Think you test your ads enough - no you don't (23)
Just watching a BBC documentary, here in the UK, about facebook. It showed an advertising company in London who test ads for brands. Testing 27,000 variations of the ad to get the most effective one.
27,000? Hold crap!
Yeah, I just watched that. They were TBG Digital.
Haha, just came on here to post the same thing! TBGs website crashed within seconds of being mentioned on that show - you'd think they might have beefed up their load balancers beforehand...
Yeah but I bet they were working with a damn bigger budget than most of us. I suppose if you started with 100's of images and multiple headlines/body copy/landers and were running high CPM bids it wouldn't take that long to get the stats to kill off variations and test new ones. But damn that's some data.
Were they testing broad for the UK?
27k variations is crazy though.
It's not THAT many when you think M/F x InterestedIn x RelationshipStatus x AgeBracket x GeoTarget x Images
You can get 27k really fast.
There's actually a better way . . . you can use Taguchi methods to significantly reduce the number of ad combinations. For example, with the right set-up, you can effectively test 81 ad combinations using only 9 ads (plus a little bit of analysis). It's the way to go when 1) Testing is expensive, or 2) You need to test FAST.
If you have the budget and time, brute force multivariate will give you the best results. But if you can get close for 1/9 the time or budget, why not?
Originally Posted by inversion
How can we transfer TBG Digital's ONE Media Manager is an automated software platform that empowers marketers to develop highly effective Facebook advertising campaigns that deliver results.
Build thousands of ads in minutes
Make mass changes to ads within seconds
Automatically optimize ads in real time
Manage all ads in one central screen
Schedule ads to reduce wastage
To use this for Media buying not just FB.
Just 8 variations spent $60k within 14 days. On FB.
http://www.google.com/url?sa=t&sourc...rZFT2voijZAAoA
links of documentary?
There's not a lot of good documentation on Taguchi, unfortunately. This guy provides a free spreadsheet for the situation I was talking about above (test 81 using 9 ads).
http://www.splittestaccelerator.com/spreadsheet/
For Facebook, just use replace the second ad copy line with "Image"
yep, stuff like that is where it is moving towards.
Im looking to implement something like this along with the military application 'big deep green'
but that is awhile off [so you are safe for now :P]
Originally Posted by blackberry
Use a UK VPN or Proxy if outside UK to watch iPlayer. Failing that, it will probably end up on eztv.it before the end of the day.
Let's face it, TBG will have ad creation down to an art for facebook. Can't see them going "Do'h!" all over the office all day due to refusal emails from fb LOL.
But damn, the brands they represent are spending some SERIOUS coin to have this level of testing performed.
What I'd give for a whizz through their database ...........
Couldn't get tired of kissing Emily Maitlis either 
I found Split Test Accelerator shared on a BH forum so here it is
Tnx inversion for the taguchi method share 
Here is Download link: http://www.mediafire.com/?3n185bof0i917c6
Password: stm
Yeah, 27k ain't that big of a deal. As mentioned above, when you get into multivariate tests, the unique ad numbers add up quickly. Back in the day I built an ad uploader and tested over 100k ad variations on POF.... not including demographic targeting variations.
Originally Posted by kyleirwin
Originally Posted by krisicash
PT
@polishedturd
Do you know this guy? I was wondering if he is still actively developing the software? I checked the forum he has setup and there is almost no activity over the past year.
I've actually purchased the full license to STA. It is a great solution for those who a) want faster results, and b) want the data hosted on their server!
He's written some great stuff, and the product is well worth it. But, there is a learning curve.
I don't think he is actively developing. Looks like he has moved on to another project: http://www.fractalplanner.com/
@inversion
So in your opinion (as one of our resident soothsayers of all things analytic 
), do you presently think it is worth purchasing and going through the learning curve to yield the fruit this tool offers? Or are there better alternatives for this type of rapid testing/prediction? (such as something coupled with your ROI Calc spreadsheet?)
Thx!
@jakz -First thing everyone should be doing is A/B SPLIT TESTING. Someone who is not consistently doing that will not end up doing anything with Multivariate testing. Only jump into Multivariate testing when A/B split testing is becoming inefficient from a time standpoint.
Once an affiliate is regularly split testing, then yes, I recommend learning about Taguchi multivariate testing. But start small. Use the spreadsheet link I posted above which provides the simple case where you have 4 variables, each with 3 options. You can use that for FB ads, PPC ads, and even banners. Do that regularly.
The Split Test Accelerator is the next logic step for those that want to use multi-variate on their landing pages. It's a really awesome tool when it comes to that.
Can Split Test Accelerator be used with CPVLAB/Prosper or is it more of a stand alone thing? I noticed on the product page it has a built in tracker.
Yeah, it can be used with other tracking. I run things through CPVLab first.
Home > General > Affiliate Marketing Forum