Tumblr Growth
I was curious how fast Tumblr was growing as a business, so I decided to spend a few minutes looking into the data. It turns out each post on Tumblr is assigned a sequential ID number that you can use to estimate the total number of posts on the site. For example, my post earlier today was ID 48896513 and a post of mine from a week ago was ID 47971938. By subtracting those two numbers, you learn that all Tumblr users posted over 940k items in the last week. I repeated that exercise over all my posts for more than a year to arrive at the chart above, which shows total Tumblr posts per day.
The results are interesting, I think. You can see that growth has been fairly steady since the beginning of 2008. The trend is good, though not quite as dramatic as some highly viral businesses that I’ve had data access to in the past. I’m not sure what the cause is for the unusually high post rate between August and October 2007, but I imagine it has to do with the affiliate spam problem that Marco has mentioned on a few occasions. Not surprisingly, the number of posts per day falls significantly over the Holidays.
Ok, enough of that… back to work.
Update: Marco sent me a nice note with some additional info which I’m quoting below with his permission:
As you mention in your Tumblr Growth post, we did indeed have an affiliate spam problem in August 2007. (The same thing happened this August as well, actually, but we were quieter about it.)
But the volume of posts from the affiliate spammers is insignificant. It was only about 3500 accounts, each only creating 0-3 posts before getting suspended.
One explanation of the giant spike, that interestingly ends right when we launched Tumblr v3 on November 1, 2007, is that v3 introduced two significant changes:
1. Post IDs were no longer “wasted” on failed photo uploads. Previously, every file upload attempt would first generate an ID, then try to process the file. Starting with v3, uploads are attempted first, then given an ID only if they succeed.
2. We started capping feed imports. This is the big one, and the variable most likely responsible for this dip. This came in two parts:
2a: Limit of 5 feeds per account.
2b: Accounts have a 100-post-per-day limit (or 75 for photos). Feed imports above that every day are rejected. But an account could still create 100 feed-imported posts EVERY day. Around the v3 launch, I made a change: if a feed import causes an account to break its per-day limit, that feed is permanently disabled. This dramatically cut down on accounts that people were treating as feed readers, importing a bunch of random high-volume news feeds (Google News, Digg, Engadget…).
If you had a way to distinguish feed-imported posts and only measure “original” post growth, I think you’d see a very differently shaped curve.
Thanks Marco.
