Optimized E-commerce Content: A Pragmatic Roadmap to Boost Your IAG

Lucia Patalacci, Manager of Data Technologies and AI at Converteo, helps our clients discover the opportunities offered by AI and generative AI to improve business processes and identify high-value use cases.

Key Takeaways: 

• The adoption of generative AI aims to increase operational efficiency and enhance marketing performance.
• Generative AI enables the automation of time-consuming tasks such as writing product descriptions, verifying product attributes, and SEO optimization.
• Despite its immense potential, a strategy of gradual and iterative integration is recommended to adopt this emerging technology while carefully tracking and measuring its impact.

Last year, a McKinsey study highlighted the immense potential of generative AI in the marketing sector. According to their estimates, the application of this technology could account for up to 10% of overall functional spending on marketing and 4% of spending in the sales sector. In financial terms, this amounts to approximately $463 billion for marketing and $486 billion for sales, respectively, worldwide.

In recent months, within the pilot projects we have conducted for our clients, we have sought to validate these estimates through practical experience. These initiatives have proven promising, offering significant efficiency gains, cost reductions, and growth in key marketing performance indicators.

Varied Use of Generative AI

Generative AI can be used in marketing in various ways to enhance either the efficiency or creativity of e-commerce content and optimize SEO. Here are some challenges for which we have been approached by our clients:

Generation of Category Content and Product Descriptions: Writers are seeking effective ways to reduce the time spent on writing product descriptions for e-commerce sites. They need to continuously create unique and engaging content, often drawing inspiration from images to enrich their product descriptions aimed at the general public. AI has proven particularly effective in generating product descriptions, significantly reducing the processing time and effort required for ideation by writers. The implemented solution maintains a high level of quality and specificity by providing suggestions tailored to the tone and various styles of the brand, defined by the writing team. The observed benefits are considerable: a 75% reduction in human time, equivalent to a return on investment (ROI) of approximately four and a half months.

Search Engine Optimization (SEO): Our clients are looking to reduce the production costs of articles while improving the SEO performance of their website. They are interested in the automated generation of articles in HTML format to strengthen their positioning on search engines. AI has enhanced digital content, particularly technical elements such as page titles, image tags, and URLs. Through our project with Groupe SEB, we were able to produce about one hundred pages via AI, which, according to Antoine Pourron, International SEO Manager at Seb, exhibit “a level of quality that is relatively good to be published as is on our websites.” The primary objective is to maximize the visibility of the group’s e-commerce sites, representing nearly 75 e-commerce sites worldwide out of the 200 sites managed by the group.

Reduction of SEO Noise: SEO professionals analyze the relevance of the content of each URL page based on the keywords that generate the most traffic. As this activity requires handling a large volume of data, they are interested in automated solutions capable of identifying and automatically redirecting pages with significant semantic overlaps. This prevents overly similar pages from competing with each other for the same search queries. The combination of generative AI with a statistical classification methodology has enabled us to automatically measure the semantic similarity between traffic-generating keywords and those extracted from the content by generative AI models. The results allowed us to build various scenarios for grouping URLs into similar classes, pages to be redirected to those with higher traffic, potential semantic silos, as well as opportunities for internal linking to improve the user experience (SEO) on pages covering similar topics.

How to Develop an Effective Strategy to Leverage This Innovative Technology?

In the constantly evolving landscape of digital marketing, the use of AI, which is also rapidly and widely expanding, becomes a crucial element for staying competitive in terms of reducing operational expenses while achieving contextual efficiency gains and generating a positive ROI in the short term.

But how can you develop an effective strategy to leverage this innovative technology? Here are the lessons we have learned and our convictions.

• “Think Big, Start Small, and Iterate”

It is essential to adopt a gradual and iterative approach when integrating AI into the company’s overall marketing strategies. Start by identifying a small but significant use case that has a substantial impact on operational processes in terms of time-consuming, low-value activities that can be easily automated without major organizational repercussions or requiring significant changes in organization, culture, or human resource management. This approach helps to limit risks and investments while delivering tangible results in a short time frame.

• The Quality, Quantity, and Relevance of Internal Data Play a Crucial Role

To ensure that the results generated by AI consistently align with the assigned task, it is essential to build a qualified database to specialize the LLM model, providing it with the most accurate context possible based on the use case. This approach significantly influences the model’s learning process (fine-tuning). Furthermore, the management rules that feed the AI model provide a secure context for the requested task, although they may, as constraints, limit the creativity of the responses.

• Demonstrate Precision in Query Writing

The process of optimizing AI responses also involves the use of fine prompting techniques, which consist of the information, instructions, examples, and ways to organize them that must be transmitted to the model to obtain qualitatively and quantitatively optimal and contextually consistent responses. For example, the instructions differ if a poetic description is needed versus an explanation or technical procedures. The “chain of thoughts” (CoT – Wei et al, 2022) technique is also known, allowing AI to acquire complex reasoning skills through intermediate reasoning steps presented using examples.

• “Humans in the Loop”

Understand where the limits of this technology lie (and be ready to make corrections if necessary), ensuring that the responses are consistent with specific queries in order to better measure the gap between prescribed rules and excessive creativity (so-called “hallucinations”) in the results. For example, it often happens that the responses do not respect the required word length or mention the names of competing brands but are nonetheless very rigorous regarding the use of discriminatory or offensive expressions.

• Measure ROI

The applications of AI in limited fields allow us to measure tangible short-term benefits: measuring AI response latency compared to human time, operational costs of tools, and SEO performance gains provide a reliable estimate of the coverage of the investment in a Proof of Concept.

• Integrate AI into Existing Business Processes

Our experience has confirmed that the automated component of AI can be gradually integrated into existing processes without substantially disrupting the data architecture or existing processes.

• Looking Ahead: Towards a Fully Chatbotized E-commerce

In the future, the rise of chatbots based on generative AI could profoundly change the customer journey, shifting from traditional search engines to an interaction experience based entirely on natural dialogue. This new type of interface would lead brands to completely rethink their SEO strategies to offer more personalized and conversational shopping experiences.