Robert Scoble asked earlier today if Twitter is Now Reliable?
I have been using Twitter a little more than normal over the last week, maybe it will last. During that time I have been looking at the reliability of consuming Twitter content via various tools (no I haven’t tested everything).
Last night was actually a very good opportunity to do some data analysis, because Robert was very active on Twitter, with lots of updates in a very short period of time.
Could you follow the conversation on every platform?
Twitter Web Interface
The first interface I tested was the Twitter web interface.
I grabbed a screenshot of Robert’s twitter updates, and then went to my friends feed on Twitter to compare how many of Robert’s tweets actually made it to my Twitter Web page, to allow me to follow the conversation.
What I have done is highlighted in yellow the tweets that didn’t make it though to my feed.
4 out of 20 messages made it through to my Twitter home page “with friends”
20% Deliverability = Fail
This is a slightly different situation to the main Twitter web interface, because of the clumping I previous objected to. They only pull some of the updates.
When you click on that huge lump of tweets you get something like this
I have used numbers to represent the order that the tweets were actually written, and you can see that they appear out of order on FriendFeed.
When grabbing the tweets in the clump, Friendfeed is able to make a specific call to retrieve Robert’s tweets for that time segment using the Twitter API, thus all the tweets appear.
To follow people who don’t tweet very often and to be sure that you see every one of their tweets, Friendfeed might actually be a better interface for twitter than twitter itself, if it wasn’t for the clumping of high activity.
If Twitter continue to only deliver 20% of tweets to other interfaces, it makes more sense to use Friendfeed, but reading them out of order, you might as well click through to Robert’s timeline on Twitter.
Tweets in clumps out of order = fail
I have been testing out Twhirl for a while, and used the same data to see what is picked up.
I haven’t used the same screenshot as for the web interface, and the data was compiled from my Twhirl history probably 3 hours later than the first screenshot.
You will see that Twirl again only has 4 out of 20 tweets
note: I only have Twhirl pulling in data 40 times per hour – as it is only pulling data that is the same as on the web interface, this isn’t an issue.
20% Deliverability = Fail
20% Delivery = Low Traffic
If you were using an email marketing solution, and it offered only a 20% chance of the message being delivered to your prospects’ email box, without allowing for open and CTR, it would be a major business liability.
What is the point of having 10,000+ Twitter followers if your messages are only going to be delivered to 20% of them?
In addition there is the problem that only a small percentage continue to use Twitter – heavy attrition.
Direct messages seem to be a lot more reliable, and I haven’t got the patience to manually determine if this deliverability affects people with less than 100 followers.
Twitter is not reliable, they just throttle the amount of updates you see
A number of people have raised one issue with this post, the use of @
What is being filtered from what you receive are @ messages to people you are not following, but this isn’t a universal rule.
I don’t follow Chris, maybe I should… but then I am only following just over 300 of the people who are following me – there are others I am not following, I am slowly increasing the people I follow, but priority is regular readers and commenters – my core audience.
For me seeing @ messages to people I might not be following is extremely important, as it plays a vital role in discovery of other friends using Twitter.
People use the @ reply mechanism to ensure that people who would be most interested in the tweet see it, not just as a reply in conversation.
They are not aware that by using @ they are also preventing, or reducing the number of other people the message will be sent to.
Just imagine Robert discovered through his Microsoft contacts that a Yahoo deal was finally on the table. Rather than use a normal tweet, he uses @techcrunch to make sure Michael Arrington sees it.
That would then potentially limit who sees that message to those following both Robert and Michael, plus an arbitrary number of additional random people who are just following Robert, or maybe none if there is some kind of threshold based on the number of people following both.