Visual search – Finding products with images

Not only the result, but also the user experience is crucial

In online retail, a well-developed and easy-to-use product search is essential for customers. For a high level of convenience and optimal results, this now includes visual search. This means that users can search for similar items using images of clothing, for example.

I tested four apps from the following fashion providers for this function:

• European clothing company C&A
• German fashion mail order company Bonprix
• Swedish clothing retailer H&M
• British fashion mail order company ASOS

In all of these apps, users can upload pictures from the device's database to the search function or take pictures themselves with the camera to search for similar products.

How conveniently and reliably does this type of product search work with the four providers? Here are my results:

How well can the AI "see"?

Whether the pictures are taken vertically or horizontally is not important; the artificial intelligence recognizes the type of garment. With the H&M app, I only noticed that more skirts, pants, and dresses were suggested in the results when the picture was taken horizontally instead of just blouses or tops.

The apps can also handle incomplete shots quite well. The apps from H&M and Bonprix also suggested blouses when a partial shot of the blouse was taken, while the app from C&A picked up the pattern but suggested other garments besides tops (which doesn't have to be wrong).

I tested another challenge for the apps: What results do they provide when there isn't just one garment in the picture, (which can happen when shooting on a clothing rack or in a photo not taken specifically for the search).

If the two garments are hanging next to each other, the C&A app did not return any results.

The Bonprix app recognized both garments – even when they were partially hanging on top of each other – and gave me two individual result lists below each other that matched the two products well.

The ASOS app automatically focused on one of the two garments and only gave me results for the dark blouse.

Cropping option

When there are multiple garments in an image, but only one should be relevant to the search, features that allow you to set a focus on an overall image are very handy.

For this purpose, H&M and ASOS each offer the option to crop the image in their apps. This way, you can select one of several pictured garments or target only the fabric pattern or color for one item.

Filter and sort function

Finally, when I get a list of results, which at ASOS for example often included over 400 suggestions, sorting and filtering options are extremely useful for users, just like in webshops on websites. Here, too, the apps from H&M and ASOS stood out positively. Both offered features to sort or filter the results according to various criteria.

User experience

Ultimately, it is not only the variety and quality of the results that is crucial for a comfortable shopping experience, but also how pleasant and intuitive the use of the app is for users.

The start in terms of visual search is good in all four apps: the search function is recognizable at first glance and the camera and photo search functions are simply designed and intuitive to use.

The apps from H&M and C&A also provide tips on what customers should look out for when taking pictures in order to achieve the best possible results.

The operation of functions such as the cropping option is also important. Here, I had a hard time with the H&M app: First, I didn't realize that I could change the cut at the corners of the grid, then I had a much harder time controlling the exact cutout than with the ASOS app.

For consumers, visual search functions offer a convenient search option in everyday life. Retailers should keep these options in mind for their own offerings.