Artificial Intelligence for e-commerce data

Shalion’s Keynote at Beauty Innovation Days 2020

Today, February 20th of 2020, Valeria Bellani (CCO) and myself, Enrique Herreros (CTO) from Shalion gave a talk for the Barcelona Beauty Cluster. The venue took place in the beautiful Pompeu Fabra University. We had the pleasure to listen and share innovative advances, looking out to how the beauty industry advances. Our talk was offered in this forum for 2 reasons:

1. Some of our clients are cosmetic selling companies
2. Fashion and beauty are booming on the Internet: sales are forecasted to be 18% higher than 2019, with a global figure of 620.1 billion dollars (source: Digital Overview 2020).

We can all agree on the great importance that online commerce is acquiring and its relevance in the growth strategies of companies. But perhaps we are not yet so aware of how big of a challenge it is for brands to manage this promising sales channel. The thing is: ecommerce does not follow the rules of the offline world. It changes them completely. And this causes Category Management to get uncontrolled quite easily. In other words, we can’t access the necessary information to optimally manage the distribution and sale of our products.

ECOMMERCE DIMENSIONS

So, what are the new rules imposed by ecommerce?

Assortment & Stocks

We no longer directly manage the stock or the assortment but it is the retailer who manages the inventory. And there are thousands of retailers and intermediaries, plus new ones appearing every day. The sale of our products depends on them, they are the ones deciding whether to sell the product and to replenish its stock when needed. Retailers are the ones responsible for choosing the assortment according to the target and the occasion of purchase. In summary:

• Thousands of retailers
• New retailers popping up constantly
• Retailers managing poorly your portfolio
• Out of Stock increasing your opportunity cost

Pricing & Promotions

With regards to the pricing, more and more retailers use dynamic pricing systems to assign a price to their products. Thus, the price varies constantly and become nearly unpredictable. Moreover, shoppers have access to offers of the same product at different prices. And the easy comparison with competitors’ discounts generates an escalation of promotions that ruin margins. In summary:

• Dynamic pricing causing continuous price variations
• Harmful promotions
• Offers based on user location
• Competitors tracking you

Search & Digital shelf

The share of shelf now depends mainly on search results, that is, how often our products appear on the first page of the most popular search keyword sets. And this dramatically varies with the search algorithms that the retailers use. This ranking problem may artificially prioritize products of their own brands, or of certain sponsors, or simply be poorly developed. In summary:

• No EANs available
• Multiple titles and variations for the same product depending on retailer
• Low performing SEO
• Multiple languages

Content & conversion

The conversion to purchase is highly influenced by the content included in the product pages. This content is defined by retailers and often does not respect the manufacturer’s guidelines. As a result, products are displayed with inaccurate titles, incomplete descriptions or poor quality images. All these causes poor indexing results of our products in searches and ultimately deteriorating our brand awareness, SEO, conversion and customer engagement.

• Unstandardized and low quality product images
• Poor descriptions
• Reviews are spread all over the web

A BIG DATA PROBLEM

To gain control over online retailers, manufacturer need data for each of these dimensions. Therefore, retailers need to collect their product’s data couple times a day at least in all the retailers that are meaningful to their business. Let’s do a quick example of the human resources needed for such a task for a medium manufacturer:

• Imagine you sell 500 products
• Your products are sold, in average, in 6 main retailers
• That is, you need to track 3.000 product pages
• In average you want to look those product pages couple times per day. That’s 6.000 visits
• Manually extracting this data would take you around 1 minute per visit
• In total, humans would spend 6.000 minutes gathering data, 100h/day
• A human works for 8h/day so you would need 13 people working full-time non-stop to get the data

Obviously, one must be insane to think this is the way to go. Any company willing to automate the scraping task as much as possible will create rules to reach the elements of interest (stock, price, category, title, etc). Such rules vary by retailer, by time and sometimes even by category inside the same retailer. We will discuss in a later post what the difficulties one finds when trying to scrape data from retailers at scale. For now, just keep in mind that a set of rules is known as an “extraction template” or “scraping template” and that those are needed in order to somewhat semi-automate the scraping process.

The Scraping task

Now let’s go one step further. Such template also need great operational energy. You are guessing right: if the website’s source code changes it might affect the extraction template, breaking some rules and losing data. When gathering data to feed our Product Monitoring tool, we need to consider hundreds of retailers, which burdens the operations workload. At Shalion, we have developed an AI tool able to automatically scrape the important product information from previously unseen retailers.

In the second part of this post we will introduce the AI weaponry that allows us to empower our client’s Category Management and Product Distribution providing:

1. (near) Real time data
2. Scalability
3. Accuracy
4. Synthesized, curated and analyzed data
5. KPI convertible information

Shalion is a company focused on collecting and measuring e-commerce data. Our EcomPilot product, supported by sophisticated proprietary technology of crawling and scraping, enhanced by Artificial Intelligence, allow us to provide you with the most accurate data for your decision making in ecommerce.

“Brand monitor” enables you to follow the offering of your products and your competitors within the different e-sellers allowing you to react and to take action regarding:

• Stockout of your products. Which retailers do I have stockout of my products?
• Price and historical price changing.
• ¿Who is wining the Amazon buy box (the first recommend seller option for a search of a product)?
• What is the score of my reviews in the different e-retailers?
• What % of products are in promotion?
• What is the coverage of all my assortment within the different e-retailers?
• How well is described my products in the different e-commerce sites?

And many more information we could provide.

Do you want to know more about our tool? Do you have the feeling that your products are being sold in multiple ecommerces but you don’t have all the control that you would like? Let us help you, drop a line to info@shalion.com