Retail AI predicts consumer behavior for targeted marketing
Not all of travel club Secret Escapes' members want to see all 140 last-minute sales it runs every week. AI marketing personalization tool Jetlore helps prevent over-messaging.
Chances are, if you have 10 million travel club members scattered across 12 countries -- making you a top-10 volume travel retailer in Europe -- someone's going to be interested in the latest limited-availability, four-star hotel flash sale you're about to post on your site.
The problem is, with 250 deals running at any given time and roughly 140 deals opening and closing weekly, finding that someone without blanketing too many members with too much email -- to the point where they ignore them -- becomes a difficult task. That was travel vendor Secret Escapes Ltd.'s technology issue: Controlling overall email noise while still growing sales via email marketing automation.
Enter retail AI, which examines data points of consumer behavior and, combined with rules-based marketing automation, cashes in on sales by predicting consumer behavior in future transactions. This means companies can create targeted promotions instead of bulk mailing everyone all the time.
Secret Escapes added retail AI vendor Jetlore Inc. to its tech stack, which uses Salesforce as its foundation.
One-to-one marketing has gotten so precise, said Ollie Miles, global head of CRM at Secret Escapes, that his team doesn't look at emails as marketing campaigns anymore, but instead as transactional tools.
"It's bookings and margins that we care about," he said.
Retail AI in action
While all that might sound pretty complicated, the bottom line is that Secret Escapes can figure out, for example, that a couple suddenly looking for pet-friendly romantic getaways just got a dog, and they want to travel with it. The company can then offer a different list of last-minute hotel deals than before -- without wasting everyone's time filling out forms, manually changing preferences and updating Salesforce records.
Via HTML, Jetlore tracks image and link clicks in combination with website pixel tracking and transaction data. After analyzing tens of thousands of data points for a particular customer -- and customers like them -- the artificial intelligence component makes recommendations for future offers Secret Escapes can send to that person. It's sort of a B2C analog to the more familiar B2B account-based marketing.
Before implementing Jetlore, Miles said Secret Escapes evaluated several other vendors, including Centage, Peerius -- acquired by Episerver in 2016 -- and AgilOne. Jetlore was the best fit, he said, because of the nature of Secret Escapes' business -- last-minute sales that expire quickly. In tests, getting effective results for the other vendors' platforms either took too long to dial in, were inconsistent or degraded too quickly. It was also by far the easiest to use, in Miles' view.
B2B benefits in a B2C model
Jetlore, also used by marketing teams at consumer-facing brands such as eBay, appliance manufacturer LG and clothier Uniqlo, doesn't directly integrate with Salesforce Marketing Cloud at Secret Escapes. Instead, Jetlore generates HTML snippets for Marketing Cloud email templates, built dynamically in Salesforce's AMPscript language, that also provides tracking data.
The concept and execution of B2B CRM are very well understood, said Jetlore CEO and co-founder Eldar Sadikov, referring to how companies with typically hundreds of accounts have teams manually updating CRM records, with some automation help, to keep them current and keep the sales coming. But when the numbers scale up and margins are smaller in the B2C world, all that record-keeping becomes too costly to maintain.
"B2C companies -- retail, media, hospitality -- they have thousands, millions or at least hundreds of thousands of customers," Sadikov said. "The challenge is that there's no human going in and filling out a lot of data about each of the consumers. There's very little structured data about those consumers. All you have is typically what they ask for in their registration form -- name, email, maybe your gender and that's about it. Beyond that, you have a huge log of activity those consumers are [doing] online. How do you take that huge log and map that to meaningful, structured data?"
Retail AI can help build profiles for consumers when companies can't otherwise afford to do it manually, according to Sadikov. For instance, clothing e-tailers can use artificial intelligence to figure out a person's style and fabric preferences, size, what colors they enjoy and many other data points. Platforms like Jetlore can take those data points and map them to current promotions and help determine which customer cohorts would be more likely to act on them.
AI performance metrics fine-tune campaigns
Of course, marketing tools can be pretty rudderless if there's no reporting on their performance and fine-tuning in successive campaigns.
To keep tabs on how the AI is -- or isn't -- creating sales off its recommendations, Jetlore sends daily CSV files that break down campaign and product performance, broken out by country, into Secret Escapes' Salesforce Marketing Cloud, which dashboards the data. Secret Escapes then uses Marketing Cloud's SQL engine and data tables to merge data and append additional product and campaign detail to build out full performance reports and adjust the algorithms.
Secret Escapes saw an immediate 11% email click-through rate improvement in the United Kingdom, its biggest country market, after implementing the retail AI tool. Subsequent tests showed click-through maintaining most of those gains. Email sales conversion rates also improved by about 5% right away, and have held through subsequent performance monitoring.
Don't get lazy and use it as a silver bullet.
Ollie MilesSecret Escapes
Miles said he believes the key to Secret Escapes' success with AI-driven email marketing personalization -- and his biggest piece of advice for other brands that might try their hand with similar tools -- is keeping up with the pulse of their customers and adding that intelligence to their system. Knowing when to use it, when not to, and where to use it on the customer journey is key.
You understand general buying trends among your customers, and they might change every six months or even more quickly. Adding your own sales knowledge -- and common sense -- to the AI's findings is crucial to keep performance consistent. It's not one of those set it and forget it technologies; it needs consistent human input, Miles said.
"Don't get lazy and use it as a silver bullet and say, 'My content is 100% personalized, just press go.'" Miles said. "These things work less well the higher up the funnel you get. With us, behavior data starts losing relevance very quickly."
