AI company Phrasee recently announced funding to expand into the US market. We sit down with CEO Parry Malm to discuss their product, company, upcoming growth, ethics policy, and how they've gotten where they are today.
Yesterday, August 30th, London-headquartered Phrasee announced landing $4m in Series A funding to expand into the US market. Phrasee uses AI to power email subject lines, Facebook and Instagram ads, and push notifications that outperform copy written by humans. They work with clients across the globe, including Virgin, Domino’s, and The Times. Phrasee won 2017’s Most Innovative AI Company by CB Insights, and was one of the first AI companies in the world to implement an AI ethics policy.
We spoke with CEO Parry Malm on Phrasee’s product, company, and how they’re positioning themselves to excel in the US market.
Parry Malm: I started my career sending out emails for a huge media company. I used to always test out subject lines and wonder, “why does one not work?” Then I joined an ESP, heading up the account management part of the business. Every customer would ask me, “what’s a good subject line?” And I’d always say, “well, test a bunch of stuff out.”
It always flummoxed me—why something works and why something doesn’t. I just had this gut feeling that there was a business opportunity there.
I got together with Dr Neil Yager and Victoria Peppiatt, my two co-founders. We founded Phrasee because we believed there was a better way. People have been trusting their gut instincts for far too long, listening to self-proclaimed human experts who didn’t have any real science to back up what they were saying. That’s where Phrasee came from—to allow people to use advanced technology to generate better language, to get better results from their email marketing campaigns.
Phrasee solves a couple unique problems here. Firstly, it can write more subject lines than a thousand humans with a thousand typewriters ever could. Secondly, using our deep learning module, it can predict the efficacy of individual subject lines. So what this effectively gives you is a thousand people on a thousand typewriters in the background, but all you get is a small amount of high-performing copy.
PM: No, we don’t iterate off of what humans write in the first place, and there’s a simple reason for that. What if what the human has written sucks? All we’re doing then is making it less bad. We’re not about making stuff less bad. We’re about making stuff awesome.
On a client-to-client basis, we customize user interfaces where customers specify certain information. Let’s say they have a facebook promotion that’s 50% off shoes. They would enter in that it’s for facebook, the discount is 50% off, the product is shoes, and let’s say it’s a back to school sale.
They would feed Phrasee all of the information. Phrasee would take that information and produce a huge volume of different subject lines. Those subject lines, in the back end, are fed through our predictive model, which is a multi-nodal neural network. It would output a number of different variants for the advertising campaign. They then test these out in third party platforms, such as ESP, an app push system, or the Facebook and Instagram advertising platforms. The winning option surfaces and that is deployed to everyone else. The results from that test is then fed back into the Phrasee system, which then further trains the deep learning algorithms.
PM: Absolutely. We learned this very early on. Businesses are very different in terms of two things: business model and brand voice. We can’t break down brand voice to a couple of adjectives like bold or mild. That’s an unrealistic and inaccurate way of describing it. Brands spend billions of dollars every year building their brand voice on every possible marketing channel online and offline. Why would you throw it away for the sake of a couple more opens?
We’ve built a very sophisticated system in which our team of computational linguists trade language models on a brand-to-brand basis. These language models are 100% tailored to a brand’s business model and voice.
PM: We’ve had customers in the US for quite some time, and they already represent a substantial portion of our revenue. For various operational reasons, we weren’t in a position to actively expand to the US. We’ve been serving these huge customers from our office in London, and servicing them effectively.
But we’ve decided it’s time for us to expand. Our subject line product is world leading. There’s nothing else like it. Our Facebook product and push notification product are already getting results that are beating my own expectations—and I’ve got pretty high expectations.
We officially founded Phrasee Incorporated on June 1, and it’s already become revenue generating. Our belief is that the US market is ready to adopt advanced technology. The time is right to make the technology accessible to many more users.
As far as our ambitions—we finished our third fiscal year with 286% year-on-year growth. I wouldn’t be surprised if we matched that this year or even exceeded.
PM: Yes. So what a lot of companies seem to do is raise $60m in venture capital, and then spend it on stuff—nobody really knows what. It’s almost as if the de facto startup business model is to throw good money after bad and sometimes never make a profit.
Maybe I’m old-fashioned, but I think business is better when its fundamentals are sound. So our first years, we focused on that. As of November last year, we’ve been turning a profit.
We saved money, had cash inflows, and now we’ve invested that in product and expansion into the US. The reason we’ve taken on this new money is not because we’re broke and burning cash on frivolous things. We’ve taken it so we can expand even faster into more exciting things.
PM: There will be a lot of lessons for us, and I really don’t know what those lessons are yet. None of them really faze me. The way we run Phrasee is very similar to the way we sell Phrasee to our clients. We say here are a number of subject lines, which we think are going to work. But test them out. Find the one that works best, then send it out to everybody else.
We do the same thing. We’re going to try out a huge amount of different strategies and tactics. The ones that work, we’ll double down on. The ones that don’t work, we’ll chalk up to experience.
PM: We expect to have 20+ people on the ground by the end of February, which will mark the end of our fourth trading year. As far as revenue goals, though, to be honest, I’m not actually focused on just getting a bunch of revenue in. I don’t think that that’s a good way to launch a business into a new territory.
The key is cutting through the marketing technology clutter and using our voice to let people know that there’s a better way. There’s actually this cool software out there that can help you skyrocket your online advertising rate. We’re focusing on starting the snowball.
PM: And we’re not even on it, so it should be 7001.
PM: Well, first of all we’re going to have really well-worded subject lines and Facebook ads. One thing I’m really proud of about Phrasee is our brand voice. We say the truth even when it’s uncomfortable—and we say it in an engaging and fun way.
PM: We’re super stoked about our AI ethics policy. Our fundamental belief is that we as business owners, and our customers as marketers, bear a moral responsibility to ensure that we are using advanced technology for good. Every week we hear another scandal. Especially in marketing, there are various actors who recommend using advanced technology for spurious means: finding vulnerable groups of people to target with extravagant product offers, or recommending guilt or anxiety based messaging to buy more stuff.
We don’t think that’s right. We shouldn’t be exploiting consumers, and we don’t want our customers to be doing that. For us, it was very important to make a stand. It’s hard to find any AI companies in the world who have any sort of ethical framework. We felt it’s one area we needed to lead in.
As far as how it works cross-culturally, I think it’s actually even more pertinent in the US. There have been many scandals of the use of advertising for political motives. Brands need to take the lead and vet vendors. Vendors need to stand up and say we are doing this, we’re not doing that. You can use spurious tactics and you will get some short-term sales. Maybe that quarter you’ll be happy. But next quarter you won’t be. We don’t want customers for a day. We want customers for a lifetime.
PM: Well thank you. I wrote it myself.
There are so many companies out there who view business as purely transactional. But we know that we’re dealing with people. Not just our customers, but our customers’ customers. As marketers, we have a moral responsibility to ensure that we’re using marketing for good.
One of our points in our ethics policy is explicitly outlining the types of companies or organizations we won’t work with. We explicitly will refuse any business from any political party. We don’t want our technology to be used like that.
PM: I’ve never been one to follow rules. I’ve never been one to believe in the status quo dictating what we should and shouldn’t do with our lives. I think that that fundamentally inhibits creativity and innovation.
I surround myself with people who are better at things than me. My cofounder Victoria is a very process-driven person. If she weren’t here, Phrasee would be absolute anarchy. Neil is much more calm and analytic. If he weren’t here, Phrasee would be quite emotional and passionate.
Another skill I’ve definitely grown into is realizing that Phrasee is no longer just about me, it’s not the Parry Malm show. It’s a team of 48 awesome people, each of whom brings something unique to the table, and each of whom are more skilled than me at specific tasks. And that’s what I’m really proud of.
It’s also shown me that we have an incredible group of customers who go out of their way to help us out. They speak to the media. They speak about us to their friends and colleagues. It’s shown me that all the time we’ve spent building this real, customer-centric culture at Phrasee—that’s one thing we’ve gotten absolutely right, and we’re quite proud of. We love our customers.
When I started Phrasee, we had many people who told us it wasn’t going to work. The idea didn’t make sense. People weren’t going to buy it. Then we got one customer, then another customer, then another. It showed me that just because the experts tell you something, doesn’t mean you have to believe them.
