The 'Twitter Effect': Possibilities and limits

Is there a "Twitter Effect," by which the rapid spread of information through the microblogging service can crush a Web site with traffic? As I see it, the answer is yes, but it's not as simple as it might appear at first blush.

First, the background: Mashable CEO Pete Cashmore, whose Twitter postings more than 52,000 people follow, concluded his February 2 tweet mentioning that a blog post on "how to use Twitter to find your next job" took that site down with an inundation of traffic. He suggested calling the phenomenon the Mashable Effect.

Next, Pingdom, a company that tracks site availability, suggested instead the more general and more apt Twitter Effect:

It will be interesting times if Twitter is about to join the ranks of Slashdot and Digg as a potential site crasher...

The Twitter Effect formula = (Original tweet * followers) + (retweets * followers of retweeters) + (retweets of retweets * followers of those), and so on.

This way, tweets can spread out like the branches of a tree or a root system and reach a very large number of Twitter users. The spread is basically only limited by the size of Twitter's user base. If the tweet contains a link to a site, this site is bound to get a significant amount of traffic as the tweet spreads.

Overall, Pingdom has it right. People like to share links with their followers over Twitter, and good ones get passed down the chain, or retweeted (look for the all-too-easily missed "RT" in front of a tweet). But there are some caveats that probably should be noted before people assume their 140-character bright ideas will bring the Internet to its knees.

(Credit: Twitter)

I see three circumstantial considerations. First, It helps to start with somebody like Cashmore, with his relatively gargantuan Twitter audience. Memes from lesser figures would take much longer to spread, if they did at all, giving Web sites more warning that heavy traffic was on the way and time to respond.

Second, at this stage in Twitter's development, it helps when the tweet is about something the tech-skewed Twitterverse cares about, with Twitter itself being high on that list.

Third, plenty of Web sites have had well over a decade to figure out how to handle heavy traffic problems, so it'll be the smaller sites that are likely to have insufficient capacity to handle the traffic.

Then I see three more general caveats, too.

First, a key part of Pingdom's back-of-the-envelope math is the retweet effect. But many people have symmetrical relationships on Twitter--I follow you, you follow me. Consequently, the amplification factor of retweeting will be reduced by some echo-chamber factor that Twitter users are seeing their own tweets retweeted.

Second, there's the interest factor. Some fraction of people just won't click on the link or retweet it. So Pingdom's formula mostly governs a theoretical maximum spread. But that's kind of like guessing every neutron in a fission bomb strikes another uranium nucleus.

The Twitter Fail Whale is seen less often these days, but Twitter's infrastructure does potentially limit the spread of viral information.

(Credit: Twitter)

Last, there's a scaling problem: If the retweet network is so well established that memes are constantly bubbling up into virtual flash mobs, Twitter itself will be inundated with traffic. It's true that the service handles heavy loads vastly better than it did a few months ago, but even if Twitter holds up, will people? If your Twitter stream is constantly awash in tweets and retweets, many of them repeats of other retweets you've already received, will you spend your day glued to Twitter for the latest updates?

For those of us who get hundreds or thousands of e-mails a day, it takes more for an interesting message to stand out above the crowd, and Twitter doesn't get some magic exemption from the signal-to-noise problem just because it's popular. And of course, any popular medium will be a target for spam and opportunistic marketing efforts that degrade people's experiences.

Pingdom does have it right in broad terms, though. By virtue of the retweet network, Twitter can be a powerful, fast-acting amplifier for information well suited to viral spread.

But it's probably good news more than a reason to panic: although some Web sites may crash as a result, my guess is that Twitter more often will just bring Web site publishers the traffic they crave.

[asilentthing]

I think it needs to be a shorter term. That takes up a ton of that 140-character limit :)

[Bertil_Hatt]

Actually, there are already formulas on how to estimate the impact of one link, based on what makes that link repeatable. If we assume Twitter has a scale-free structure (the most likely assumption) and that generally, having a large following doesn't impair your re-twittablity (unlikely, as you are most likely to quote from someone famous) then any link might end up 'percolating' in the whole graph: there is no intrinsic unmanageable disadvantage, unless you have strickly no friends.

Being suggested by a twitterati makes your first-stage impact that much larger ? but what does kill you is that those twitterati know each other, meaning if #1 has many friends, so does #2, meaning that more re-tweets will reach an even larger base. This effect usually dominates the damping from "common friends don't count twice" (aka clustering) in explaning how an better-connected prescriptor has more influence.

[Dango517]

If you could find the magic formula you might be able to recreate it but then again it's magic. Magic tends to go it's own way, it's ephemeral , continually changing that 's why it's magical. Those that find it seldom keep it for long. Today opinion shooting star, is tomorrow's has been.

Let's say you did find the formula, would you then use it for good or evil? Maybe nuke a web site for instance. In all likelihood, sooner or later your target or targets would become wise and take evasive action ending in a zero sum gain or possibly creating more effective counter measures then your own then where would you be?

In the final analysis it all boils down to sufficient wire, ports, adapters and servers few if any web services have enough of these to deal with the occasional wild fire so these are simply allowed to burn themselves out.. The law of the land is averages and it will prevail.