How to use the image similarity module

In this article, we explain how you can use the image similarity module.

Are you still manually checking if your retailers uploaded product images with the new packaging? Our image similarity module shows you instantly where images match and where they don't. 

The image similarity module consists out of 2 components: 

- Reference image editor

- Image similarity in action 
- The image similarity dashboard 

- Using the DCC to automate 

The reference image editor

Uploading reference image can be done in the Reference Image editor, or via our API (link with article). 

  • Go to settings
  • Click on Image similarity module

On top, you can filter your product selection to upload reference images per tag, set, or retailer: 
  • Tag(s), specify a tag corresponding with 1 or more products, for instance “tag: shavers”
  • Set, choose an image set you create(d) for specific retailers. On default there is only one “Default reference images” set.
  • Retailer: specify the products you want to see for a retailer. In the example above I choose Bol, which leads to 5 products.

Use the Search bar to quickly navigate to subsets of products using specific keywords, or even directly move to a specific product using a product code, such like a GTIN or MPN:

  1. Pressing the icon seen on the right will make the first 5 columns (not)sticky. This can become handy in case you want to scroll through an image set, horizontally:
  2. Click on the plus (+) icon to add an image. In case there are already images present, you can click on the image to see and update.

Image Similarity Module in action

For the Philips Shaver S9000 Prestige SP9860/13 on Bol, we want to upload a range of reference images.

As manufacturer of Philips shavers, we have access to a set of reference images. We have several (variant) images that our channels can choose from, although we want to know which ones the retailer uses, and if they at least have a set of required images live.

To do this, we will upload a set of reference images. Click on the plus (+) icon to start. Next pop-up will look like this:


Let’s start off by clicking on the highlighted block seen underneath ”1”. This indicates room to upload a reference image as equivalent of the image position 1 (on Bol).

Clicking on this button will lead you to another pop-up where you can add images. You can also add variant images, if needed. After adding images for a position, you can hit the CONFIRM button.

Potentially, you receive feedback in case the image is not accepted. Common errors are that the image is having a smaller dimension than accepted (less than 150px) or is bigger than 5 mb. 

After uploading a first reference image, you can add more. Just click on the plus (+) icon you see on the right:
We will add now an image + variant. You see here an image which is very similar to the variant uploaded:

We find it OK if the retailer is using one of these images. After uploading a few more images, we want to create an additional image set. This use-case indicates a scenario when a retailer uses a preferred image-set, and another retailer requests a different kind of image-set. To serve this scenario, we give the option to add a new set by clicking on the plus (+) icon;

In this scenario, we want to create an Amazon specific image set, for the retailer Amazon DE: When creating this set, the images uploaded in the default reference images set, will be the default (seen greyed out). You can override the images for this set though, by clicking on the position.

Adding images will lead to the scenario below:
Greyed out images are inherited from the default image set. For the Amazon set we rearranged the last 4 image positions – since Amazon is in this (fictional) example having a different agreement.

Now, after uploading an image set for 1 product, we will see the first checks within 24 hours after upload. The checks – and the similarity results will be present in the ISM dashboard, connected in the account.

The ISM Dashboard

The ISM dashboard gives insights in the performance of your reference images vs. The actual images on the retailer. You can filter similarity results based on EQUAL and/or DIFFERENT and can even see the reference images you just added or updated (including the ones done via API). By clicking on the filters seen below, you can easily get to your preferred view:

Click on the label to show / hide results. Either choose to see

  • Products with Equal results (same product image(s) vs uploaded reference image(s))
  • Products with Different results (different product image(s) vs uploaded reference image(s))
  • Products with Check in progress (just added or edited reference image(s))

The products with label: Check in progress will get an updated status withing 24 hours within upload.

You can check the details for each product by using the “expand” button

If you click on an image, you see more details:

Here you can browse through each product image (in the example above there are 9 images available) on the retailer, and even switch between retailers (in case this product is listed and available in your account.

You can always add or edit reference images (if you’re account admin) by clicking on the button you see on the top right.

Automate scoring via DCC

The DCC (link) gives an automated option to track if a reference image is Equal to the equivalent of the product image listed on the retailer. After choosing the Image Similarity field, there are 2 conditions you can choose of:

Please note that to get a score, the product should be in scope for ISM and should have at least 1 or more reference images. Only then will we rate it in DCC as Y/N. If the product is not marked for in scope, or reference image is missing, the score will always be Y. If the product contains a reference image that is not equal to the product image on the retailer (hence labeled as DIFFERENT), the DCC score will be rated as N. Performance can be tracked in a content dashboard.