02-07-15

# traffic allocation

If I have traffic allocation set to 50%, shouldn't the visitors be almost equal?  Does that take a little while to even out (I just started 3 days ago).

Also, is it safe to assume that Bounces are the opposite of Engagement?

Level 1

greg 02-08-15

## Re: traffic allocation

Yes, the number of visitors should be about equal, but it may not be exactly equal. It's like flipping a coin: If you flip 10 times you're not always going to get 5 heads and 5 tails... But the more you flip (ie, the larger your sample size), the closer you'll get to a 50/50 split.

You're correct about bounces and engagements; they're exact opposites.

Level 2
tammyh 02-09-15

## Re: traffic allocation

Hi, I'm having a problem with this as well.  I understand that it won't be exacly 50-50, but some of my current experiments are seeing splits of 80-20 even though the settings (for A/B testing) are set to 50-50.

What do you think is the cause of this?

Level 1
Highlighted
greg 02-09-15

## Re: traffic allocation

How long have the experiments run? What's the total sample size (users tested) and sample size per variation?

If the sample size is >1,000 and you're still seeing an 80/20 split despite having it set to 50/50, then there could be some issue and hopefully someone from Optimizely can check it out.

Level 2
tammyh 02-09-15

## Re: traffic allocation

It's been a bit more than 3 days. Total sample size for one experiment (as an example) is just over 5,000 visitors--but one page has about 4,400 visits and the other has about 800. Will check in with our sales contact and see if they can help us look into it. Thanks!
Level 1
Roberto 11-09-16

## Re: traffic allocation

Hi there!

I got same issue. I understand that it won't be exacly 50-50 but 65-35 is not good.

Anyone has any idea about that problem?
Level 1
CouchPsycho 11-09-16

## Re: traffic allocation

Hi,

just one idea and something i myself experienced in earlier times. Is the experiment setup completely new or do you have modified one, which already existed and which already bucketed users into variations before?

Kind regards

Michael
Kind regards

Michael

Level 11