rotating offers (10)
when is a good time to cut out a network offer when testing?
between 2 of the same offer,
i have 411 clicks 21 conver 26c epc,
and the other 370 clicks 27 conv, 39c epc.
should i just let em both run, or is the worse performing one slowing down the better converting one?
thanks
Use a split test calculator like this: http://visualwebsiteoptimizer.com/ab...ce-calculator/
The answer in your case: keep testing
http://www.charlesngo.com/statisticalsignificance/
I have a similar question to voan.
I want to cut my offers. I have 4 in rotation. In split test calculator I enter number of visitors(25) and conversions( 0) then press calculate significance. I get "Yes!" So I am good to cut that offer, correct?
Note: I didn't change the variation [700 and 150] and P =0
I'm a bit confused by your example... Did you only enter the values for 1 variation and left the second one with the default values? That would explain the "Yes", as you would compare 0 conversions to the default of 150.
You always have to compare two variations to each other, so e.g. your top performer in the "Control" column with the worst performer in the "Variation" Column.
The calculator won't tell you if a variation is "bad" or "good", just how it performs compared to another one.
Put in the impressions under "Number of Visitors" and the conversions under "Number of Conversions".
E.g. variation A with 500 impressions, 10 conversions vs. variation B with 700 impressions and 12 conversions would look like this:
Originally Posted by andyvon
what about when my offers aren't converting. I read caurmen's post How To Cut Offers And Landing Page and still a bit confused...
We cut offers when
1) spent clicks worth 4x payout with no conversion
2) or use AB split test calculator to find the best offer
Now if my setup is 10 banners> 4 Landers> 4 offers ($6 payout)
-$47 spent for total impressions
-sent 25 clicks to each offer
to cut do I simply
$47/4(offers) = $10.75
$10.75 < $24 (4*$6)
So I need more clicks...yes?
Yep, you need more data 
The amount of data you need on a macro level really depends on the conversion rate. We have offers that pay $100 and convert a .2-.4% and are some of the best offers onr ou network.
BUT they require a ton of data to optimize to and therefore typically a ton of money.
the 2x or 4x spend rule doesn't make much sense to me. Because statistical significance has nothing to dow with click cost and payout. It has to do with clicks and outcomes. Click cost could be super expensive or super cheap. Payout could be super expensive or super cheap. Both are immaterial of confidence intervals.
If you want to see what you think things should come in at I'd always do your math first and set your reasonable expectation. Do you Marketing Math prior to running you first clicks based on your projections of click costs and conversions rates. If you're not familiar with marketing math here's a post I did on it a while ago http://www.oooff.com/php-affiliate-s...ore-of-it-all/
I'd suggest keeping a excel spreadsheet where you can pop numbers into it and have a range of expectations it calculates out for you. Make it so you have variables at the top. And then have rows of data with 2-5% variations either direction on each row. Then you can how far off your projections you can be and still be successful.
Originally Posted by Smaxor
Thanks so much. Your blog is golden, it cleared up so many calculation questions I had.
@smaxor - I generally agree with you on the 4x rule - if you're comfortable with doing the expectations math that's a better way. Practically speaking I've always found the 4x rule is likely to tally with a fully-worked-out confint / expectations spreadsheet to within about 85%.
Would be very interested to hear if your experience differs on that - there may well be edge cases I haven't considered.
Home > Questions and Answers > General Questions