Understanding Algo-Commerce [a Q&A with Feedvisor]

Like it or not, marketplaces (think Amazon) play a critical role in selling online today.

There are positives, of course, like getting products in front of more people than a company's own acquisition efforts would allow. There are also challenges and complexities that marketplaces bring, however, like marketplace rules, inventory and prices.

Feedvisor operates in the latter by using machine learning algorithms to help third-party Amazon sellers remain competitive in the marketplace. Their revenue intelligence informs sellers in real-time what's the best price for profit for each individual item.

As part of our "4 Companies You Didn't Know Worked in Artificial Intelligence" feature, Website Magazine caught up with Feedvisor CTO Eyal Lanxer to learn more about the company's offerings.

What are some common pricing scenarios sellers face on Amazon (to the detriment of profits)?

One of the main challenges in selling on Amazon is to determine the optimal price per product and time, given numerous factors and market conditions. Such factors are typically broken down to four main groups: The seller's characteristics (e.g., positive/negative reviews, shipping method/time), business goals and constraints (e.g., minimum margin); competitors' characteristics, business goals and constraints, as well as their pricing strategies; the marketplace's ranking system (the Buy Box in Amazon's case); and the consumer demand per product and time.

In addition to the complexity derived from these market conditions, prices on Amazon can change almost instantaneously, and for high-volume sellers, it's incredibly difficult to ensure every item's price continuously reflects market conditions, while also remaining priced to sell at a profit.

How does machine learning help improve those scenarios?

With the help of machine learning algorithms, the above aspects of the market are analyzed and factored into an optimal pricing decision. Such algorithms track the arena over time and apply the desired price adjustments, ensuring that sellers respond to the changing market conditions and are able to make a profit on their items. As the learning phase progresses, the algorithms are more capable of identifying the seller's competitive edge within the arena and helping them realize higher profits.

What/if any rules can be set by the seller to ensure automatic adjusting of their pricing doesn't impact their bottom line and or become non-compliant with minimal advertised pricing (MAP)?

One of the main advantages of applying machine learning techniques for pricing is that these free the seller from defining numerous business rules for pricing their products. Nevertheless, there is a minimal set of configuration points required to enable such pricing methods, typically - cost structure and minimal margin - per product. Based on these settings, we are able to determine floor prices that will ensure that we don't cross the seller's bottom line or breach their MAP constraints.

What expectations do you have for machine learning when it comes to marketplace selling in the next year or so?

Tracking marketplaces over the past years shows that pricing on these platforms is becoming more automatic and more dynamic. Hence machine learning is a huge area for growth in marketplace selling, and not just in the next year or two. Because these marketplaces are so dynamic, and often complicated, it's unrealistic for sellers to scale and remain competitive without some sort of artificial intelligence solution. Machine learning and AI are a necessary addition because they can analyze data and make business critical decisions in a fraction of the time that humans can, enabling sellers to focus on the core aspects of their business.