When to shout, when to whisper: Making AI-driven marketing really personal

David Reed, director of research and editor-in-chief, DataIQ

According to the philosopher Ludwig Wittgenstein, “the limits of my language mean the limits of my world.” So what if you could extend your language and thereby increase your reach? For digital marketers under pressure to grow their engagement, conversion rates and revenue, anything that might provide greater reach is worthy of checking out.

Persado is hoping to make that possible through what it calls a “marketing language cloud”. Using artificial intelligence and machine learning, its platform Persado One is able to take marketing collateral, such as subject lines, copy and even images, and find the right emotion and level of intensity for a campaign’s audience, segment or individual target using non-personal, anonymised data.

Assaf Baciu, Persado“It’s an AI platform to generate language that resonates best with every individual. That has far-reaching implications because marketers have got audiences across digital channels and, at the last mile, they need to generate language. At the moment, most clients neglect that completely,” co-founder and SVP product and engineering, Assaf Baciu, told DataIQ.

Baciu joined the business from Nuance Communications, which developed the speech recognition and analytics system used by Siri. Persado has just closed $30 millon in investment from Silicon Valley Bank, adding to the $66 million capital it has already raised. What the platform is able to achieve in terms of uplift - across all of its clients’ campaigns, the average uplift in engagement versus control is 49.5% - has been so successful for clients Citi and American Express that they also invested during the second round of fund raising.

“We have proven that AI and ML works,” said Baciu. “We ask them to put in our content and test it against their best version - ours wins every time.” 

Getting to this point has involved a blend of academic theory, mathematics and big data. Alex Vratskides, co-founder and CEO, was previously at Upstream Systems which was typically running high-cost SMS mobile promotions, such as offering a Mercedes SLK every day for 90 days in a promotional draw for Italian subscribers to a mobile phone network. 

“The cost of those promotions is in the millions, including the cars, TV ads and online campaigns and a one-time SMS to prospects,” explained Baciu. With 80% of responses coming from texts, rather than ads, Vratskides wanted to understand the factors which would optimise response rates. A mathematician from Columbia and Cornell Universities, he recognised that the 300 to 600-word messages were a finite set that could be modelled. He hired in Baciu and built a developer and data science team to crack the problem. 

“Deep learning is a well-known concept and so is emotion in language. We have been able to classify over 1 million words and phrases into 15 emotions. Over 60% of response in digital channels comes from the emotional content of language,” said Baciu. Having identified the most important emotions out of an initial range of 100, Persado built the analytics and content engine which marketers can now plug into whichever marketing engine they have in place. 

Using Persado, marketing messages can be tuned in to the most appropriate emotion and also adjusted for the level of intensity required, dialling up the urgency and strength (“Attention please!”) or dialling down for a softer tone of voice (“We thought you’d like to know…”). Even fonts and word spacing can be adjusted using the AI engine.

“99% of the time we generate language that performs better than what our clients were using,” said Baciu. By loading in campaign metrics, users can analyse the exact impact and also generate creative briefs for agencies based on the AI-identified emotional profile of audiences.

Often, the results can be surprising. You might expect Japanese consumers to respond to different triggers from German ones, but Brazilians are triggered by a combination of ambition and anxiety, for example. Insights like this - and the results which Persado claims to deliver - may well justify Baciu’s statement that, “this is the biggest advance in personalisation since people started talking about personalisation.”

Director of research and editor-in-chief, DataIQ
An expert commentator on all things data, David has been editor of DataIQ since its inception in 2011.

Sign-up to hear about the latest DataIQ news, content and events.