When measuring navigation precision on Ebay its clear that categories that have structured data catalogs, consistently yield higher levels of navigation precision with an obvious positive effect on sales. However, though very large and important sections of site, cannot be structured with catalogs, the situation can be helped.
After reviewing how navigation precision levels vary between ecommerce retailers, we wanted to see how precision levels may vary between different product categories, within the same online store. For this review we choose Ebay. As one of the world’s leading online marketplaces it has both a very extensive set of attributes and values used for navigation as well as a very high amount of products that can be tagged with these attributes and values.
When sellers want to sell a product on Ebay, they have two possible options to provide the product details and make them visible on the marketplace. The first, is to upload the listing data and manually provide all the possible details - “item specifics” about the product in the listing. The seller can do it by using the different attributes Ebay offers for this category of products. Some of these item specifics are mandated and have to be filled, but most of them are optional. For example, a seller who sells t-shirts can add to each listing he uploads the size, color, material, sleeve length, neckline, etc. In this case, the seller bears all the responsibility and burden of tagging the products, in order for them to show up in relevant filtering results.
The second option to provide listing data to Ebay is to associate it with an existing product in Ebay product catalog. This option is more likely to be used in the categories where products are branded or easier to define. All the seller has to do is choose or match his listing to the product from the catalogue, and Ebay will take care of the appropriate tagging, by inheriting the product data to the specific listing. For example, a seller who wants to sell a Seiko SKX009 Divers Watch, only needs to select it from the list. In this case, Ebay is the one bearing the burden of the proper data tags creation.
The difference between the products eligible for each option, is the ability to catalogue them. This characteristic, of course, varies from category to category. Some categories are quite difficult to catalogue, because they are mostly brandless, or lack specific structure, such as t-shirts. Other categories, such as digital cameras, are highly branded and structured, which makes them easy to catalog.
To further complicate matters, the listings can be submitted to Ebay either one by one through the “Sell your item” flow or in bulk through the API or feed. In the first case the selection of the matching Ebay product is manual, thus visual and relatively simple. In the second case, in order to match his listing to catalog the seller needs to either directly submit a matching Ebay product id (ePID) or provide product identifiers like UPC or ISBN to allow for matching to be completed offline. Clearly, for the sellers with considerable feed sizes, providing the matching Ebay product ids can be a very complicated process.
Now when the process of getting the navigation data to Ebay is clear, we set out to examine the differences in the data precision levels between listings in categories that are structured / catalogued and unstructured. For this purpose, we visited Ebay.com, Ebay.co.uk and Ebay.de, and reviewed the following categories: Men’s T-Shirts, Men’s Sneakers, Men’s Watches, Digital Cameras and Fragrance. In each category we selected 1-2 attributes to refine the results, and checked how many of the first 100 results actually match the values we selected.
As discussed, the Men’s T-Shirts category is a fine example of an unstructured category. When surveying navigation precision on Ebay sites, we selected the ‘Long Sleeve’ and ‘V-Neck’ values from the Sleeve Length and Neckline attributes, respectively. ‘Long Sleeve’ alone measured at about 74% for Ebay.com and Ebay.co.uk, while measuring 85% in Ebay.de. ‘V-Neck’ alone measured around 75% for all three. When selecting both values together, out of the first 100 results, all three Ebay’s had a precision level of about 65% on average.
The Men’s Sneakers category is structured only a little better, but the strong presence of brands in this category allows for at least partial cataloging. Here, we selected the ‘Mid-Top’ from the Shaft Height attribute, and the ‘Red’ value from the Color attribute. All three Ebay’s improved on their precision, ranging from 82% to 92% precision for the first value, and more than 85% on average for the second. With both values combined, while Ebay.co.uk and Ebay.de hovered around 73% precision level, Ebay.com maintained at 86% precision. We can see the relative improvement in precision in this category compared to the previous one.
The Men’s Watches category is much more reliant on brands, which makes it more structured. However, it does have a very long tail of unbranded products, which still makes cataloging complex. Narrowing down the results to watches with leather armbands and blue dials, the results were mixed. For leather armbands, Ebay.com achieved 71% precision, which is significantly lower than the 90% and 93% achieved by Ebay.de and Ebay.co.uk, respectively. However, when selecting both attributes together, Ebay.com achieved only 59% precision, Ebay.co.uk did better at 76% precision, and Ebay.de topped them with a solid 85% precision. The results demonstrate the fluctuations possible in a category with such a long tail.
The Digital Camera category is the opposite of Men’s T-Shirts, as it is one of the prime examples of a well structured category, which is entirely based on brands. This time we selected the ‘Compact’ value from the Type category, and ‘Bluetooth’ from the Features attribute. All three Ebay’s returned impressive results which measured at 96-98% precision, both when we surveyed each value by itself, and when we combined them.
The final category, Fragrances, is unique in that while it is not solely based on brands, the products in this category are quite easily tagged by the seller (only requires tagging name, scent, formulation, and volume). We selected the ‘Spray’ formulation, and the ‘Jasmine’ scent, and surprisingly, all three Ebay’s returned results with 100% precision for each of the values separately. When narrowed down the results using both values, the precision remained at 100%!
As we can see, there are significant differences in the precision levels of different categories. The more structured the category is, the higher the precision levels (with the exception of the unique Fragrance category). When a category is well structured, it allows ebay to catalogue it, and then take over the tagging process, which improves precision level.
Join us in the next post, when when we follow up on the navigation precision on Ebay, look into the same categories and review how the listing data coverage level.