How traffic is divided
In TDS.SO you list several URLs in one destination, separated by |. Each click picks one of them at random with equal odds, so with large volume the traffic splits evenly: two URLs get about 50% each, three get about a third each. Small volumes will not be exactly even, which is normal.
To test on part of your traffic only, for example one country, put the URLs into a rule for that country instead of the main URL.
Running a fair test
- Change one thing between variants: the headline, the landing page or the offer.
- Tag each URL differently, for example
?variant=aand?variant=b, so reports can tell them apart. - Send the variants the same kind of traffic at the same time.
- Wait until each variant has enough conversions. A few hundred visits with a handful of conversions often show a random difference.
- Compare final ROI, not only clicks.
Split test vs A/B test vs multivariate
In everyday use "split test" and "A/B test" mean the same: whole variants compared against each other. A multivariate test changes several elements at once and needs far more traffic to say which element made the difference. For most affiliate setups, a series of simple two-variant tests is easier to read.
Reading results in TDS.SO
The click log shows which URL each visitor was sent to. With the tracker on, postbacks tie conversions to clicks, so you can compare conversion and ROI per offer or per tag. See redirects and the A/B test guide in the docs.
Questions
How long should a split test run?
Until each variant has enough conversions for the gap to be more than noise. There is no fixed number of days, and low-volume tests need longer.
Does a split test always divide traffic 50/50?
No. With several URLs separated by | in one rule, each click picks one at random with equal odds, so the split is even on average and only approximate on small volumes.
What can I split test?
Landing pages, prelandings, offers, or the same offer on different networks. Change one element at a time so the result is clear.