'Personalisation at another level is the next step'

Lars Hard

  is the CEO of Sweden-based Expertmaker that helps automate and optimise data-driven decisions and processes for retail, manufacturing and the Internet of Things (IoT). In an interview to Furquan Moharkan of Deccan Herald, Hard says that eCommerce portals are incorporating artificial intelligence (AI) — which is a culmination of one of the most fascinating technology topics today, and the newest way by which people shop.

In what areas have eCommerceportals applied AI?

AI is mainly used for recommendations and personalisation, both on the sites and in email marketing. It has so far been mostly successful in product areas like music, films and other cheap volume goods driven by personal taste. However, it has also been tried with more limited success in other areas. Product categories that are not ideal for simple recommendations can still benefit from the technology, though, impact would be less.

What are the reasons behind the fact that eCommerce portals felt the need of incorporating AI?

The simple reason is that it drives revenues through personalisation and recommendations. It is estimated that Amazon makes 30 per cent of its revenues from simple recommendations. As competitive pressure increases, so does the need for personalising more difficult product categories such as apparel and others that do not fit with the simpler technology that requires high volume, low price and taste-driven kind of products. All these require more advanced AI. So we will see an increased activity as the market matures.

How has AI impacted the eCommerce business model?

Recommendations and personalisation have become essential for companies to thrive. The increased sales volumes through recommendations is factored into how the margin is calculated.

What are the further avenues of eCommerce where possible use of AI can be explored?

The products that do not meet the criteria of low price, high volume taste-driven, such as apparel and pharmaceutical products, will be the next area to be attacked using more advanced AI models. Image recognition can be used as a stand-alone, but it is still difficult to get that right. When working, it obviously lends itself to searching for products and automatic matching of combinations, for instance, clothes and accessories.

Image recognition can also be used as part of an advanced recommendation and personalisation solution to understand the customer on a deeper level. Natural language processing (NLP) search with real time interaction will also help eCommerce companies evolve, enabling users to search and interact using natural questions and answers.

How has this impacted the profits of the companies?

Most companies will reduce both cost and improve margin and revenues. While the search functions will mainly drive revenues, recommendation and personalisation are powerful technologies for driving revenues for both upselling and cross selling, and also reduce cost by helping companies learn what is in demand. This will help them focus their resources on those products, thus reducing unwanted inventory. Price optimisation is all about finding the right price points for the market. This is naturally interesting as it creates an optimum price. When taking it further to individualised pricing, it enables companies to drive additional revenues and also create a happier customer as the personal price point is met. This creates both higher revenues and a more loyal customer. It can also be used to reduce cost by making sure that surplus inventory is minimised.

As NLP and image search are introduced, they will improve the user experience and enable users to find products they otherwise might have missed, which leads to higher sales. Both NLP and image technologies are particularly interesting for mobile devices as the interface is constrained.

How do you see the scope of AI in eCommerce in the coming years in India?

As eCommerce players have already adopted simple AI technologies for easier recommendations, we see advanced personalisation to be the next step. This will enable better recommendations and personalisation for many new product groups. As competition increases, other parts of the eCommerce operations need to be improved too. We see demand forecasting, and price optimisation as the next step outside personalisation.

Price optimisation can quickly become a complex multidimensional challenge and taking it further to individualised pricing makes it even more complex, but it will be important to be the most competitive player in the market. There will be tests on image recognition and NLP to see what really works too. Not only is the technology complex, but it will also take time to get the users to change their behaviour, and start to use the new technology for more than test purpose.

Given the fact that AI in itself is still evolving, how do you think, it will affect R&D of eCommerce portals?

Different AI technologies work on different product categories. What works for film recommendations, does not work so well for apparel or hardware products. As new AI approaches and new ways to apply AI becomes available, new opportunities open up and there are more than one way to solve the problems. It is important for companies to test what works for their particular circumstances.

Optimisation technology can be difficult as many optimisation problems quickly become very high dimensional. With the increasing speed of computers, both new and older approaches, can be used to attack these complicated challenges. Companies need to start out with a simpler scope and evolve over time, testing various approaches.

Any particular difference between the kind of AI used by the US, and Indian companies?

AI technology is progressing across the world. Both commercial products and open source ones are available worldwide, making it possible for Indian companies to compete with the same technology. It is more a question of picking the right approach and platforms for the right problems and products. However, skilled resources are in high demand and not easy to find some platforms that are more expensive to work with, at least for the short-term.