The realistic future of AI for e-commerce
Artificial Intelligence is changing the game for e-commerce marketing
A true one-to-one experience is the holy grail of e-commerce, and for good reason. The customer experience is the driving force behind successful e-commerce businesses. The days of negative revenue and buying market share in Southeast Asia are quickly coming to an end, and sustainable e-commerce is going to require a focus on user experience.
Today’s e-commerce customers demand you know how, what and when to communicate with them. Thus, if we want to deliver an e-commerce experience down the the individual level, we’ll need artificial intelligence.
One interesting application for this is with email marketing, which continues to grow in importance as a channel for e-commerce companies, but which arguably offers some of the worst user experience for many consumers. Day in and day out, the emails keep coming and keep filling up inboxes with irrelevant information that can ultimately just be downright annoying.
Also Read: E-commerce startups must evolve now or fall by the wayside
With all the excitement around creating a more personal e-commerce experience, concepts like segmentation, marketing automation, and personalisation have been drawing the most attention. While the available tech in this space is smart, literally every single user is on their own journey, and only AI can adequately sift through the data to act on an individual user-level basis.
How, then, is AI positioned to change the current landscape and what are we building towards? Let me break it down.
A word of caution: content is hard
Almost everything you read online about email marketing optimisation talks about content. What’s the best subject line? What kinds of images or designs work well? How can we personalise products and images based on user interest? While intuitively this seems like the obvious choice to apply AI algorithms, the reality is that really good AI for content may still be a ways away.
The problem boils down to not having enough data. Is it true that just because someone clicked on a promotional email that they wouldn’t also respond to a newsletter? If someone bought a pair of shoes and you’ve got a full line of fashion apparel, are you sure they wouldn’t also shop for dresses? If someone identifies their gender as female are you sure they won’t also buy something for their boyfriend or husband from the men’s section?
In an increasingly fragmented e-commerce space with thousands of choices and limited loyalty, most e-commerce companies lack the raw data to make content recommendations entirely meaningful. While good technology exists, it’s really geared for the biggest of the big companies and isn’t applicable outside of an absolutely massive data set.
An alternative to segmentation
Segmentation is a sensible step toward a more personal customer experience, but let’s just run through this scenario. Say you want to create a subset for the people who have ever bought something from your store. You should absolutely be sending different messaging to these people, but recognise that you’re still treating all 12,286 customers the same.
For most marketers striving for a one-to-one experience, the most common next step is to delve into more advanced segmentation tactics: gender, age, location, etc. Take it a huge leap further and you’ll hear people talk about dynamic content.
While I recognise the thought process, the problem with each of these tactics is that they require a mountain of manual labor to create the supporting content. If you’re not careful, it could completely eat away at your ROI.
Also Read: You’re not missing this one out: the 7 e-commerce trends to follow into 2017
The key question, then, is how to treat users individually without creating a labor cost for every individual user. At a bare minimum, it requires that you have data on every user. For example, past purchase behaviour may only be available for a small portion of your database.
One thing you are sitting on top of for every single user in the email marketing world is the email channel data itself — send, open and click history, which makes a very good candidate for AI as it’s systematically and readily available to marketers.
Think about it from your own perspective. You find a cool product. You leave your email. You receive branded newsletters for the first couple of days. It’s enticing. You click through to the site and it’s great. But as time goes on, it becomes less and less interesting. After a while, it becomes invasive.
That can’t be managed with segmentation. There are too many different users at a different level of interaction with the brand, but as a marketer, you’re already sitting on top of all the data you need. You know exactly how someone is interacting with your brand and what the state of their relationship is. You need artificial intelligence to look across every engagement to determine the right frequency and timing of communication for each individual user journey.
Also Read: Artificial Intelligence 101: A map to understand where we are now
It’s a practical application of AI, because the data is readily available, requires no additional manual work (if executed correctly), and therefore can be fully automated.
Rethinking marketing automation
Marketing Automation (drip campaigns, trigger campaigns, or web-behaviuor-based automations) offer the power of generating user response, engagement and sales without lifting a finger once they’re in place. But without AI-enabled automation running, marketers can at times generate more problems than they’re solving. As CRM and CLM marketers create increasingly complicated systems of automation, it’s entirely possible for a single user to be eligible for multiple messages on a single day, and a simple rules-based approach to determine which messages to send isn’t nearly enough.
AI has to be baked in, and looking across the entire ecosystem in an integrated fashion. It needs to make sure you’re delivering a good user experience by analysing everything that’s possible in the next 24 hours, predicting that and then only sending out the best possible message and not five.
The takeaway
The broad emergence of AI applied to all kinds of technology can rapidly improve the relationship between businesses and users, but a realistic approach is grounded in data — readily available data is the place to focus for now.
We’re already making serious headway at Ematic on the AI front, but we are rapidly seeing all sorts of efforts to leverage AI technology in the email marketing world. As you weigh the value of future AI implementation, just be conscious of these three things:
Look for tech that’s leveraging data inside or close to the email world;
Carefully evaluate the manual labor needed, and,
If you do venture down the path of AI for content, all bets are off, just make sure your data set is massive.
With these three things in mind, AI is a thrilling reality more and more e-commerce brands will soon be taking advantage of.
—-
Paul Tenney is Founder and CEO at Ematic Solutions, a Singapore-based SaaS provider that offers email intelligence platform that helps digital marketers nimbly build and deploy sophisticated and high-ROI email marketing programs.
A seasoned executive, entrepreneur and world traveler, Paul has been working in the B2C, Retention-Based Email Marketing space for over 10 years. From his early days working with clients like Hewlett-Packard, Macys.com, eBay/PayPal and many more, he got a front row seat to just how clunky and difficult email marketing can be, blighting what is otherwise the most brilliant marketing channel on the planet.
Paul will be speaking at Echelon Vietnam 2016. Join us for two days of high level conferences, interactive workshops and engaging exhibitions. Time is running out, so get your tickets now.
Image Credit: Pixabay
The post The realistic future of AI for e-commerce appeared first on e27.
