Marpipe is bringing A/B testing on steroids to marketers

Automation is the boogeyman of the professional and working class, and the advertising industry—for all its creative pomp and circumstance—ain’t avoiding it. We’ve already covered how artificial intelligence is changing the game for copywriting.

No, robots aren’t writing Super Bowl scripts (yet). But automation is taking over parts of the creative process.

Take Marpipe, a platform that just raised $8 million from investors. It’s aiming to replace the “most annoying process” of digital advertising, Dan Pantelo, founder and CEO of Marpipe, told Marketing Brew: creating and testing ads.

Here’s how it works: Marketers upload their campaign assets into Marpipe’s platform (think: images, background colors, and copy), and it spits out tons of different variations and combinations.

Unlike copywriting software, creatives still have to input something, but Marpipe throws it all together.

So, if you had a pet store...your assets might include images of a chihuahua, a kitten, a rabbit, a tennis ball, and some chow over five different background images and colors.

• The ad copy might also plug a deal you’d like to offer (free shipping, for instance) or tout sustainable chew toys.

Then—because it’s plugged into Facebook’s advertising network—a campaign across a selected audience (Marpipe suggests a testing pool of at least 250,000 people) runs with tons of different assets. Performance is then measured by either purchase conversions, leads, traffic, or reach (or several at once).

From there, Marpipe ranks the best-performing creative, giving results within days.

“Most creative doesn’t work well; the process of guessing and checking is a very, very inefficient process...No one wants to just drag and drop stuff all day,” Pantelo told Marketing Brew. “The most granular visual variables have such a profound impact on performance. That brings up the fundamental question: How do we determine what our ad creative should look like and what it should say?”

Most companies allocate between 10% and 20% of their ad spend on testing, according to Pantelo, and it’s this chunk that can be made more efficient. Gone are the days of A/B testing two ads against each other—why not test 150 of them and know what works at a more granular level?

Back to the pet store example: You might find that chihuahuas always outperform rabbits, yellow backgrounds always outperform white, and free shipping is a better pitch than sustainable products.

The little things

Though only integrated into Facebook and Instagram, Marpipe will soon have access to run ads on Google’s Display Network and another programmatic advertising platform (Pantelo declined to name the company).

• Marpipe is looking to work with brands that spend at least $50,000 a month on advertising. It currently charges marketers $2,000 a month, opting for a flat fee over a percentage of ad spend.
• Average users test between 50 and 150 pieces of creative, about every four days.
• Clients so far include Quip and Guitar Center.

As the industry braces for a loss of personalized data since the third-party cookie is going away, creativity is becoming more important for marketers than audience targeting capabilities, Andrew Silberstein, CEO of DTC skincare brand SolaWave, told Marketing Brew.

SolaWave has been using Marpipe for the last four months and has tested as many as 192 pieces of advertising for a digital campaign. It usually lets campaigns run for as long as four weeks before checking to see which pieces are working best.

“The smallest, least obvious design changes can dramatically change results,” Silberstein said, explaining how things like letter capitalization and small design elements can drive surprisingly different results

An example: SolaWave recently ran eight advertisements; half of them featured a wavy line down the middle, while the other half showed a straight line (and a wave in the bottom left corner). The ad with the straight line performed 3.5 times better in terms of getting customers to add products to their shopping carts.

“It’s not about replacing human beings—that’s part of it—but the software helps plan more comprehensive testing...it pushes us into testings we wouldn’t think to test,” he said.

Zoom out: Arguably, if SolaWave only tested two ads, it wouldn’t have to run 190 others. Aren’t more ads just more opportunities for duds?

But Pantelo likened Marpipe’s testing to the stock market: A diverse portfolio provides more options than going all in on two companies. “By running more, you maximize your possibilities of finding winners that are winning more,” Pantelo said. “If you only have two darts, you’re probably not going to hit a bullseye, but if you have 50, you’re probably going to hit it.”