Like many who spend a lot of time reading on the internet, I love twitter. It’s an invaluable source of information. One especially prized by journalists and infovores. But the product has stagnated. In particular casual users have struggled with it. One billion people have tried it (!) but only about a quarter of those stayed with the product. So it was no surprise when twitter announced in early June current CEO Dick Costolo would step down.
Since journalists and tech analysts love to write, many chimed in with Twitter product suggestions. My own take from 2013 for getting casual users on board was “Instead of curating a list of people to follow, think stations. You first pick from a couple of categories such as News, Tech, Celebrities to show what you’re interested in. Then Pandora style you like and dislike tweets until you are happy. It would work for the vast majority of human beings who will never curate their lists, and frankly find competing with stars on twitter far too depressing.” This is a popular notion, and I’ve seen plenty of variations on the idea. In particular I liked Ben Thompson’s post from last year Twitter’s Marketing Problem: “Instead of trying to teach new users how to built a curated follower list, build the lists for them. Don’t call them lists, though; embrace Twitter’s TV connection and make them ‘channels.’ ” So here we are in 2015, and Twitter just announced they’ll ship a version of this feature “later this year” as Project Lightening.
By far the most high profile critic (and simultaneously most fervent booster) of Twitter has been one of their largest stockholders, Venture Capitalist Chris Sacca. Prior to Costolo’s resignation he wrote an 8500 word post on what Twitter should do, followed by another 1800 words after the announcement. His lengthy posts wandered a bit, but what jumped out at me was this quote:
Logged in users are always nice to have, but Twitter’s logged out visitors will continue to be increasingly valuable. Between search and referrer intent, geographic and time context, TellApart data, cookie data, data gathered from developer tools, and inferences from engaged participation in the stream, there is no ceiling on the potential to make money from this logged out audience.
“Logged in users” are users who have a twitter account. Twitter knows who logged in users are since, well, Twitter has their email address and they have to log in each time they use the product. So any action they take on Twitter ties back to their identity. In turn that identity ties to what interests them, since Twitter knows who they follow and what they click on. In contrast “Logged out users” are people using Twitter who never created an account. They’re anonymous. Though as Sacca notes above they can still be tracked to a limited extent.
Given Twitter’s shortcomings, many heavy Twitter users, including me, also use Nuzzle. Nuzzle has a brilliant simplicity, as great products often do. It aggregates links from your twitter feed. So even if you’ve been away you can find out what people were linking to while you were gone. This avoids tediously paging back through your tweet history to see if you’ve missed out. Nuzzle’s CEO is of course Jonathan Abrams, most famous for founding Friendster. In a recent (and excellent) interview Business Insider described Abrams as the “CEO of the company everyone thinks can fix everything that’s wrong with Twitter.” And in discussing upcoming versions of Nuzzle, Business Insider asked a question I was really interested in: “How will you make the not-logged in experience work? You must require some kind of log in surely?” Abrams responded:
We do, but if you install a news app, most of them don’t require you to log in, they just show you a bunch of news, they just pull up politics of business, sport whatever. So we’re going to have a new version of our app where if you don’t want to log in, you’ll be able to browse and look through different feeds, and then if you find a feed that’s relevant to you, whether it’s the Hilary Clinton news or the Bitcoin news, Google news, you’ll be able to just save it and then that will be then in the app a tab or something you can come back to the next day that feed will be right there.
That’s the way news works today. Sure. But as a big Nuzzle fan, and someone who hopes someone of Abrams’ stature winds up at Twitter, this response was a real bummer. Similar to Chris Sacca’s quote above. Of course that’s not an obvious reaction. So let me explain.
Consider Benedict Evans’ tweet below. He’s complaining on Twitter about Twitter, as one is wont to do. His complaint is Twitter knows his precise interests by who he follows, but then provides idiotically inappropriate recommendations on what to read.
Clay Thompson is meaningless gibberish? Yes. At least to some.
This complaint brings my concern into focus. In my post 2015 is a transition year to the (somewhat creepy) machine learning era, I argued our baseline expectations for computers are radically changing. We expect computers to be personalized digital assistants, anticipating what we want. This is true of traffic on my commute home, calendar reminders, choosing restaurants, voice interaction (Apple Siri and Google Now), health suggestions (Apple Watch says stand up now) and much more. It’s becoming true for ads as well. Providing crappy ads with no context is not going to cut it in the machine learning era. Obviously someone like Benedict Evans, doing tech analysis at a VC firm, has expectations above the norm. But that norm is where we’re all headed. It’s not a stretch to believe irrelevant ads will soon be viewed by mainstream people as hallmarks of a crappy product.
How ads work in this new era is companies like Google, Apple, Facebook, and yes, Twitter, have a lot of data about you. This is somewhat creepy. At least today. This data feeds into machine learning algorithms to organize your photos, customize voice interaction, provide reminders on commute times, and make recommendations on what you might want to read. There’s a common misunderstanding these platform companies sell your private information. They don’t. Instead, just like print magazines back in the day, they sell placement of ads in front of likely buyers. So they sell placement of car ads in front of people who want to buy cars, but don’t provide the car companies themselves with your information. Car companies just pay to get an ad in front of people without knowing exactly who those people are. Even Apple does this with their iAds platform (albeit indirectly since Apple’s primary business model is selling devices). There’s a caveat to this with ad networks, but for this discussion we’re focusing on Twitter and similar internet platform providers. In fact Facebook and Google made real identity a requirement a while ago. Identity is the one of the few areas where Twitter would do well to blindly copy Facebook. This becomes all the more critical as we enter the machine learning era.
Of course Twitter is full of very smart people. Undoubtedly they’ve discussed the pros and cons anonymous users many times. And their recent acquisition of Whetlab shows they understand the future is machine learning. But the elephant is the room is the 1 billion people who’ve already tried twitter. How does Twitter accomplish the nearly impossible task of getting them to try again? Anonymous users might help. Tactically. But what differentiates Twitter from Facebook and other social networks is Twitter is foundationally built on interests. That’s what makes Twitter Twitter. It’s why Twitter’s so great. People from around the globe come together to talk about shared interests and concerns. However the casual user problem gets solved, identity will be essential to creating a premium, and differentiating, experience tailored precisely to people’s interests. Yes, it’s true some newspapers toss out anonymous ads on untargeted stories to anonymous users. But even there the percentage of revenue coming from ads versus subscription is in decline, as pointed out on, well, Twitter. And news articles themselves are getting folded into platforms like Facebook, because, among other reasons, those platforms have advantages in targeting ads to real identities. Chasing anonymous means chasing the past. Anonymous users are Twitter’s great, and falsely alluring, temptation.
To wrap up I want to play around a bit by rewriting a paragraph taken from one of my favorite science essays “The False Allure of Group Selection,” by the great science writer Steven Pinker. Which of course is where I got the title for this post. Pinker’s essay argues group selection has a false allure and explanatory power in genetics, and should be avoided as obsolete dogma. In my version below I’ve swapped out phrases such as “group selection” for “non-logged in business strategy” as needed. The original portions are in italics, while my swapped-in words are in straight text. The unaltered version of Pinker’s paragraph is in the appendix if you want to compare.
I am often asked whether I agree with the non-logged business strategy, and the questioners are always surprised when I say I do not. After all, the anonymous user strategy sounds like a reasonable extension of the internet today and a plausible explanation of the social nature of humans. Also, the anonymous user true believers tend to declare victory, and write as if their theory has already superseded a narrow, reductionist dogma that ad based businesses destroy all privacy. In this essay, I’ll explain why I think that this reasonableness is an illusion. The more carefully you think about building platform internet businesses on anonymous users, the less sense it makes, and the more poorly it fits the facts of human psychology and a future built on machine learning.
Comments? Questions? You can reach me, where else, on twitter at @ntaylor963.
APPENDIX – MORE LINKS AND READING
My related posts:
- 2015 is a transition year to the (somewhat creepy) machine learning era. Apple, Google, privacy and ads.
- My 2013 post on twitter: Area man reports “I’m in a promising local blog.” Which is why Twitter needs Pandora style timelines.
- My 2014 post on twitter, suggesting casual users get their own app, separate from power users. What’s depressing is how most of the suggestions from this post feel current today. Twitter should copy Facebook the Conglomerate or Facebook Paper, not Facebook the web site.
- I’ve written about group selection and genetics as well. So if the Pinker quote piques your interest, more here: Let’s get real about Selfish Genes, Group Selection and Altruism. Towards a less impoverished view of humanity.
Other posts mentioned in this piece:
- Ben Thompson Twitter’s Marketing Problem. Also see his explanation of why Twitter needed a new CEO, written months prior to Dick Costolo’s resignation: Twitter Needs New Leadership.
- Chris Sacca What Twitter Can Be and What Twitter Says, and What Investors Hear.
- Steven Pinker The False Allure of Group Selection
Original and unaltered version of Pinker’s paragraph misquoted above:
I am often asked whether I agree with the new group selectionists, and the questioners are always surprised when I say I do not. After all, group selection sounds like a reasonable extension of evolutionary theory and a plausible explanation of the social nature of humans. Also, the group selectionists tend to declare victory, and write as if their theory has already superseded a narrow, reductionist dogma that selection acts only at the level of genes. In this essay, I’ll explain why I think that this reasonableness is an illusion. The more carefully you think about group selection, the less sense it makes, and the more poorly it fits the facts of human psychology and history.