Today, we are going to talk about a trend that is increasingly interesting digital professionals: visual research. At a time when images represent a significant part of the activity on the web, we are trying to make the most of them.
In this article, I offer an introduction to what visual search is, what it can be used for a site and how to optimize your site in this perspective.
What is visual search?
Visual search is not a “simple” search for images. It’s the idea of reproduce with a machine, a computer program, human behavior : we manage, as humans, to give meaning to shapes, spots of color. For example, if I show you this picture …
You are going to tell me “she is a witch” whereas in absolute terms, it is only a black spot made up of a multitude of small colored dots.
Visual research is therefore…
- For a computer program, the ability to recognize images as a human being would do and make sense of it, to “understand” what is depicted above.
- For the person doing this type of research, it’s a different way to get information, starting from an image rather than from a written term.
What can visual research be used for?
What is the use of having a computer program capable of understanding an image?
Examples of using visual search
The most basic and oldest example is reverse image search, which I already told you about on the blog. She allows to find the origin of a visual or look for it in other sizes.
But today, the stakes go far beyond: visual research allows for example to take a photo of a product that makes you want with your smartphone … to buy it online without having to search for its name, reference, etc.
For example, with my Amazon application, I scan a pair of sneakers… and I am directly redirected to the product page on Amazon if the shoes are sold there. I took part in the game and, if Amazon does not sell the desired model, it nevertheless redirects me to a selection of similar products.
Another example: I have a crush on a sculpture in a doctor’s waiting room. I open the Pinterest app and with Pinterest Lens, I photograph the sculpture so that the application gives me lots of suggestions of pins relevant to the photo, to feed my decorative inspirations.
Interest in brands
The first interest is the opportunity to considerably shorten the journey of a potential client. Instead of doing a conventional search on a search engine, which exposes it to a crowd of competitors … he has direct access to the product for which he had a crush or a similar product. The probability of purchase is therefore higher.
In addition, visual research is not not dependent on the person’s “expertise”, it abolishes certain inequalities that may exist on a traditional search engine. When you do a search on Google, it is sometimes difficult to find the right wording.
Imagine a Sunday handyman who wonders what the new ceiling fixtures for electric bulbs are called: he may type “bulb that clips into the ceiling”, “name bulb recessed in the ceiling” … instead of typing “box DCL ”which is the technical name. By photographing the “ceiling owl gadget”, anyone can know what it’s called.
Another advantage of visual research: easy presentation a sample of products. Showing a product “in context”, along with other products, can inspire the user to buy more than just the thing that initially interested them.
Many brands are doing it … and to give a dizzying statistic, Pinterest claimed in early 2018 to record the record number of 600 million visual searches per month.
Visual research is a still new and very complex subjecteven for web giants. Clay Bavor, vice president of virtual reality and augmented reality projects at Google, said at a conference:
“In the English language there is something like 180,000 words, and we only use 3,000 to 5,000. If you are trying to do voice recognition, there are actually quite a few things that ‘you have to be able to recognize. But now think about the number of objects in the world, separate objects.
There are billions of them, and they all exist in different shapes and sizes, so the question of visual research is considerably more complex than what we have seen with the text or even with the voice “.
How can we therefore prepare our site, on a small scale, so that images are better understood by the different visual recognition engines? Here are a few tips.
Optimizing your site for visual research
Optimize the image files themselves
Pay attention to a few important things:
- The name of the image – Use clear terms indicating what is on the image, and separate the words with dashes. Avoid accents and other special characters. For example, “cloak-wool-hanger-red-woman.jpg” is preferable to “cloak-2208393.jpg”.
- The weight of the image – A good resolution is essential but still think of not uploading too heavy images (preferably aim for less than 200 KB). It weighs down web pages, which take longer to load … at the expense of a good user experience!
- Theattribute alt – He describes the image if it cannot be displayed. It must be written in “natural language” but contain keywords that correctly describe the image. For example “Women’s woolen fitted coat, red color”.
- The legend of the image – The alt attribute is not visible on the web page, it is information that helps the search engine and visually impaired users to understand the images. However, the legend appears on the web page for all visitors, humans and robots. It can therefore help to better understand an image and gives you the opportunity to provide a little context around the image.
Use structured data
On an e-commerce site, it is essential toadd structured data to your products, in the site code. It’s a way to bring out all of their attributes for search engines.
You can read Google’s advice on this on this page. Indicate at least the name of the product, its availability, its price and the associated visuals.
Integrate your images into a sitemap
Sitemaps, sitemaps for search engines, help you discover your content. As such, it is entirely justified to also include your images in sitemaps. You can associate information such as the legend, the geolocation of the image if it is associated with a specific place, the rights associated with the photo …
Google provides you here with information on sitemaps for images.
Use Search Console data
Google Search Console provides you with image data, which let you know what terms people have typed to find your visual content and how much of your traffic it represents.
These keywords can also be used to feed the descriptions of the images (in particular the “alt” attribute) and their context on the page.
If you use Search Console, go to the menu Performance> Search results and at the top of the page, set “Search type” to “Image” instead of “web”.
You get a graph and, just below, a table with the main keywords that allowed you to access your images.
An essential contextualization of the image
It will be easier for a search engine to understand the meaning of an image if it is unique, placed in a precise and relevant context.
On No Tuxedo for example, I often use “neutral” images from image banks: they are therefore found on the web in a multitude of different contexts, to illustrate a host of different subjects. This is clearly not ideal for visual research.
Conversely, if you post your own photos, you use them in a very targeted way in content that is very relevant to what the image represents, this is an asset in helping search engines to understand its meaning.
Work on the hierarchy of your content
To strengthen the understanding of the meaning of the image, be sure to link your content to each other in a relevant way. Not all internet users will have the same degree of precision in their searches … so creating a bundle of content around each product of an e-commerce site, for example, promotes better understanding of the visuals.
For example, there are people who go for a “Rolex Lady Datejust 1970 watch”, others for a “ladies Rolex”, others for a “luxury watch”, others for a “luxury 40 years woman gift” . The more the site will offer an organization of its contents in line with research intentions of Internet users, the more likely it is that the visual research work will be facilitated.
Submit multiple images
To facilitate visual recognition, do not hesitate to offer several different views of the product on an e-commerce site.
You do not know from which angle a user could take a photo of the product they are looking for: from above, from the front, from the back … If your site has different angles, it will be easier for search algorithms to identify the image.
Not just any image!
Also bet images that show what you’re looking to be visible for. Think of the famous books “Where’s Charlie?” »: Finding something in a very crowded photo is a real challenge. It will be the same for an image search engine!
The more the product is put forward in a sober and refined way, the more we will understand what is the element to identify in the photo.
At the time of writing, the visual research is not yet a systematic reflex but we can already put in place some good practices to prepare for it. A little experimentation never hurts when you have a website!