ImageKit: How Noon manages the images in its e-commerce product catalog

About Noon

Noon.com is middle east’s home-grown e-commerce platform that was launched in mid-2017. With its operations spanning across the gulf countries, Noon is currently the leader in the middle east’s e-commerce market and is all set to rival Amazon in this region. From fashion items to kitchen and baby food products, Noon offers a full-fledged online market-place for local consumers, bringing them the best of varieties and collections. It has partnered with various local retailers and store chains to widen its offerings in local goods, and build a more viable supply chain. 

Client Requirements

Noon is on a steady rise exploring new e-commerce verticals and expanding its product catalogues. This requires Noon to constantly work on its product image library by optimizing new images, identifying duplicates and labelling images, and transforming images to the correct aspect ratio. Along with that, it needs to maintain optimum quality throughout the process and deliver images via its CDN and storage. 

With the rapid expansion plans lying ahead, Noon needed to upgrade its image optimization and delivery workflows in order to keep up with future demands. Along with that, they needed the final transformed and optimized images to be stored in their existing storage and delivered using their existing CDN. They needed to bring the following changes;

1. The entire image resizing and optimization process should not compromise the output quality of the image. 

2. The new image optimization system/service should be able to scale well with their ever-evolving needs, as there would be large volumes of images coming in every day in different batches.

3. Eliminate duplication of images within the catalogue and avoid processing of the same images, by detecting similar images in advance.

Noon was already getting this done by another third-party provider. But the costs ran high and the results were unsatisfactory, as the response time was too long and the final renders had high error rates. Also, Noon first needed to upload the images to the provider’s storage and then download the final renders after the transformations were done. This added to their already extraneously high costs.

How did ImageKit help?

Overview:

Using a one-time mode, ImageKit facilitated an integrated image optimization ecosystem, where Noon did not need to upload images for transformations like they did with their previous provider. This was not a usual offering of ours but it worked perfectly for Noon as it alleviated one of the most pressing issues in their image workflow.  

Additionally, ImageKit provided a metadata API that procured data about the images in the storage, including the perceptual hash of the image. This hash can be used to detect visual similarity between any two images and prevent duplication.

Details about the project

1. Image transformations and optimization

ImageKit made it very simple for Noon to get started, as all they needed to do was integrate their existing storage with ImageKit’s solution. Once it was integrated, they could easily process all images and download the named transformations, without any re-uploading. This expedited the entire image delivery workflow and helped create an effective sync between teams. 

With the new process in place, Noon is now able to build new image catalogues, create new variants of the same and publish them online in a quick and hassle-free manner. Apart from being faster, ImageKit’s bandwidth-only pricing structure helped drive down the costs even further. This was a drastic improvement from their engagement with the previous provider as they had more control and fewer resources invested in the process.

2. Duplicate Image Detection

ImageKit provides a feature in its metadata API, which returns the perceptual hash value of any image. As the API can be used on all types of images, the perceptual hash value can be used to detect the similarity between the subjects of any 2 images. This entire detection process works independent of the image dimensions. This way, any 2 images having different dimensions can still be classified as similar under this feature. 

This feature enabled Noon to conduct a preliminary ‘duplicates detection’ process for its entire image library, by fetching the unique hash values and comparing them. It allowed them to easily eliminate any duplicate images from the library before the image optimization and transformation stage. Along with identifying duplicates in the library, this feature also allowed Noon to identify duplicates within catalogues and the repeated use of ImageKit’s transformation APIs. [used on more than one image]

3. Response Rate and uptime

With real-time image optimization and transformations, Noon was able to process large volumes of new images and image updates across the website. Most of the requests that come in are for fresh processing due to which, there are always sudden spikes in the demand. ImageKit renders all outputs in real-time and also helps in driving down costs when the demand is low. This is because ImageKit bills customers only based on the final output size of the images optimized. It does not charge for the storage or the number of requests. 

With this, ImageKit has been able to maintain a close 100% SLA with Noon. It is effectively driving down the number of retries and failures that Noon faced in its image processing requests with the previous provider. 

4. Named transforms

ImageKit provides a variety of resizing and cropping image transformations using URL parameters. This way, Noon could use ImageKit’s width and height resizing on the master image and get the desired image size. It then creates a unique name to associate with a particular transformation, making it easy to remember and refer to a particular transformation. 

This immensely helped Noon, as the places where catalogue images are published/shown are fixed, and therefore, the transformations used for every image are also fixed. Using named transforms, they can specify an alias for every transformation, making it easy to identify each image.

“Using ImageKit has helped increase the number of catalogue images that we can process in a day while reducing the retries that were being done earlier for the same. It has also helped cut down the cost significantly by eliminating unnecessary uploads to their storage.”

-Ankush Thapa (Principal Engineer, Catalog Team at Noon)

In Conclusion

ImageKit’s customized image processing services helped Noon to uplift their entire image delivery workflow and expedite their entire media production process for the website. This dramatically reduced the time taken for adding new product listings, updating existing ones, and replacing images for product variants. Along with that, they eliminated a major hurdle of image duplicates within their catalogues, allowing them to scale smoothly with new products. Being an e-commerce site, Image library management is one of their biggest investments. With ImageKit, Noon manages to churn out large volumes of high-quality renders and delivers them across devices via their existing CDN, without facing any cost issues. 

If your company is facing a similar challenge in image processing that is holding you back in optimizing your web performance, reach out to us at 

support@imagekit.io for a quick consultation session. You can also try out the product by creating a free account here