Home > General > Affiliate Marketing Forum

Multivariate testing?? (5)

07-14-2016 01:56 AM #1 ebaskin (Member)
Multivariate testing??

do you use it?

I'm thinking to go with a service to find the best variations for my landers... worth the hassle or no?

07-14-2016 03:09 AM #2 crysper (Member)

It's a lot to tell about this, but yes, it's worth it if you have enough traffic and you are serious about affiliate marketing(you are for the long run, doing things systematically, having processes, etc)

The people that I see that are very successful with multi-variate fit this profile:
- they run lots of traffic daily
- they run more like/are a business(they set systems, processes, optimize every part of the funnel, etc)
- they have at least a VA, unless they are a team
- they work very close with the networks or advertisers(so they know the offer won't pull out after they work on optimization)

I'm not saying that this won't work for the regular affiliate with high traffic, just that high traffic is not enough if you can't physically focus on the landing page optimization part of the whole. If you have a few high traffic and already profitable campaigns, you can give it a try, otherwise is better to stick to A/B.

07-14-2016 08:34 AM #3 offersdirect (Member)

We have tested multivariate extensively, but its a tricky issue. The problem is the Frankenstein effect, where best converting elements in different segments produce unexpected sub-pair results.
So unless you have really a lot of volume and have extensive experience using A/B testing I would not recommend it.

07-14-2016 07:39 PM #4 ebaskin (Member)

What service can be integrated with Voluum? You can PM me if you like too.

I asked VWO but they don't have an easy way of doing it.

07-14-2016 09:05 PM #5 crysper (Member)

Quote Originally Posted by offersdirect View Post
We have tested multivariate extensively, but its a tricky issue. The problem is the Frankenstein effect, where best converting elements in different segments produce unexpected sub-pair results.
So unless you have really a lot of volume and have extensive experience using A/B testing I would not recommend it.
That's why we use full-factorial and treat combinations more like stand alone A/B variations in LPOptimizer. Many times I had to tell people that you don't need separate performance data for each element like in the usual multi-variate testing because it's useless, you need the performance for the variation(or combination). It won't help you if you have a good performing headline in one variation, but the same headline performs very poor in another.

Home > General > Affiliate Marketing Forum