AI And The Agency: Lingerie Brand Cosabella Replaced Its Agency With Artificial Intelligence

This is the third of three stories in a mini-series on how artificial intelligence is affecting the work that agencies do. Read the previous stories about Xaxis and Publicis.Sapient

After years of double-digit growth, global lingerie brand Cosabella suddenly lost momentum in 2016.

“We decided we needed to cut ties [with our agency] and change something up,” said Courtney Connell, marketing director at Cosabella.

Cosabella had three options: hire another agency, hire more in-house marketers or adopt an artificial intelligence platform that can handle marketing and media buying autonomously. In October, after assessing multiple vendors, Cosabella chose Adgorithms’ AI engine, Albert.

Albert’s machine learning powers marketing in email, mobile, search, social and display. Marketers enter high-level parameters like geos, channels and target audiences and set a budget and KPIs around return on ad spend.

“After that, [Albert] makes every single decision,” Connell said, including identifying targets and keywords, moving budgets between channels, identifying fraud, controlling bids and executing buys.

After three months with Albert, Cosabella saw a 336% increase in return on ad spend. In Q4, revenues increased 155% and the brand saw 1,500 more transactions year over year, 30% of which came from new customers. In Albert’s first month, Cosabella decreased costs by 12% by increasing returns by 50%.

On Facebook, return on ad spend was up 565% within Albert’s first month. By the end of month three, Albert had increased conversions on Facebook by 2,000%.

Cosabella didn’t have to hire any new talent to bring marketing in-house with Albert. Its 10-person marketing department does creative production in-house, feeding Albert images and copy to serve dynamically.

“He can mix and match [creative] however he pleases,” Connell said. “He might start with an ad and if he sees that getting fatigued, he might roll out a new combination.”

Connell checks the Albert dashboard every morning, but because the tool self-optimizes, her team only must check on campaigns once or twice a week. It takes them less than an hour to produce graphics and copy and other materials for a campaign.

“All of the big idea, strategy and campaign creative is happening in-house,” Connell said.

As Albert buys and optimizes media, it makes suggestions. For example, it told Cosabella that creative featuring people performed 50% better that ads featuring just the product.

“The beauty of Albert is we don’t have to optimize campaigns,” she said. “He’ll make suggestions on budget or different microsegments he’s seeing movement on.”

While systems integration can be bumpy for clients, Connell described ramping up on Albert as “painless” and requiring “no technical investment whatsoever.” All a marketer has to do to is link their accounts, like Google AdWords and Facebook, to Albert. It ingests and optimizes ongoing campaigns for about two to three weeks before deploying its own.

“Ever since then it’s been very easy,” Connell said. “We just give him the creative concept to make sure he has enough fresh content.”

Albert only takes a matter of days to weeks to set up because it can work with just pieces of a marketer’s data, said Or Shani, CEO of Adgorithms.

“We really didn’t want to be in the position of the marketing clouds where it takes six months to a year to onboard,” he said. “We want to get started very fast and show value even if we don’t have all the information.”

Eventually, Cosabella will get Albert talking to the other AI vendors it’s onboarded since cutting ties with its agency, including Emarsys for email marketing and Sentient for customer acquisition and real-time merchandizing. Albert can ingest customer lists for lookalike targeting, but Cosabella wants to hook it up the company’s CRM system to keep that information flowing in constantly.

Connell estimates the development work will take up to three hours.

“We want to hook Emarsys up to Albert so he can model high- or low-value customers and adjust his budgets to spend more to attain a certain customer,” she said. “We all want them talking to each other.”

If an email company creates a higher lifetime-value customer, for example, Albert can change its calculations to spend more to target that customer.

“At the end of the day, Albert is just as good as the data you hook it up to,” Shani said. “The more information he can get, the more accurate he can execute.”

Before technologies like Albert, mid-size companies like Cosabella, which has 100 employees, didn’t have many options to leverage AI, Connell said.

“A lot of companies have proprietary technology, but that’s just not reachable for small to mid-size companies,” she said. “Now that’s totally changed.”

Connell doesn’t miss working with an agency at all. Bringing marketing in-house has allowed Cosabella to communicate better and work more efficiently on marketing without the agency as a middleman, she said.

“There’s nothing I miss about advertising agencies,” she said.

When it comes to measurement, attribution and reporting, Albert is more accurate than Cosabella’s agency ever was.

“He knows that he’s shown someone a Facebook ad or if they click on a search ad and make a purchase,” she said. “He gives you reports on assists versus actual sales so you can see how the channels work together to get the customer to make a purchase.”

This isn’t the first time Albert has snagged an agency’s business. A large CPG brand recently fired its agency after piloting Albert for four months, Shani claimed.

“Everybody tells us we’re the agency killer,” he said. “We are a threat to them, there’s no way around that, because were trying to take the pieces of the puzzle they make a lot of money on and that’s a risk.”

Connell doesn’t foresee a time when Cosabella will return to working with an ad agency. There’s just too much data for a human to process and make real-time decisions on.

“I would never have a human do this type of work ever again,” she said. “Albert is looking at very small, subtle patterns 24/7. When everybody else should be sleeping, he’s out there making decisions.”

She leaves marketers with a word of advice.

“If you’re not up to speed, get up to speed and get a tool like this,” she said. “If you want to survive, that’s what you need to do.”

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Kevin R

    I follow automation trends closely, but I missed this one. It doesn’t surprise me thought. People think that automation will kill low wage burger flipping jobs (and it will) but a wide range of white collar careers are also headed for a quick demise. As the article notes, AI systems can process information 24/7 at a level of detail that no human can match.

  2. I’m fascinated by this and how rapidly AI finds its way into all kinds of industries while putting everyone else in that industry out of business. Great article!