Phrasee Uses Artificial Intelligence to Write Effective Email Subject Lines for You
A major problem with marketing artificial intelligence today is that marketers don’t first identify the business problem they’re trying to solve with AI before implementing it. Phrasee (@phrasee), a marketing AI tool, has a laser focus on the business problem their artificial intelligence solves: Email marketing optimization.
Phrasee uses machine learning to generate and optimize email subject lines so people open, click, and convert more often. The company works with large companies like Domino’s and Virgin Holidays to beat humans 98% of the time in head-to-head subject line writing tests. We spoke with CEO Parry Malm (@ParryMalm) to learn how Phrasee does it.
In a single sentence or statement, describe Phrasee.
Phrasee is AI that generates and optimizes email subject lines, and gets you more opens, clicks, and conversions.
How does Phrasee use artificial intelligence (i.e. machine learning, natural language generation, natural language processing, deep learning, etc.)?
Phrasee has a few core applications of AI:
1. Natural Language Generation (NLG)
We’ve built a programming language for language. Our team of computational linguists create generative algorithms on a client-to-client basis. This insures that we can generate huge volumes of language at scale, ensure the language is natural sounding and grammatically correct, and align the generated language to a customer’s brand voice and business model.
2. Natural Language Processing (NLP) / Sentiment Analysis
Our proprietary sentiment analysis tool is much more advanced than just a simple “good or bad” model. We identified the key marketing response sentiments (urgency, directness, curiosity, friendliness, and entropy) and can extract the hidden sentiment within unstructured language data at scale. This allows us to quantify the semantic components of an email subject line, and then use that as a decision tool.
3. Machine Learning / Performance Prediction
Once the language is generated and quantified, how can you measure goodness vs. badness? Having a huge volume of language is pointless unless you drive towards language you expect to work, and avoid what you expect to underperform. Our machine learning prediction model is highly sophisticated and can tell you how a given subject line (either generated by Phrasee or written by a human) will perform—before you click the “launch” button.
What do you see as the limitations of artificial intelligence as it exists today?
The biggest limitation is not in technology itself, but in humans’ unrealistic expectations of what AI can do. People think either apocalyptically (“The machines are coming!”) or quixotically (“Just imagine if AI could do this!”), but none of those perceptions solve actual business challenges.
Therefore, the lack of understanding of AI is the main limitation. For example: “AI” itself doesn’t really mean anything, but it’s a catchy buzzword. It’s a catch-all phrase for numerous application areas such as NLG, NLP, machine vision, machine learning, and so on.
What do you see as the future potential of artificial intelligence in marketing and sales?
In a perfect world, you could type into a chatbot, “I want an ad for milk, with a superhero theme.” The AI will spit out an awesome looking advert of Superman with a milk moustache, and a catchy tagline.
We are quite a ways from this. But this, to me, is where we’ll get to.
The media buying business model has already been massively disrupted through the programmatic buying revolution. Advertising and creative agencies are next.
What makes Phrasee different than competing or traditional solutions?
Phrasee focuses on the sweet spot of effective language: subject lines that perform, and are also on-brand.
See, it’s easy to clickbait people into opening an email. For example: “Free Beer!” Who’s not going to open that? But if you’re not actually offering free beer, then you’ve just created a negative brand experience.
So that’s why our generative algorithms are trained to avoid being misleading, and are tailored to a brand’s voice.
Coupling this brand specificity with our market-leading performance prediction model is why we beat humans in 98% of head-to-head tests. But, hey, like Tom Petty said: even the losers get lucky some time!
Who are your prototype customers in terms of company size and industries?
We exclusively work with enterprise customers like Domino’s, Virgin Holidays, and Charming Charlie. Our key sectors include retail, travel and holidays, telecoms, and financial services.
What are the primary use cases of Phrasee for marketers and sales professionals?
The use case is quite simple. Do you want more opens and clicks from your revenue-driving email marketing campaigns? If so, give us a call.
Any other thoughts on AI in marketing, or advice for marketers who are just starting to explore the possibilities of AI?
Often, a CEO will walk into a meeting and proclaim, “We need to invest in AI”, and throw a bunch of money at pie-in-the-sky concepts that aren’t realistic or may never actually make them any money!
Here’s what I’d recommend to avoid this situation: spend 95% of your effort on identifying the business problem you’re trying to solve, and then 5% of your time identifying the solution. AI might be a solution for your problem. But it might not be!
This is why Phrasee has been so successful: we identified the problem. Brands are using ineffective subject lines in their email marketing campaigns. Then, out of numerous options, we decided that an AI approach would be the most effective.
About Paul Roetzer
Paul Roetzer (@paulroetzer) is founder and CEO of PR 20/20, author of The Marketing Performance Blueprint and The Marketing Agency Blueprint, and creator of The Marketing Artificial Intelligence Institute and Marketing Score. Full bio.