Home > Paid Traffic Sources > Native

1 Weird Trick To Improve Your Campaigns :) My approach to systems - Please Critique!! (7)

02-19-2017 08:56 PM #1 mass_marketer (Member)
1 Weird Trick To Improve Your Campaigns :) My approach to systems - Please Critique!!

Quick(ish) post as I realised I've been running my campaigns with no real system, just acting intuitively rather than using a system based on the data.
I think the latter is key to managing risk and finding profitable campaigns and before I start native (used to run Adult), I thought it would be sensible to lay out a system that I can follow (and tweak over time), to bring discipline to my campaigns. I've written the steps below; would be great if some of the more experienced guys could take a look and provide some constructive criticism .
(I'm hoping this eventually becomes something anyone struggling to be systematic can learn from)

Thanks a lot! (P.S. I'm planning to run skin in Tier II geos)

Step 1 - Find an offer

  1. Ask AM for top performing offers in Tier II geos
  2. Check whether your desired traffic source has sufficient volumes in the geos you plan to test
  3. Decide on an offer and geo to run
  4. If spend offer payout x 10 without a single conversion, kill the offer


Step 2 - Ads
  1. Use spy tool to search for recent, high traffics ads
  2. Download top 10 images (and associated headlines) and resize as appropriate for the TS [Note: This feels inefficient/slow, need to find a way to streamline/automate this process –suggestions welcome ]
  3. Test 5 images with the same headline to find the best image based on the CTR after [a statistically significantly] number of impressions [Note: I'm not sure what statistically significant will look like yet]
  4. Decide on which headline to run with the 5 images based on ripped ads / your insight into what makes your market tick/clickbait generator
  5. Decide on the best image out of the 5 previously tested and run with 3 – 5 headlines to determine the best headline (Each headline should represent a slightly different angle)
  6. Rinse and repeat the outlined steps to continually find higher CTR images and headlines


Step 3 - LPs
1. Use spy tool to search for recent, high traffic LPs running your offer in the same geo, as well as other geos
2. Use spy tool to search for LPs in the same vertical and geo, but for a different offer
3. Rip LPs
4. Review LPs for any backdooring / malicious script OR get coder to recreate
5. Have pages translated as required
6. Run two ripped LPs at a time
7. Study other LPs to determine what works in this vertical and also what adjustments you can make to the ripped landers
Step 4 - Bidding / Placement optimisation (this is definitely the trickiest part)
1. If the bidding platform is not performance based (i.e. impressions are solely driven by price), map the bidding landscape as per Caurmen’s post back in 2015
2. If not, start with an ‘average’ CPC bid – ask account manager what average looks like (as I understand starting low means you end up with the very worst traffic on Native so you can't tell whether you actually have a decent campaign on your hands or not)
3. Cut placements riddled with bot traffic (e.g. super high/low ad or LP CTRs) as quickly as possible – currently using noIP so this should help with bots though will need to keep an eye on the bleed rate
4. Cut placements based on a combo of CTR and ad spend - haven’t quite decided what this will look like yet, I think spending 2-3x the offer payout isn’t the smartest move given the extent of awful placements on native
5. Increase spend on converting domains

That's it...for now!

02-22-2017 12:26 AM #2 mass_marketer (Member)

Anyone???

02-22-2017 03:40 AM #3 ackbar22000 (Member)

looks good, try it !

02-22-2017 10:24 AM #4 caurmen (Administrator)

First thing that springs out when I read this: cutting ads and placements based on CTR.

Don't do that! You'll end up with ads that throw a lot of clicks - which then don't convert.

The only time you should cut ads or placements on CTR is if the cost per click they're generating is so high that you can't possibly make a profit.

Otherwise, always cut ads based on conversion rate or return on investment.

Hope that helps!

02-22-2017 08:47 PM #5 vortex (Senior Moderator)

Caurmen has already pointed out one of the shortcomings. I would also suggest not to test landers 2 at a time, but 5-10+ to narrow down to a winner. That way you'd be casting a bigger net. And don't think that testing 2 landers across 5 rounds of split-testing will let you test 10 landers - because every time you start a split-test you'd need to keep the previous winner as the test control. So effectively the more split-tests you run, the more landers you're re-testing. Hope that makes sense!



Amy

02-25-2017 10:47 AM #6 mass_marketer (Member)

Thanks Caurmen and Amy - noted and incorporated into my system.

02-25-2017 02:17 PM #7 durell217 (Member)

Wow I like this system but I definitely agree with Caurmen. You want to optimize based on conversion and not purely CTR. Made that mistake in the past and got burned badly for it.

Home > Paid Traffic Sources > Native