Laura Bliss is CityLab’s west coast bureau chief. She also authors MapLab, a biweekly newsletter about maps (subscribe here). Her work has appeared in the New York Times, The Atlantic, Los Angeles magazine, and beyond.
If there’s a more effective way to say “stand clear of the closing doors,” this start-up wants to help transit agencies find it.
You’ve been here before: crumpled against hundreds of bodies in a subway car at rush hour, withering in stale air. The train slows down and jerks to a stop. Oh no, you think. Then the dreaded announcement flips on.
“Ladies and gentlemen,” crackles a jaunty male voice, “we are delayed because of train traffic ahead of us. We apologize for any inconvenience. Thank you for your patience.”
You bristle. If they really wanted to assuage your frustration, they’d get newer trains on tracks, speed up repairs, and keep expanding service. The metro’s apologetic plea for patience does little to soothe you.
But what if the message were a little brighter, a little peppier, a little more direct? It wouldn’t speed up the train, but one tech-ified marketing company thinks you might be a happier rider if you heard, instead: “Thank you for your patience, ladies and gentlemen! We are delayed because of train traffic, but we should be moving shortly.” It doesn’t sound much different, but millions of marketing phrases, processed by the power of artificial intelligence, suggest that it really might inspire a dash more patience.
The company is Persado, a start-up that uses AI to “optimize” marketing messages so that customer behavior yields greater returns to their clients, which include the likes of Verizon, Microsoft, and American Express. They’re trying to expand from corporate marketing to other realms of nudge-induced decision making. In the example above, which tweaks a classic message from the New York City MTA, Persado’s representatives explain via email that its computers put politeness at the front of the message (“thank you”). And, because of the frustrating context of the message, they added in a sprinkle of positivity at the end to convey a greater sense of assurance.
How does it work, exactly? Persado’s computers take in a pre-made company message—for example, one you might see in an email blast, street signage, or in the store—and run it against a database that spans roughly 6,500 marketing campaigns, representing roughly 45 million unique phrases used in those ads, with data on how effective they were at prodding customers to buy or act. The AI spits out a comparable message that packs a quantifiably bigger punch, as defined by its formatting, its syntax, its descriptive and emotional language, and its “call to action.” (Persado tests its AI-backed wording against original copy-written language, and says that, in the context of email campaigns, their messages improves click-through rates by an average of 68 percent.)
Of course, the train may very well not be moving shortly. And it won’t fix the structural problems plaguing transit networks. But Persado believes smarter wording of onboard messages might inject some much-needed amiability and calm into the rider experience. The company ran a few more New York City MTA announcements through its machinery and came up with these alternatives, provided to CityLab via email:
MTA: Stand clear of the closing doors, please.
Updated: Please be careful of the closing doors.
Here, Persado explains in the email that “front-loading” politeness (i.e., “please”) is a nicer way to frame commands. “Stand clear” is apparently a “technically worded verb phrase“ that lacks emotional resonance—“be careful,” they say, is a simpler, clearer way of stating what’s immediately important. It connects the requested action to one’s “immediate sense of well-being.”
MTA: For your safety, please do not block or hold the car doors while the train is in the station. And please, do not lean against the doors.
Updated: Reminder: please do not block, hold, or lean against the car doors. Your safety is important.
Persado explains that this announcement contains an emotional appeal already, so updating it was just a matter of whittling away excess words. Apparently, the word “reminder” is better at grabbing attention when you want to talk about “safety”—which is an effective emotional conjures.
MTA: Ladies and gentlemen: we are delayed because of train traffic ahead of us. We apologize for any inconvenience. Thank you for your patience.
Updated: Thank you for your patience, ladies and gentlemen! We are delayed because of train traffic, but we should be moving shortly.
Once again, Persado has put politeness at the front of the message (“thank you”), and “added a positive element at the end to convey a greater sense of assurance.” Of course, the train may very well not be moving shortly… but for the time being, it might help to have people believing it will.
MTA: A train is now arriving. Please stand away from the platform edge.
Updated: Attention: Please stand away from the platform edge. Your train is now arriving.
Here, Persado restructured the phrase to be more persuasive, according to its data: the intro commands more attention, it’s immediately followed by a call to action, and then the important information follows. And notice the strategically placed “your”: a direct appeal that adds a little sense of personal ownership is good for inspiring action.
MTA: If you see an elderly, pregnant, or handicapped person near you, offer your seat. You'll be standing up for what's right. Courtesy is contagious and it starts with you.
Updated: Dear passengers, look around you: Do you see an elderly, pregnant, or handicapped person near you? Please don’t ignore them—offer your seat. Your courtesy is appreciated!
The new phrasing has an intimate salutation, and it asks for a specific, direct action that’s worded as a question, which apparently “works well as a means to encouraging courteous behavior.” It then conveys a bit of guilt with “please don’t ignore them.” Then it deletes “courtesy is contagious”—too conceptual and logic-intensive, according to Persado, and “contagiousness” is a poor metaphor under tightly packed conditions—and replaces it with “a direct and simple expression of gratitude.”
MTA: This is an important message from the New York City Police Department. Keep your belongings in sight at all times. Protect yourself. If you see a suspicious package or activity on the platform or train, do not keep it to yourself. Tell a police officer or an MTA employee. Remain alert and have a safe day.
Updated: Important message from the New York City Policy Department. Keep your belongings in sight at all times. If you see a suspicious package or activity on the platform or train, tell a police officer or an MTA employee. Remain alert and have a safe day.
Almost no changes here: According to Persado, the only thing worth changing here is scrapping a few unnecessary words. Otherwise, its computers read this as an effective example of direct, emotionally appealing instruction.
Yes, copywriters, these robots are coming for your job. Persado more or less relies on the authority of its computers; their engineers don’t get too far into the behavioral psychology of its methods. But these language changes are, more or less, behavioral economics distilled into software; it’s the psychologist Daniel Kahneman’s famous theory of two-tiered decision making: We rely much more on our emotions than we do reason to process and respond to surroundings and inputs.
Amid declining ridership numbers nationwide, and ever-shrinking budgets, transit agencies might just want to do everything in their limited power to improve the passenger experience; more thoughtful messaging could go a long way.