OfferFit gets $25M to kill A/B testing for marketing with ML

“A/B testing is dead” proclaims the copy on the homepage of OfferFit, a three-year-old, Boston, Mass.-based startup founded and led by George Khachatryan as CEO, a PhD mathematician and former cofounder of education software startup Reasoning Mind.

It’s a bold proclamation, but one the company is confident it can back up for brands seeking to optimize and personalize their digital marketing efforts more easily and with far better results than prior methods. (“A/B testing” refers to the practice of sending half of recipients one type of communication and the other half a different one and seeing which message performs better in terms of metrics such as open rates, click throughs, activations, sign-ups, purchases, subscriptions, etc.).

And investors seem to agree: today the company announced a $25 million series B funding round led by Menlo Ventures, joined by Ridge Ventures and earlier investors Canvas Ventures, Harmony Partners, Alumni Ventures Group, Carbide Ventures, and Burst Capital.

In addition, Capital One Ventures, the VC arm of the recognizable and popular credit card and banking merchant, committed an investment following its success using OfferFit to automate sending personalized mass marketing messages about its financial services products to customers.

What OfferFit offers

Key to OfferFit’s success at winning backers and customers-turned-backers is its approach to digital marketing: it uses machine learning, specifically reinforcement learning, in which algorithms are trained to take actions that result in either “penalties” or “rewards,” essentially gamifying the learning process and relying on trial-and-error, similar to how human babies learn.

Reinforcement learning forms the backbone of OfferFit’s automated marketing solution, which ingests data about its clients’ customers and marketing efforts-to-date, and automatically figures out the optimal messages to send at the optimal times on the optimal channels to each and every single customer — even if the userbase is in the millions, as is the case with large enterprises such as Capital One.

“The beauty of this is it’s not a one time thing,” said Jean-Paul (JP) Sanday, a partner at Menlo Ventures, in a video conference interview with VentureBeat. “You can test you and don’t have to declare a winner. It just always optimizes and it stays on — the lift actually improves over time.”

And even if and when end user behaviors change — as they often do throughout our lives, as we grow and enter different levels of school, the workforce, get married, have children — OfferFit can deliver the right messages for the end-user’s stage of life.

“If your user patterns and behavior changes, it picks up on that and starts saying, ‘this is a new emerging behavior,’” Sanday explained. “When a new channel shows up, or somebody starts spending more time in a different app, it will detect that and change the marketing to accommodate it.”

OfferFit’s ML solution is also flexible enough to work across different key performance indicators (KPIs) without retooling. Whether the customer is seeking to drive open rates, engagement, click throughs, or nearly any other conceivable, measurable result, the platform can optimize its messaging times and channels to achieve the customer’s goals.

“What frequency with which you send messages, what day what time of day, it all gets kind of figured out by the system and so you just apply experimentation at scale,” noted Sanday.

The ‘Holy Grail’ of automated personalized marketing at scale?

Sanday admitted he was hesitant at first to invest in OfferFit because it seemed too good to be true.

“When I saw this, initially I said, this is like the Holy Grail again.. I don’t know, I’ve been pitched the ‘Holy Grail’ so many times,” he told VentureBeat.

But Khachatryan and his co-founders’ deep mathematics background, along with the opportunity presented by a more mature ecosystem of connected messaging applications and toolsets, won him over to the central conceit of the platform and the innovation it facilitates: a one-stop shop of algorithms for optimizing and personalizing marketing across sectors, channels, audiences segments, and timespans.

“The model is going to go out and based on actual [end-user] behaviors, start understanding,” Sanday explained. “It will give you [customer] a series of things to put in front of users like subject lines, creative offers or incentives of all different types. And it won’t hallucinate or give them 90% off or anything, it will operate within the constraints that the customer sets up.”

Specifically, OfferFit claims to have achieved such striking results as a 120% increase in average revenue per user (ARPU) at Liberty Latin America, a telecom company, resulting in an addition $1 million in annual value. For Brinks home security, OfferFit says it achieved a 450% growth in value by driving contract extensions from existing customers, equivalent to $5 million annual benefit.

The company services customers across sectors in retail and ecommerce, travel and hospitality technology, media and entertainment, telecommunications and utilities, financial services and insurance, as well as healthcare and wellness.

Moreover, Sanday was careful to note that OfferFit did not aggregate end-user data across its customers, nor did it co-mingle data from its various customers into a pile. However valuable that might seem — creating cross-company customer profiles — OfferFit seeks to maintain the privacy and data security of both its customers and end users.

Sanday said this was also not necessary for the platform to optimize its suggested messaging.

“The way you manifest to your utility provider, for example, does not necessarily always tell me what’s the right thing to do for your credit card offer,” he noted.

What’s next for OfferFit with its new cash

Now that the company has demonstrated its value to large notable customers and secured additional investment, it plans to “continue investing in our product.”

According to its webpage announcing the funding round, that means it will build out additional integrations to marketing software platforms, allowing OfferFIt’s ML smarts to leverage existing workflows and software tools to push out the best messages at the right times for its customers (and most importantly, their end users).

In addition, the company plans to expand “our self-serve and content generation capabilities.” According to Sanday, this may ultimately include a generative AI component of actually generating raw marketing copy and visual assets, though he stressed these would of course be subject to approval of a human marketing manager or equivalent for every customer before being pushed out to end users.