Ecommerce sales in the United States grew almost 15% in 2019, and thanks to the events of 2020, that percentage is set to continue its increase. With online retailers competing against each other harder than ever before for a share of the market, excellent search engine optimization has become crucial for improving conversion rates and profitability. One of the best ways to improve SEO is to provide search engines with adequate contextual information. The addition of structured attributes from product content helps search engines understand the context as well as the consumer intent behind search terms. Which words do potential buyers use to locate a specific product? And which words lead buyers to items that they end up purchasing? Deepening their understanding of consumer behavior helps retailers provide search engines with the most useful and effective content.
If this process sounds expensive, that’s because it often is, in both time and funds. Most retailers add structured product attributes manually, by outsourcing the process to third-party labor providers or via crowdsourcing platforms. Firms with the capacity even invest their own employees to validate the quality of enriched data provided by product suppliers before onboarding products and making them available to the consumer, search engines, and online catalogs. Even with outsourcing and crowdsourcing, this process takes excessive time and money, and is extremely error-prone. As product catalogs for some companies increase by thousands or even tens of thousands of products per day, it becomes almost impossible to continue the process manually or to rely on human labor for validation.
To enrich their product catalogs and keep up with the constant stream of new products to be onboarded, businesses are increasingly turning to automation solutions. The value of automation in this context cannot be overstated: businesses can add millions of dollars in revenue through both cost savings and additional sales. One retail enterprise, implementing the CrowdANALYTIX DataX solution for automating the creation of product attribute data, saw conversion rates increase 3x, product onboarding speed increase by 90%, and product returns decrease by 35%. The resulting increase in revenue was immediate and sustained.
Automating the creation of product attribute data essentially teaches AI to think like human beings do when it comes to tagging, annotating, and classifying products, whether they are represented by images, PDFs, unstructured text, or other types of data.
Once the process of producing product attribute data has been successfully automated, businesses can begin assessing data quality. Ongoing assessment leads enterprises to deep insights into customer behavior, changes in demand, and search engine optimization that can help them make lucrative business decisions and save even more. After automation, data quality can be assessed using three key measurements:
- Completeness
- Accuracy
- Conformity
If a company can leverage AI automation to regularly measure and maintain these important parameters at high standards, it can better understand consumer intent in searching and take action on that understanding to improve customer search results, and therefore conversion rates.
How do businesses determine the quality, both initial and ongoing, of the product attributes being automatically added to their catalogs?
Accuracy
The accuracy of product attributes is fairly simple: it is a percentage measurement of the applicability of the data given. For example, if a customer searches for “denim top,” they may see results including denim tops as well as items classified as “blue shirt” or as “navy button-down,” but they should not be presented with pants or with yellow tops or with, for instance, household supplies.
A search for “denim tops” on Nordstrom includes both accurate and inaccurate results, as you can see below. This would lower the Accuracy level and therefore the quality of the company’s search results.
Consistency
Consistency is the requirement that product attributes be uniformly formatted across product types internally. For example, a single business should consistently use either “trousers,” “pants,” or “bottoms” to label a single type of product, rather than vacillating between terms. This makes it easier for customers to surface products within the business, and easier for them to find the same products repeatedly.
Conformity
Conformity is much like consistency, except that it refers to adherence to external standards rather than internal ones. Maintaining conformity requires some research, since the business needs to attempt to use attributes well in-line with those used externally by similar companies for the same types of products.
For example, a failure in product attribute conformity might result in a search for “floral perfume” returning only four results, while a search for “floral perfume for women” results in ten.
Because most businesses only market floral perfumes as products for women, conforming to that attribute is important in helping the right customers surface the products they want to buy.
By assessing and monitoring all three of these important attributes in product onboarding, companies can gain better search engine performance, which can lead to increased conversion rates and higher customer satisfaction. ROI is quick and profitability can be maintained when businesses choose the right automation solution.