Slamby provides world class categorization solutions for Classified Ad Sites. We are constantly developing and improving automatic and language-independent categorization to reach 100% accuracy.
In order to be the best on the market, we are aware, that we have to understand the Classified Ad market profoundly. That is why Slamby is conducting various research projects to gain a better understanding of the Classified Ad market, and to develop the best solutions for it.
Automatic categorization usage in reality
Slamby is providing a smart solution for Classified Ad sites that is able to categorize new Ads based on their title and description at a very high accuracy level ~95%. In late May of 2015 we have started to work together with our new client, Agroinform.hu. They have implemented Slamby Classifier, and this way, their customers are provided with automatic category recommendation during the Ad Posting process. After 2 weeks of working together with them we asked a few dozen of their customers how they like this new feature on Agroinform.hu in the ad posting process.
Categorization experiences before using Slamby Classifier
Agroinform.hu has a very broad and complex category tree. They had more than 9 thousand categories in a three-level category system with most of their categories being leaf categories. Whenever a customer wanted to post an ad, he had to choose from these three levels and from an abundance of leaf categories. What he saw, was not too informative, because the category tree narrowed down step by step in the category levels. Furthermore, they had to face the problem that if someone had picked a wrong category at the beginning of the categorization process, the final category could be wrong, and it could mess up the site’s database, ad search, filtration, ad recommendation and user experience.
Automatic Categorization with Slamby Classifier
When we have started working together with Agroinform.hu, we received their database, and set up Slamby Classifier. Slamby Classifier read all their ads, and checked their categories. In this way, Slamby Classifier learned the category tree and categorization on the site. After this Agroinform.hu integrated Slamby Classifier, and when a customer started posting a new ad, and typed in its title and description, Slamby Classifier automatically recommend the appropriate category.
The result looks like this on the site:
<- The Title of the ad
<- Recommended categories
Different users with different problems
2 weeks after setting up Slamby Classifier we asked 20 users, who posted a new ad on Agroinform.hu, and received recommended categories. Based on our interviews and our buyers’ feedback we were able to distinguish 2 typical user profiles.
They are those, who have a few products, or work for a seller company and have to post products, with which they are not very well acquainted. For them selecting a category took about 1-1.5 minutes on the average.
We have found that it is a significant problem that users need to understand an unknown category tree, which leads to frustration. Based on the answers we have received it is interesting that after a certain period of time users select the first relevant “Other” category and finish the category selection process. In this case the line before selecting the “Other” category was about 1 minute.
Experienced users sell more than a dozen products monthly from the same group of categories. We have interviewed a user promoting 100 posts a month on average. The average time this user spends on category selection is about 30 sec. The user at hand has on an average 100 items per month, which means that his ad posting time adds up to 5 hours and his category selection time up to 50 minutes monthly. Since users have to select the same few categories every time, category selection becomes a time-wasting, yet tiresome process, which leads to a lower user satisfaction. Thanks to our automated category recommendation process the user could save time and was more satisfied during ad posting.
What are the most important results of automatic categorization?
- At least 50% less time spent on categorization.
- Users become more committed to the site.
- A clearer category tree.
- Less time spent on category moderation.
Even the most experienced users spent about 30 sec on categorization before using Slamby Classifier. After our automatic categorization solution was deployed, this time reduced remarkably; and it takes a maximum of 3 sec to find a category. It is more than 50% time saved, even for expert users! With automated category selection there is no more time wasted during ad posting, which means almost an hour every month for a user posting 100 ads per month.