Artificial intelligence enables retailers to dive deep to find marketing strategies that work. But humans still have a role to play.

All it took for U.K.-based personalized picture book retailer Lost My Name to reduce its cost per acquisition 17% was to automate its Facebook advertising campaign.

While Lost My Name allocates some money in its budget for TV ads in November and December—when the vast majority of its sales occur—it spends roughly 70% of its paid marketing budget on Facebook ads, says Anne Thouas, the retailer’s head of awareness marketing. Until last year the retailer managed its Facebook advertising campaigns in-house. But as the retailer’s business grew and expanded into new markets, Lost My Name reached a point where it either had to hire more staff or an agency to help manage its Facebook campaigns or turn to an advertising automation vendor, preferably one that used artificial intelligence and machine learning, she says. After all, a number of vendors have popped up over the last few years with artificial intelligence-powered platforms that they promise will enable retailers and other marketers to efficiently run a large number of tests, crunch data, and quickly learn and adjust based on those results at a pace that would be near-impossible for a human to achieve.

Marketing offers an ideal use case for artificial intelligence given the large trove of data marketers have to work with, says Joe Stanhope, a Forrester Research Inc. vice president and principal analyst. “Our brains can only do so much,” he says. “They have a certain amount of cognitive capability. The customer experience is moving so fast that humans can’t keep up. We need machines to close the gap between what we can do as people and what we should do to engage with customers.” However, the technology is still in its early days and, for the most part, it is only beginning to help retailers improve their performance and boost engagement.

The customer experience is moving so fast that humans can’t keep up. We need machines to close the gap between what we can do as people and what we should do to engage with customers.
Joe Stanhope
Forrester Research Inc. vice president and principal analyst

Even so, Lost My Name was intrigued by the idea of using an artificial intelligence-based system to automate its Facebook advertising campaigns. The retailer’s Facebook ads largely consist of direct-response link ads that seek to drive consumers to the retailer’s site, along with some retargeting ads directed at the retailer’s existing customers and consumers who have visited the retailer’s site. In particular, it liked how’s technology enables a retailer like Lost My Name to choose a goal, say drive sales or clicks to a retailer’s website, and set a budget. The technology then tests the retailer’s ad campaigns to determine the optimal budget distribution for the stated goal.

To gauge whether automation would help it scale its Facebook advertising program, it pitted automation technology developed by Inc. against a staff member who manually allocated the retailer’s budget. Within a few days of launching its test, Lost My Name saw the technology shifting dollars around in ways that, on the surface seemed overly aggressive, Thouas says. For instance, on one day the system allocated 200 pounds ($255.80) out of a 2,000 pound ($2,558) budget to a particular ad set. The next day it increased that spending to 800 pounds ($1,023.20).


“We would never make such a big swing,” she says. However, the technology’s algorithm paid off as it boosted the retailer’s conversion rate by 36.1% and reduced its cost per acquisition 17.0%. Those results convinced Lost My Name to switch to The vendor typically charges 5% of the ad spend but Lost My Name received a discounted rate because the companies share an investor. has transformed the way Lost My Name’s approaches Facebook advertising, Thouas says. “We were way too quick to give up on creative that didn’t seem to work,” she says. is more patient with ads that fail to deliver immediate results. Instead of giving up on them, it tests them again a few days later. The method has helped the retailer discover that “bad luck” may cause an otherwise-effective ad to fail to produce results, she says.

Lost My Name is just one of a number of retailers that are finding that artificial intelligence and machine learning enable them to conduct more experiments and quickly examine, and react to, the results of those experiments. However, a number of retailers have also found that, despite the benefits of artificial intelligence, the technology still has its limits.

“AI-based marketing automation has almost unlimited promise,” says Susan Etlinger, industry analyst at Altimeter Group, a business management consultancy. “It can help marketers find and correct for unintended biases. It can help marketers discover that they’re underserving a particular segment.” But, no system is perfect and the systems are only as good as the data they have to examine, she adds.

AI-based marketing automation has almost unlimited promise.
Susan Etlinger
industry analyst at Altimeter Group, a business management consultancy

Take a brand like Cosabella, which has been pleased with artificial intelligence’s benefits, but still sees a role for marketing agencies.

Adgorithms LLC’s artificial intelligence marketing platform Albert has enabled the lingerie brand to make fast decisions, which is particularly useful for a fashion retailer that needs to react swiftly to trends. For instance, it recently found that certain keywords such as “pajamas” and “lingerie” drove particularly strong returns in the middle of the of the night, or that consumers are looking for, but failing to find or buy, certain products in a category such as maternity lingerie.

“Artificial intelligence enables us to better understand how we can drive sales of our brand collection,” says Guido Campello, the retailer’s CEO. “An agency doesn’t enable you to dig as deep as quickly as AI.”

After years of working with a marketing agency, the retailer last October turned to Albert to help it manage its marketing and media buying. The retailer made the change because its sales suddenly slowed after years of double-digit growth. After Cosabella input the key performance indicators it was looking for, the channels it sought to compete in, budget parameters and creative materials, the system drew on the retailer’s customer data to quickly identify user behavior patterns across paid search and social media. Those insights enabled the vendor offer the retailer suggestions for how it optimize and scale its campaigns, he says.


At the same time, the retailer enabled Albert to take what Campello calls “substantiated, pragmatic risks.” “It’s like we have someone on staff who can process data incredibly quickly,” he says. “That enables us to make our ad spend as efficient as possible.”

Albert’s ability to run highly targeted tests enables the retailer to move quickly to ensure it is telling “different stories to different customers at the right time,” he says. It can talk about a maternity lingerie campaign on Monday, develop the creative the following day and have results it can analyze by Friday. That’s a far cry from the agency the retailer used to work with that would take up to five or six months to develop a campaign, Campello says.

The first month that Cosabella began working with Albert, the technology helped it quickly and efficiently scale its ad campaigns. Initially, the Albert system cut back the retailer’s ad spending by 12%, but then quickly restored the spending back to prior levels once it found what worked. Within the first few months, the retailer produced a 336% return on ad spend, and a 155% increase in revenue that it directly attributed to Albert.

However, Cosabella also learned that Albert requires supervision. For example, the retailer found the technology was spending too little, when a higher budget would have produced more profit. “When you’re dealing with an agency, it is more than happy to increase your budget,” Campello says. “But Albert could see that if we wanted to increase our spending we’d be less efficient with our ad dollars. We had to make the call that that was OK.”


Those types of quirks are an inherent part of dealing with artificial intelligence, Etlinger says. “Machines are literal,” she says. “So you need to keep a critical watch on what they’re doing.”

But with supervision artificial intelligence can significantly bolster a retailer’s marketing efforts. For instance, Albert has enabled Cosabella to move its marketing in-house without it having to hire any new talent. Its 10-person marketing department produces its creative in-house. It then feeds Albert images and copy, which the technology can mix and match in ads that it dynamically serves to consumers based on the products and content those shoppers look at on the retailer’s site.

Despite the gains Cosabella has made thanks to artificial intelligence, Campello believes there is a place for specialized agencies which is why he plans to hire three or four agencies throughout this year that have specific specialties, such as maternity or plus-size, to complement Albert’s output.

“While Albert can handle the day-to-day collection and reporting processes, the agencies can add a human dimension, such as content development,” he says. For instance, an agency could help the retailer offer a sports bra tie-in with a gym that’s running a membership campaign aimed at women.


That approach makes sense, Etlinger says. After all, those type of relationship-building marketing efforts can’t be automated.

[email protected]