Without data there’s no AI — but without getting people, process, and principles aligned, there’s no AI strategy for your business. Learn how to ensure your AI deployment is successful, and how to avoid loss of money, trust, and your brand’s reputation along the way, when you join this VB Live event!
Register here for free.
Evo, the online sports retailer, is a data-driven company in a lot of ways, says Nathan Decker, Director of Ecommerce at the company, but he also says that for years artificial intelligence had been a bit out of reach.
“We don’t have the budget or resources to have a full-time data scientist in house,” he explains.
But as the AI vendor world continues to grow rapidly, an AI strategy is now accessible now to companies that were previously just pressing their noses against the tech store window. At Evo, they’ve aimed their focus on solving real-world problems — for instance, getting the right brand stories to the right customers at the right time, or targeting and timing.
“Human beings, myself, and my ecommerce team, had come up with some rules that were honestly fairly arbitrary,” he says. They didn’t want an automated email to go out more than once a day, so they stack-ranked the 15 different messages they could send to anyone so that, should someone qualify for more than one particular message, the most relevant would be prioritized.
It worked fine, he says, but once a customer triggered a selection of automated messages by hitting the evo website, one message would be sent and the rest would expire. So they had to go back and add additional rules, which initially started out quite simple — cart abandonment triggers, for instance. And those seemed to be working out so well they went back and built out additional rule sets to fill in the gaps of this manual system.
And soon that manual system not only became overwhelmingly complex, but there was no way to measure whether they were getting it right in their customers’ eyes. Were they choosing the right prioritization? The right frequency? These kind of questions could be answered in regard to specific email campaigns, but tests would add another layer of complication to implement, and the results would be temporary in nature, because at any given time the metrics they were measuring could change.
“We were frustrated with this reality,” Decker says. “But AI lets us home in on the right strategy.”
AI enables the shotgun approach: rapid-fire, large-scale experimentation that is impossible to do manually, but learns across a huge variety of frequency and audience and message settings, he says. Over time, as the model trains, it continues to optimize for the performance metrics the team selects, such as open rate, click-through rate, unsubscribe rate, and so on.
Depending on how much data you have and how many messages you’re sending, it takes a while to train the algorithm, as the tool zeroes in on the right combination for each customer.
And at first, they actually saw performance decrease. In the beginning, every potential mix of message and frequency is on the table, and some of those combos are just not going to cut it with customers. But the longer the system trains, the smarter it gets, homing in on how to make audiences happiest, and performance grow.
But even at the end of the official algorithm training period, the system never stops learning, keeping a test group within the normal send and message cadence, which it uses to continually try to find better formulas, and continues to try things that are more or less aggressive and while optimizing for the metrics chosen.
The company is not stopping there, Decker says — and AI is no longer just the purview of huge enterprises.
“There is a high barrier to AI, but those barriers are coming down,” he says. “The vendor community is enabling even small to medium companies to take advantage of AI.”
To learn more about how AI is changing the game for every company, where to get started, and the five fundamentals of AI readiness, don’t miss this VB Live event.
Don’t miss out!
Register here for free.
Attend this webinar and learn:
- What you need to do to prepare for AI beyond the data science team
- Real-world examples and research findings
- Top 5 best practices for strategic AI implementation
- Nathan Decker, Director of Ecommerce, evo
- Ken Natori, President, Natori Company
- Jessica Groopman, Industry analyst and founding partner of Kaleido Insights
- Rachael Brownell, Moderator, VentureBeat