How many impressions to judge if banner is good or not ? (7)
Hey guys,
How many impressions do you give to a banner before you know if you keep it or not ?
Thanks 
Too broad a question, what source, vertical, angle/broad/niche?
I dunno.
It depends on if you're going CPA or CPS, what your niche is, what your payout is, etc...
I do CPA right now and the payouts are around 5-10 dollars, so I know within around 20-30k impressions whether or not an ad is working as it will typically generate me a conversion by then or will show some promising data like high LP CTR on CPV lab.
It's different for every niche. If you are trying to build a list and get someone to pull out a credit card...well that could take 100k or 1 or 2 million impressions to determine.
Like Maynzie said...depends on the niche really.
Your best approach is to figure out what CTR and signup rate you need for the offer/traffic source combination you're promoting, then use a statistical calculator like this one to determine when you're reasonably certain you will/won't hit that with each banner.
(I've got a mini-tutorial on doing this coming up soon - watch this space.)
I generally make pretty quick decisions with banners based on intuition. When you've been working in the same niche long enough, you get a good idea of what has potential. I can normally see within 5-10K impressions whether it's going to be a contender.
You'll need a lot more data to get anywhere close to statistical significance. But if every affiliate relied on that, we'd all have been broke before we made our first dollar.
Depends what you're talking about, I'd say - signup rate or CTR. Unless your clickthrough rate is absolutely tiny, you'll hit statistical significance on CTR fairly rapidly for lower confidence levels: 10k impressions on a banner with a CTR of 0.28% is above 0.2% at 80% confidence - which means only a 10% chance it's actually a loser.
Signup rates take longer to converge, though, I agree. I tend to use statistical significance stuff for negative sets there - calculating the maximum likely signup rate for a low-performing banner, and killing it if there's no chance it's going to get to a decent positive number.
Intuition and "feel" definitely work too once you have the experience. A mathematician friend of mine told me a while ago that in practise, the difference between an expert's intuition and actual statistical certainty tends to only be about 20%. Add in the fact that a human brain can spot patterns in a way that stat calculations can't and an expert affiliate may well be 95% as accurate as statistical significance whilst spending a lot less on data.
I think going for the mix of data + experience is obviously best. Don't forget all the data we get is pretty limited in the beginning, I bet everybody has seen clear winners turning into losers faster than expected and vice versa.
Home > General > Affiliate Marketing Forum