What does Mixpanel think about adding statistical significance where appropriate? Take the Fbconnect and nonFbconnect users vs song engagement in the video. The chap in the video was fairly certain there was no significant increase, but visually it looks as if there in fact was have a slight, but significant difference between the two patterns of engagement. Impossible to tell without some sort of statistical analysis. There are probably lots of cases like these if your offering is fundamentally sound.

If I was the guy that championed and implemented FBConnect in my app and one of my co-founders thought that based on that graph there was no significant difference I'd be pretty pissed off.

> 1 billion actions per month

What a bamboozlingly humongous number! So I tried to figure out what it might mean in concrete terms. Assuming an "action" is some tuple of (32bit timestamp, 32bit customer_id, 32bit user_id, 32bit action_type_id, and say, 64bits for some custom value associated with the action)..

    >>> (((1e9 * 12) / 365 / 86400) * (4 + 4 + 4 + 4 + 8)) / 1024
    8.9183789954337911

Amounts to about 9kb/sec of raw data.

Average data point is actually 320 bytes. If we were doing a billion, it'd be about 116KB/sec.

We do more than a billion though. =)

Odd horizontal scrolling on properties under funnels. (Doesn't work for me in Chrome)

Also I'd love to be able to see a graph of my funnel conversion rates over time.

We had this before, I just didn't have time to add it back in. I'm working on it right now.

Strange, works just fine for me on 10.0.634.0 (Windows)

Same here with Chrome 10.0.634.0 dev (OS X)

Yo Suhail,

Was MixPanel once an opinion/debate site? I swear the name sounds familiar.

Congrats guys!