5. The engine returns the most relevant product in the form of a voice response providing the price, which here is 63 euros.

Visual search

Visual search is a new way of searching which avoids the need for the consumer to express verbally or in writing what it is they are looking for.

Indeed, the desired product can sometimes be difficult to describe or may have an unknown name. So taking a photo of it is a simple way to try and find it or, more likely, to find something similar.

This technology naturally relies on the image but not on this alone. The structured information which characterises your products is still essential.

The analysis algorithms which make up this technology are able to detect what is contained in the image and to associate this content with things they have learned previously.

Inspired by the human brain, this type of algorithm is based on a network of so-called artificial neurones, trained with thousands or millions of images whilst in the learning phase known as “Deep Learning”. The network is self-learning under supervision, which means that if you show it an image of a dog and the network says cat, it will be told that it is wrong.

In this way, the network will readjust itself by changing its parameters to avoid the same error in future. In this process, each neurone specialises in the recognition of one part of an image (eye, ears, nose, moustache, etc.).

Once image detection is performed, the information is extracted and used to conduct a search of products which have the same characteristics.

FEATURES OF VISUAL SEARCH USED ON AN E-COMMERCE SITE

A young woman spots a pair of sunglasses in a photo in a magazine.

1. With her smartphone, she takes a photo of the product in the magazine

2. The image is analysed by the algorithm described on the previous page

3. This transmits the characteristics that it has detected in this image to a programme that will transform these into text…

4. ...in order to simply carry out a search on the combination of words corresponding to the criteria,

5. The results are presented on the screen.

In the majority of cases, therefore, this means that the products in these images have to be sufficiently well described with standardised characteristics in order to be able to derive associations and propose relevant results during a visual search.

In the same way as voice search, visual search relies on a search engine. This search engine is itself based on a product catalogue described by a set of features that ensure optimal semantic relevance. You must therefore understand the importance of standardising product data for search performance.

Naturally, these new types of search are only relevant in certain contexts. We will not take a photo of every product that we wish to buy and conversely, we will not only use a voice assistant to buy products.

Setting up a visual search or a voice search simply to ride the wave of a “buzz” of technologies that everyone is talking about would be a risky gamble.

If you do not think of your client’s journey and experience, your investment in these technologies will not deliver the performance nor ROI and you even risk deceiving your clients.

Before throwing yourself into these areas, take some time to analyse and understand what your clients are searching for, and in what way. But above all, find out whether the products you are selling are really suitable to be purchased via these new search methods. The answer to these challenges will come from the creation of combinations and hybrids of the different search methods (web, mobile, voice assistant, visual search etc).

CONCLUSION

Customer behaviour has changed considerably over the last twenty years. Expectations and levels of customer requirement during the purchase process have reached a height at which traders and ecommerce companies need to master some key subjects. For example, they must provide their clients with an effective and quick way to access products: search.

In reality, the challenge is vast! This feature, whether it’s on your website or an app, must be able to generate around 40% of orders on the website and should even be the first route to finding the catalogue*2. Do not forget! A customer who does not find what they came to search for is a lost customer. Because, if necessary, they will go to another website or even Google, which will take them to your competitor.

One of the keys to the success of an e-commerce site is the performance of the search function, which in turn is based on a number of pillars: relevance, speed of response, customer experience, whatever device is used and the ability to individualise. The objective is to always return results that match the user’s search, in order to deliver immediate efficacy. If this service is executed successfully, the result will be a conversion.

Finally, regardless of the means, the device, the time and the type of search performed, effectiveness and conversion will be driven by:

1. a search engine capable of meeting the current expectations of customers,

2. a search engine that is evolving in line with tomorrow’s challenges (visual search, voice etc)

3. and by a product catalogue that has structured data and is of high enough quality to be used effectively.

*1 Source : StatsCounter’s statistics
*2 Source : Among Sensefuel’s customer base, 43% of orders are generated by our technology.