Challenges
One of the biggest problems with the website was that the images were uploaded but, were not updated live on the webpages. Due to this very reason, the user experience was unsatisfactory, and the business was losing potential customers. Another problem that arose was in determining the problematic images with image verification itself becoming a challenge.
Solution
We adopted a 3-step approach with Content Filtering, Content Upload, and Content Validation. We used Automation to direct manual efforts in the right direction and with the right balance of manpower and technology, we were able to solve the issues at hand.
- Content Filtering: Products with the incorrect images were marked through the image similarity algorithm. Random sampling was done to ensure the correctness of the algorithm.
- Content Upload: Correct image was manually restored or wherever possible through the Selenium automation suite.
- Content Validation: A crawler was set to signal if the image is live. Random sampling was done to ensure the correctness of the algorithm.
Outcome
2000 images were uploaded with an average of 7 images per product. The use of automation reduced the cost of maintenance and improved accuracy.
We successfully helped our client maintain product images on Amazon and Walmart. The problem could not be solved through automation alone. Our team used a hybrid approach – manual, automation, manual-in sequence. Crawlers provide a real-time feed, and the automation suite detects any image anomaly.
Related Projects
Reconciliation
