You’ll probably never know it by looking at the final result, but I’ve been working on this post, at least thinking about it, for several months. I’ve been reluctant to publish it because every time I start, it tends to sound like a rant and that’s not what I want this to be about. As anybody reading me knows, I’m not afraid to rant, but in this case that’s not my objective – I really just want to share some lessons learned in the hope of informing the dialog.
Some of this may be obvious “well duh” sort of stuff for others out there who may be a lot smarter than me. Hopefully it won’t all be that way.
As many of you may already know, over the last several years, I’ve worked on and otherwise participated in a number of side-projects, including:
- Phweet – a phone call in a URL
- Taglets – the story of tagged things
- Twitmart – Twitter Marketplace for social classifieds
- SuperTweet.net API – An Answer for Twitter OAuth-pacalypse
- TweetPress (140plus.com) – a free Twitter blog at http://YOU.140plus.com
Each of these can be considered a failure, depending on the metric or dimension one examines. Except for SuperTweet.net and TweetPress, all of them have since been shut down. Each of them have been a disappointment, to some degree or another, for me, personally. However, they also each provided various positive rewards, in addition to the sometimes painful lessons learned.
Metric 1: Popularity
Certainly, with the exception of SuperTweet.net, they are all failures in terms of popularity – they never gained any significant traction.
With the exception of Taglets.org, each of these services require a Twitter account – the only way to use the service is using your Twitter account. This was an intentional decision, where these services were supposed to be integrated with and tightly coupled to Twitter and your Twitter social graph. I don’t recommend such a strategy for anything real you may be doing. It’s fine for side projects like these, but it’s a bad decision for something you might hope to make into a business and you only need one reason for that, according to Pew Internet:
8% of online Americans use Twitter
So by limiting your target market to only those using Twitter, you are automatically excluding 92% of potential users. And that’s the number as of December, 2010 (according to Pew) at a time when Twitter was claiming over 150 million users. It was even worse, much worse, back when we released Phweet in 2008 when Twitter had only a few million users.
There are other reasons to avoid putting too much emphasis on Twitter as a social network (which even Twitter says it is not), some of which I cover below, but obviously the 8% problem is all you really need to know.
Opening up your site to Facebook users will expose you to about ten times more users. That makes good business sense – however, in my case, I don’t particularly want to help Facebook lock-in more users and more sites, so I have elected to not implement Facebook Connect – a bad business decision, but one I accept on principle (and admittedly more weakly, on long-term business strategy grounds). I don’t encourage anyone to follow my lead on that, unless you also are doing so for something other than sound business reasons.
It’s too early to call regarding TweetPress one way or the other, but given that it is Twitter account-based, it suffers from the above problem of excluding the lion’s share of potential users that don’t have Twitter accounts, right off the bat. The service has not received any TechCrunch, GigaOm, or other such “juice” as Phweet and Twitmart did, and is only being promoted through the limited reach of my own followers and a few friends and others that have started using the service. As a result, growth has been slow, if not almost non-existant. Either way, it serves my needs and it continues to be a wonderful platform for experimenting with different technologies, mashups, APIs etc. I have no plans to shut TweetPress down, regardless of whether it ever achieves wide adoption, but it’s always more fun when such services are embraced by a larger community.
Twitmart was being used by a few users, but it never even started toward the goal of helping users build a trading community with their Twitter Friends and Followers. Nobody used it that way. Instead, it became little more than a vehicle for spammers, so we shut it down.
As noted above, SuperTweet.net isn’t a disappointment in terms of adoption. It has been steadily growing since Twitter shut off Basic Authentication on the API. Unfortunately, most the users are parasites, contributing nothing in return (with notable exceptions, @ae6rt @TheDarkNighty and a handful of donations) – certainly it has not produced the kind of community of support I had hoped it would (naive, wishful-thinking on my part). Update: I have to take this statement back – there has been a significant upswing in community support for SuperTweet.net – thanks folks.
About the 8%
This brings me to some of the lessons learned about the 8% of people using Twitter. Let’s look at the Twitterverse in more detail.
With each of these services, we have learned something about the evolving Twitter user-base. And as Twitter approaches double-digit penetration into the mainstream, it may also provide some lessons about the Internet population at large. We have seen this pattern before. The early adopters skew the personality of a service in the beginning. We saw it with Usenet, way back in the late eighties, and the pattern has repeated across many services over the years, right on up to Quora today.
In the beginning the services are populated with users that are smart, helpful, active etc. – there is a real community and it’s a regular utopian bliss. But then the real world steps in. And guess what, not everybody is so smart or sweet and nice. Some come from harder backgrounds where making a living might mean $1 a day. There are sleazeballs of every flavor with all kinds of motivations. If there is a way to exploit the service for their own personal gain, they find it, especially if it costs them almost nothing to do it (whether it’s the cost of a Usenet post, email, or a Twitter status update).
And Twitter is no different. As it’s gone from under a million users to hundreds of millions of users, the flotsam and jetsam of the population came along. As with email, the spammers and the like may be only a small fraction of the total user-base in number, but they can have a tremendous impact on the overall system. And there is a gigantic gray area of behaviors in the name of “social media marketing” and “SEO” going on on Twitter that sometimes cross the line. I read a business plan recently that literally referred to Social Media as the “advertising mecca of the future.” I suspect such is a common view among budding entrepreneurs.
One doesn’t have to look very far to find a huge array of consulting firms, PR agencies and others that will make all kinds of claims about their “secrets” and expertise in Social Media Marketing, often charging exorbitant fees – I shouldn’t even have to say it, but very few of them provide any value and most the time, engaging such a service will end up doing more harm than good by damaging your firm’s reputation.
Anyway, while I doubt the percentage of Twitter users that fit the above abuser category is very large, their presence is felt loud and clear. And they are the first to try to “game” any new system or service connected with Twitter, as well as Twitter itself of course. These are the folks that jumped on Twitmart and, I have to say, soured me on the project, and on Twitter as an ecosystem to a degree.
The Twitter “social graph” is a joke
In the bright young days of Twitter, a lot of us thought it might be a new kind of social network. And it was, for a while (and still is for a small percentage of Twitter users). But once Twitter started to aggressively market the service, attract celebrities, and celebrity stalkers, any chance of a meaningful social graph were put to bed. As Pew and other research has shown, there is nothing mutual in those relationships and no community or friendship aspect to them. Even Twitter finally acknowledged last year that they are a media company, not a social network.
Further, the Twitter Home Timeline, the stream of who you follow, has also become part of the joke. Most people that have a ‘reputation’ of any kind also follow a lot of people out of a sense of duty – they feel obligated to follow people so not to send a message that “we aren’t really friends,” even though they don’t really care much about that person’s tweets. Their “following” list is so polluted with such connections, they really never follow it at all. I had a high-profile Twitter guru tell me recently, in essence, that I’m not a serious or typical Twitter power-user because I don’t follow 500+ people. I brutally manage who I follow because I still actually read my timeline, basically every tweet in it. If I followed 500+ people, I may as well not be following any of them, because I’d never see most the tweets in that timeline/feed. But that’s exactly what a lot of people do on Twitter, making their home timeline, and thus their social graph, meaningless.
The implication is that any Twitter-related service that somehow relates to the social graph is working from bogus data and won’t be of any use to the largest segments of Twitter users, those for which the social graph is almost entirely meaningless.
Twitter is less real-time than you think
For news and events, the Twitter consciousness, the live timeline (fire hose), can provide wonderful insights as events happen. This is the broadcast platform role that Twitter serves very well.
In terms of real-time person-to-person conversation, however, Twitter is hit and miss, mostly miss. With Phweet, we discovered that most people don’t tune in to Twitter all that often or consistently. Here are some other issues that impair Twitter’s ability to serve as a communications platform:
- Many connections aren’t mutual, so direct messaging fails
- Twitter settings are too complicated, such that most people never figure out how to set up real-time notifications
- Messages get lost in overflowing timelines
- The 140 character limit kills deep conversation, or causes them to move elsewhere
Many services have emerged to try to solve the conversation problem on Twitter and they have all gone nowhere. In my opinion, Twitter as a person-to-person or small-group conversation platform is a non-starter.
In the end, Twitter is basically like a zillion RSS feeds overloaded with a messaging system (actually two messaging systems, @ messages that are semi-public, and direct messages that are email) wrapped around a search engine, and hooked in to SMS (at least in the US). Probably the original description of Twitter as a “microblogging service” is still the most accurate.
In short, people use Twitter to share stuff. Unlike a Facebook or LinkedIn style social network, tweets are generally shared with the world at large, not just ones mutual friendships. When, or if, people see what you share depends on a number of factors, not just whether they “follow” you or not. They might see your tweet as a search result, due to a retweet, as part of a Twitter list’s contents, or because the tweet was directed to them via @ messaging. And the fact that they follow you is no guarantee they will ever see your tweet, far from it.
All these things are important to keep in mind when conceiving applications or services that in some way work with Twitter.
So can you build a business, or even a viable service, around Twitter, exclusively? I don’t think so. Oneforty has done it (to whatever degree they are successful) and of course some Twitter clients have done alright (many have not done that well, especially once Twitter themselves became a competitor). Even though Twitter claims 200 million users, bear in mind that only 8% of online users use the service, and of those, only a tiny fraction make up the kind of users that can help build value in a community.
I’ve resorted to using an RSS reader to read select Twitter feeds. The reciprocal-follow practice not infrequently results in bent noses if not observed. And if you follow people who hold no natural interest, but you’re socially bound to follow, your feed is too noisy to be of any real use. Yes, there are lists you can follow, but those too are fraught with hurt feelings about who’s on the list and who isn’t. So using a conventional Twitter client to follow feeds turns out to be just not worth it.
Supertweet.net is the only reason I still use Twitter. If I didn’t hate Java so much I would probably install Supertweet on my Linode.
I’d pay a couple of dollars per month to get priority access–really hard to get a tweet through your system lately. I’d also pay a couple of dollars if you could offer a quality search service off the firehose.
For example, I’d love some kind of common sense API access where I could get notification when anyone on Twitter says “I need a WordPress programmer” for example. Maybe you already offer this through SuperTweet?
The Twitter API makes it extremely difficult to figure out all the various quotas and mysterious details and they are snobs about “whitelisting” people. I applied for whitelisting and they actually told me I wasn’t a developer! What?!
As for all the other clients out there (that offer to search for a phrase) so far I haven’t found anything nearly as useful as the client I developed for my own needs. But still I don’t get enough ROI from Twitter to bother rewriting my app for oAuth, which I still think was the dumbest business move ever.
I built a system that delivered notifications in real-time off the firehouse as you describe that we called “watchlists” – but it was never built for large scale and probably would not scale too well to millions of users. It did work and that code still exists and would not be that hard to revive, but I’m not sure I could come up with a model to support it – although I do get requests for something like it once and a while.
[…] is essentially still a mainstream failure, with only 8% of online Americans using it. It’s clear that one big reason is Twitter’s complexity and the inability of Twitter to […]