It all begins with knowledge. Knowledge has the utmost value in our lives and it’s what differentiates us from other species. The more knowledge we accumulate, the more our ability to make faster and better decisions increases.

This is exactly what Machine Learning scientists have been doing, ever since the ‘50’s:
Feeding knowledge into machines, in order to teach them how to perform certain tasks. The purpose of Machine Learning then was to eliminate human error and decrease the time needed to perform these tasks.

What has changed since then?

During the last decade or so, technological advancements in Cloud Computing have brought forth the democratisation of Machine Learning and Artificial Intelligence. Adding to that, enormous amounts of data are now available – a prerequisite to embark on AI – most of which consist of data that most e-commerce companies probably already have. We can now have machines processing vast amounts of data, making them smarter, day after day. This is what’s made AI accessible, affordable and important to people and companies alike. So important and of course easier to implement, that I believe that AI will soon become commoditised.

The AI technology is maturing, but what about you and your company? My experience has shown that, before jumping into bold decisions, it’s always better to assess your maturity level with AI, understand where you stand and then take action.

So what can you really do with AI today? Even if I limit the AI applications only to the e-commerce industry, the list is still long. I have picked though a few of them, which are feasible today and can get you real results in conversion rates, profits and operations. Although some might seem common and easy, I do believe that you will find some hidden success factors in there.

PERSONALISATION

From rule-based to AI-driven

Personalisation was a buzzword about 10 years ago. Those days, delivering personal experiences on digital properties was a rule-based game. The problem was that it didn’t get truly personal and it was hard and expensive to scale. In my opinion, that is why personalisation didn’t take off at the scale to match the hype it had gained.

AI has shifted the paradigm and, instead of combining sets of rules, it uses your accumulated knowledge in the form of data, to predict the most relevant content or the most probable action for every single of your customers.

Product Recommendation

I’ll start with the good old product recommendation. The “You might also like” or “Recommended for you” type of suggestions inspire familiarity and desire throughout the navigation flow, boost internal linking and enhance the overall user experience. Most importantly, we have seen conversion rate increases of up to 6%. Very basic practice, one might say, but also very important.

Cross and up selling Personalisation

The moment when a customer decides to make a purchase is one of the most pivotal steps of the customer journey. There is no better time to leverage the consumer’s purchasing desire, to trigger what Grant McCracken named the “Diderot effect”. On the consumer’s side, obtaining a new possession creates a spiral of consumption – and you should be there and take action, at that very moment. AI enables the scaling of cross and up selling and can significantly increase the basket value.

Personalised product list sorting

Throughout my career, I have discussed the matter of product sorting in an e-shop’s product list with every single one of our clients. This is another rule-based function that can become intelligent; and AI is the way to do it, by delivering a product list that is relevant to each customer by pushing the most relevant products above the fold. AI-empowered product list sorting is particularly effective for large product catalogues with thousands of products in hundreds of categories. It is, though, an application that, once configured properly, requires little effort to maintain.

Personalised newsletters

All the content, whether it be offers, products or any other kind of information shared through newsletters, is an obvious AI personalisation practice. Taking it one step further, you could adjust the time (day of the week and time of the day) each newsletter is sent out, according to the unique behavioural habits every person has, in terms of handling incoming e-mails, especially promotional ones. A not so common, yet extremely powerful way to increase newsletter open rates, decrease spam complaints and boost the efficiency of your campaigns.

DYNAMIC PRICING

Maximising profit through AI

You might be familiar with dynamic pricing on the airlines or car rental industry. Dynamic pricing in retail is a mechanism that uses supply and demand information, stock availability and historical data, in order to predict the price that will generate maximum profit for a specific timeframe. It is a balance between sales volume and margin and can be used in the context of offers, discounts and dynamic bundles. The value of dynamic pricing is more evident during sales periods, such as Black Friday: besides the real-time and automatic price configuration, it provides an intelligent, data-driven tool to help humans adjust their pricing strategy in such competitive times. In fact, there are case studies out there that have recorded almost 1,5% increase in profit, by using AI-assisted pricing strategies during sale periods.

ANOMALY DETECTION

Know what is happening and know it on time

Anomaly Detection is one of the hidden gems that can affect considerably e-commerce operations and performance, as it provides visibility to everything that is happening in your e-shop, in near real time – a process that humans would need days, if not weeks, to carry out.
Having an anomaly detection that can monitor streaming data requires Machine Learning algorithms that, not only instantly and accurately identify anomalies, but are also smart enough to adjust to seasonable behaviour. AI, once again, can take this to the next level and break down your KPIs by country, brand, age group or any other demographic or dimension that your data can provide, empowering you to make smart and timely decisions.

THE SUM IS THE GREATER THAN THE PARTS

Combining scenarios could change the rules

Imagine now how many possibilities arise when you combine different AI applications into a more sophisticated and inclusive campaign!

This is the power of AI.

ONE FINAL NOTE

To conclude, let me take a step back and talk about acquiring AI maturity. This is where it all starts, so I would like to share with you 4 tips to help you begin your maturity journey in AI:

1. Take top management on your side

Most of the times, top management will agree to practices that generate more revenue. So pick your first experiments wisely at the beginning, to make top management AI advocates. Sponsoring for your projects will be much easier.

2. One man show is a no show

You should set up a cross-functional team, involve tech, finance and marketing and get help outside of your organisation if needed. Apart from passion, AI needs diverse brains.

3. Get your technology sorted

AI requires technology. You can buy it, you can build it or even go hybrid, but I strongly recommend you get consultation from an expert to guide you through this decision.

4. Make it valuable

AI experiments should “comply” with specific attributes, in order for them to be of value. They have to me actionable, measurable, scalable and, of course, they must make some kind of impact. If more than one of these attributes are missing, it’s probably a good idea to invest your resources on a different experiment.

I know I just scratched the surface of how AI can be applied in e-commerce operations, but I hope I have provided some useful insights and inspiration to those seeking to get involved with it. The opportunity is here. The applications are countless. It’s up to you to strategically plan your AI roadmap!

Inspired by Jason Kataropoulos's presentation "Real life, real results: applied AI for conversion optimization" at the e-Commerce Conference 2021, held in Athens, on April 15, 2021.