Caught in the middle: Rev.com

Note: This article is long, at least by blog standards (2000+ words, or about 5 pages). The first Power of Three article to closely examine a specific problem at a middleman business, it required more reporting and analysis than most articles you’ll see here. Thank you to Ea Macom for editing it.

....................................

When news recently spread that transcription service Rev.com had lowered pay rates for some of the work done through its platform, I was doubly dismayed.

First, I have been a user of the service myself, and, like many of my writer colleagues, I have come to rely on it for fast and accurate transcriptions of phone interviews. It’s been a boon to my productivity to outsource the time-consuming work of transforming an audio recording into an easy-to-read text file—a relief to be able to click a button that simultaneously submits a recording and charges my credit card, and to know that, within 24 hours, a clean transcript will arrive in my inbox. Talk about smoothing out trading frictions—the company has been a perfect example of a middleman reducing transaction costs. At least it has been for me.

Second, my interest in marketplaces (and middleman businesses more generally) has given me a special appreciation and admiration for the work Rev does. It matches users’ recordings with available transcriptionists (or “Revvers,” as the company calls them); to date, whenever I have submitted an audio file at least one Revver has been available to provide me with a transcript within the aforementioned 24 hours. In addition, while it offers its workers a chance to work from home and to set their own hours, Rev keeps the Revvers accountable by letting me and other customers rate the quality of each transcript we get back. (As I have recently learned, the company also rates Revvers on metrics such as on-time submission.)

The tough reality for Rev and other middleman businesses is that they are often caught in the middle between the demands of their customers and the needs of their labor force.

No longer the ideal Partner?

This combination of being a Bridge, a Risk Bearer, and an Enforcer is exactly the kind of thing middlemen do to create value for both sides. Rev’s price to me is lower than the cost of doing my own transcription, so all in all I have considered Rev a valuable Partner in my work life. I have always understood that Revvers working on my files receive only a portion of the fee I pay, and I have been fine with that, since Rev has real costs and also aims to make a profit.

But news that workers would now be earning only $0.30 per recorded minute for some jobs has made me reconsider: had Rev, while a Partner to me, become a Predator to its labor force? Would I now have to choose between convenience and conscience, in much the way that so many customers have felt obliged to do with Uber, Instacart, DoorDash, and other middleman businesses that appear to be squeezing out profits at the expense of their contingent labor force? 

I decided to look into these questions. If nothing else, I knew that the situation at Rev would make for an interesting case study of a middleman business caught in the middle. The rest of this article reports on what I found out and suggests what can be learned from it by people creating or using marketplaces.

All jobs are equal but some are more equal than others

Rev pays its transcribers by the number of minutes in an audio file, not by the number of minutes required to transcribe that file. Regardless of a Revver’s skill level, two audio files of the same length can require significantly different amounts of work. That’s because some recordings are more difficult than others to transcribe.

This fact is the root of Rev’s problem, and it’s something Rev apparently didn’t anticipate when the company created its marketplace. As a Rev spokesperson put it in an email to me, “While striving to make Rev a better place to work, we discovered—and heard from our community—that complex transcription jobs require increased expertise, time and effort to complete.” Or, as one former Revver wrote on a review site, “For around $.50/minute captioned, it may sound doable, but when you consider the abundance of poor audio quality, esoteric subject matters, or the number of different people speaking during a video, that $.50/min is not compensation enough.”

The spokesperson explained to me that the company’s goal “is that our community is paid fairly per hour of their work—not per the length of the audio file.” Therefore, the new rate structure will “pay more for harder files and less for easier files.” 

How can the company know how hard a file is before a Revver transcribes it? Rev runs the file through its “speech recognition engine”—speech-to-text software that generates a preliminary transcript and assigns the file “an audio confidence score” that predicts how much time and effort it will take for a human to transcribe the file. For example, how clean is the audio quality, and how much crosstalk is there between speakers? 

Seeing this confidence score, along with details like the length of the file, enables Revvers to decide which jobs they’d like to claim, the spokesperson said.

More than a matter of fairness

Rev presents its recent rate change as a matter of fairness—and stresses that Rev does not profit in any way from the change—which sounds like an admirable, communitarian goal. And it does sound fair to pay more for harder files. (Whether they are achieving fairness is another story—when that higher pay comes out of the pay for easier jobs, the change will seem like a loss to at least some Revvers, especially if they didn’t think that even the easy jobs paid all that well.)

But fairness isn’t just a moral concern—it has practical business consequences. In general, wages that don’t seem fair create ethical qualms for employers, workers, and customers. (The New York Times, for example, has already said it is reviewing its use of Rev in light of these concerns.) As for workers, unfair wages don’t just harm morale—they actually change how workers behave in ways that can harm their employer and the marketplace in which they’re working. Most obviously, poor or unfair wages cause the most skilled workers to abandon the platform in search of greener pastures. Less obviously, workers who stay with the platform may be more likely to take the easiest jobs because those jobs yield the highest hourly rate. As one Revver wrote on a review site, “The good work is snatched up quickly, leaving the poor-quality audios for those who aren’t available to work all the time.”

That’s what we should expect to happen when pay rates don’t adequately take difficulty into account. The undesirable result of that kind of cherry-picking by workers is that in a marketplace like Rev, harder jobs tend to languish in the job pool, potentially exceeding the 24-hour promise to customers. 

Before this change in its pay structure, Rev had already been paying more for harder jobs. But the company wasn’t paying enough of a premium. For example, if a harder job took twice as long to transcribe as an easier job, the harder job didn’t necessarily pay twice as much. So to maximize their hourly rate, a rational Revver would take an easier job if given the choice. And if there was no choice because there were no easy jobs to be claimed at that time, a Revver eager for work might go ahead and take the difficult job—but might shirk a little (rushing through that job somewhat), reducing the quality of the transcript. At least that’s what economic theory would predict. And my guess is that that’s what’s actually been happening at Rev; through the writer grapevine, I’ve heard rumblings that Rev transcript quality has slipped lately.

The all-you-can-eat problem 

Rev indeed had serious business reasons for changing its wage structure. Some changes were necessary for the marketplace to thrive and for the company to make long-term profits for its founders and funders. But were these changes the right ones?

A significant cause of Rev’s challenges is the fact that it charges the same rate to customers regardless of the difficulty of the jobs. The price is directly proportional to the length of the recording—$1 per recorded minute of audio, no matter how poor the audio quality, how many people are speaking, how thick their accents, or how much they’re speaking over each other. It’s also the same price whether the conversation uses highly technical jargon or common English. 

Charging the same price to everyone can be a recipe for trouble. Think of an all-you-can-eat buffet, where everyone pays the same price regardless of how much they actually consume: the flat price is a great deal for some people and a bad deal for others. In effect, modest eaters subsidize the diners who are as voracious as Homer Simpson at that seafood buffet. You might think things come out even on average—but that’s not what happens. That’s because diners don’t show up randomly, but rather self-select depending on how good a deal they think they’re getting. So big eaters will be disproportionately drawn to the buffet. (In the jargon of economics, such a situation is an example of adverse selection, also known as the lemons problem.)

And there’s every reason to believe something like this was happening for Rev’s customers: if you have a difficult file, you’re more likely to think $1/minute is a bargain than if you have one that’s straightforward. For example, if your file is really easy, you’re more likely to take it to one of the wholly automated transcription services whose price is somewhere between $0.10 per recorded minute and free. (Temi.com, owned by Rev and presumably running the same speech recognition engine Rev uses to classify recordings by difficulty, charges $0.10 per recorded minute.) Speech recognition software makes lots of mistakes with difficult recordings, but does a good-enough job with easy ones—at least good enough considering their super-low price. 

I believe that adverse selection has been hard at work in the Rev marketplace, leaving lots of “lemons” in the form of recordings that Revvers have found unprofitable to transcribe. Indeed, the easy jobs seem to have become a minority of the jobs submitted to Rev’s human transcription service. The Rev spokesperson told me that the new rate structure that raises the wage rates for difficult jobs will lower the compensation for only “a small number of jobs (we anticipate about 5%),” suggesting that the majority of jobs are not easy.

Why not raise prices to pay workers more? 

Reading all this, you might think that Rev could have avoided the need to lower any of its wage rates, even on the 5 percent of jobs that they’re anticipating, by raising prices to its customers. Perhaps they could have gone to a system of charging more for difficult files, such as recordings about technical topics. When I brought up this possibility, the Rev spokesperson did say, “We are exploring ways in which we can charge customers higher prices and therefore pay our freelancers more.”

But that is not a simple solution by any means. The main reason, of course, is competition from rival services, especially the automated transcription services I mentioned earlier. While I may be willing to pay ten times more for a human transcription than I would for a computer-generated one because of the difference in quality, would I pay 15 times more? Or perhaps I would be better off looking for a reliable human transcriptionist on my own, maybe even a former Revver, cutting out the middleman altogether. I know that many of my colleagues have already chosen one of these alternatives to Rev.

The tough reality for Rev and other middleman businesses is that they are often caught in the middle between the demands of their customers and the needs of their labor force. As I mentioned in my previous article, it is certainly possible to serve two masters—but it’s not always easy. When a middleman business succeeds, it can almost look like it’s minting money, getting something for nothing. But in truth there’s a lot of work and know-how in running a successful middleman business: as the Rev situation shows, keeping both sides happy is a difficult balancing act, and a misstep can create problems for the whole system.

So what is to be done?

This isn’t to say Rev is stuck. Its current solution, to raise some rates by lowering others, has clearly proven unpopular with Revvers and customers alike. (It sounds like robbing Peter to pay Paul.) But other possible solutions are worth trying. For one thing, Rev could limit the kinds of files it accepts (for example, telling customers up front that a phone interview between a journalist and a source is fine but a conference call with 10 people speaking is not). It could run a submitted file through its speech recognition engine and offer the customer a price dependent on the file’s projected complexity.

It could also embrace the win-some/lose-some attitude of the middleman role I call the Risk Bearer, reasoning that it is in a better position to absorb the vagaries of different jobs than are its Revvers. In other words, it could decide that while not every job has to be profitable to the company, every Revver needs to be satisfied with their pay. As my friend and colleague Ea Macom pointed out, if Rev wants to pay fairly, one solution is to “pay by the hour of labor, trust your worker to claim the correct number of hours, and then weed out workers who are, by some algorithm, too slow.”

Whatever it does, Rev must become more mindful of its reputation than it has been. A middleman business lives and dies by its reputation with both sides. It’s not enough to be known for treating customers well—the marketplace must do right by its workers, too, even by those who may seem easy to replace. We live in an age when workers talk with the whole world, broadcasting their experiences through review sites, Reddit discussions, and other online forums. What’s more, these reviews are seen not only by prospective workers, but by prospective customers as well. And many of those customers, including me, prefer to work with businesses that are true Partners to both sides. I’m rooting for Rev to be such a business. 

Where to read more

You can follow me here or subscribe to my newsletter if you don’t want to miss a future article. I’m also happy to hear from readers. Do you have questions I might be able to answer in a future article?

About the author

Marina Krakovsky, the voice behind Power of Three, is the author of The Middleman Economy: How Brokers, Agents, Dealers, and Everyday Matchmakers Create Value and Profit (Palgrave Macmillan). She is also co-author, with economist Kay-Yut Chen, of Secrets of the Moneylab: How Behavioral Economics Can Improve Your Business (Portfolio/Penguin). In her writing, speaking, and consulting, her main focus is on the practical application of ideas from psychology and economics. Her articles and essays have appeared in Discover, FastCompany, the New York Times Magazine, Scientific American, O (The Oprah Magazine), Psychology Today, Slate, the Washington Post, Wired, and more.