Case Studies

Marketing a new product on a market oversaturated with products and ad campaigns is getting increasingly difficult. Companies struggle with ways to position their products to stand out among competitors, while the move to online selling platforms have added a new layer of work - creation of online marketing materials, like blog posts, ad campaigns, brochures, e-books, social media posts, white papers, etc.

One of the most important factors marketers need to take into account when positioning the product on the market is how it compares to competitors and what factors favourably distinguish the product. Another important part to determine is what problems is determining what paint points the product solves for the customer. Analysing these factors is key to a successful marketing campaign.

Our client decided to take advantage of the latest developments in natural language processing and has come to us with an interesting project aimed at automating the market analysis. Our goal was to develop a system that would determine customer pain points and determine what people like and don’t like about similar products already on the market. This information would then be used by marketers to create promotional content aimed at the target audience.

The system had to receive a website page describing the product in question as input, determine what the product was, gather the public’s opinion on similar products and determine main advantages and disadvantages according to the customers.

While working on this project, we had to address the following challenges:

• The system had to successfully determine what the product in question is
• We had to determine where the system would go to gather the public’s opinion
• The system had to integrate GPT-3 and produce human-like written content

As the project involved natural language processing and generation, we have decided to work with GPT-3 - a language model that uses deep learning to produce human-like text. GPT-3 is the best language generation model in the world, creating texts which are difficult if not impossible to distinguish from those written by a human. This was an easy choice for us since it is the most advanced English language model at the moment.

First, the system receives a landing page with the product’s description where it collects keywords to determine what the product is and its main features. The best option for analysing customer pain points is to pull reviews for similar products from online shopping platforms. We have chosen Amazon as the platform from which the reviews would be pulled as it is one of the largest online retail platforms with millions of products and thousands of customer reviews.

The system looks for reviews to similar products on Amazon and pulls them out, after which we access GPT-3 via an API to generate a text detailing the main advantages and disadvantages of the rivaling products. This information can then be used as the basis for any marketing content or as marketing content in itself in a form of a script for a promotional video or a social media post.

The system we have developed greatly improves marketing efforts as the text produced at the end is based on real customer opinion, while the quality of the generated content makes it indistinguishable from human written ones. The system has great potential in not only aiding marketers in creating content, but creating content altogether. As natural language processing and generation tools get better, the need to spend hours writing marketing texts greatly diminishes allowing more time for creativity.