Exposure across product discovery channels is essential for fashion retailers. To state the obvious: if customers don’t see your products when they’re searching, they’re probably not going to buy them. So, retailers spend millions every year trying to make sure that they’re top of the pile, or at least present in the search results of whichever channels they think are most important. Unfortunately, many of them are leaving money on the table by failing to optimise the process by which products get digitised and fed into channels like Google Shopping.
Intelligently optimised product data makes brands’ promotional spend instantly much more efficient and improves organic ranking and conversion at the same time. That's because machine learning can be taught the millions of variables (which bits of product information are most important for which channels, and how should those be formatted?) that go into creating product listings online.
Fashion ecommerce teams spend days in spreadsheets manually editing product information, often just to get it into a usable format. Then it needs to be tailored to each individual channel, which means more rounds of editing thousands of lines. Unsurprisingly, not every team member will know off by heart the millions of potential fields/values/attributes/names on each of these channels, so mistakes get made and information is left out.
To understand the impact this has, we need to think about how search engines (whether Google or Amazon’s A9 or another) value and order results.
Visibility is earned by maximising relevancy to search terms
The thing search engines are most concerned with is relevancy. Is this what the searcher is looking for? They live and die by their ability to surface relevant results to queries. Let’s say that within the product data for a pair of blue socks, the ‘Colour’ field is left blank. Were that information present, searches for “blue socks” would be more likely to return the product at a higher ranking. The more information the search engine is given, the better it gets at surfacing the product to the most relevant people.
Brands aren’t just missing out on incremental search visibility. Often, search terms which include attributes (like our “blue socks” example) convert at a higher frequency, as shoppers using them already know they are looking for a specific subset of product and are more likely to be ready to buy rather than simply researching their options.
Different product data schema across different channels have different degrees of complexity and depth. Often, there is a long tail of potential attributes to add to a product. It can be incredibly hard for ecommerce teams to know where to focus their energies enriching their product data so that they get the most bang for their buck.
Artificial intelligence can streamline this whole process from start to finish.
One data set for all channels
Firstly, the process of inputting data can be made much easier if it only has to be done once. Mapping to channels is a manual task that can be automated by machine learning, trained to understand the complex taxonomies of different channels and understand which values from channel A need to be changed to work on channel B, and what to change them to.
That means there’s one master product data set to work on which can then be intelligently mapped out to relevant channels, without incurring all the extra work.
Suggesting data-backed optimisations
Having just one data set makes going out to the market simpler and more efficient. That should free the ecommerce team up to make those optimisations which improve the relevance of products and thus improve their visibility.
That’s the second area that artificial intelligence can help with. By monitoring the interactions customers have with product listings across channels, Volo’s AI software can recommend optimisations, which can be actioned straight away. For example, changing a value to be more in line with customer search terms is a quick way to improve relevancy and visibility. The software can see the whole picture, and reflect customer behaviours by changing product data to better suit how consumers search.
Why it has to be intelligent
This problem is by nature suited to the strengths of intelligent software - the potential data mappings represent too much information for any team of people to perfectly recall, and software in this area currently represents not much more than a user interface upgrade from Excel. That means it cannot help team members with what to do next, or take any of the work itself away from them. Intelligent software can take away the manual work burden and help the team make the right optimisations, based on a wide pool of relevant data.