Home > Paid Traffic Sources > Native

These Ads Suck: How can they be made better? (14)

07-27-2021 03:54 AM #1 jack_l (Veteran Member)

Originally Posted by kweens

Hi everyone!

Some time ago I built a spy tool to go through the Alexa 1M (It's really the Alexa 605K) and go through all the revcontent sites in the top 1M. Did I find all of them? Definitely not, but the list I found should be a good starting point.

Below is an image with a collection of ads that my home-grown spy tool has found. These are meant to be desktop us. What was interesting is the frequency with which some ads appear--despite the ad units / widgets being viewed from the same device, ip, and browser. These are likely ads paying CPC instead of CPM. The highlighted row was the top ad that I could find which was promoting an offer on the Clickbank Network. The main advantage of clickbank is the payout speed. I don't have the same level of relationships with affiliate networks that others on this forum enjoy--most notably the weekly payouts. So the next-best-thing was clickbank, where the time from conversion to payout is about 2 weeks.

Some spidering, spying, and scraping later, I have a batch of 128 ads set up on VoluumDSP, the overwhelming majority of which perform terribad. The best performing ad has the potential to reach 0.36% CTR. And that's with slightly better than 2.5% chance, or slightly better than 1:40 odds. For instances with 0 visits, I simulated the most recent impression being a visit to get a sense of the range of CTR.

To explain each column in the screenshot below:

Creative Name: Title/Headline
Images: Self-Explanatory
Impressions: The number of impressions won on VoluumDSP.
Visits: The number of clicks from the impressions through to the landing page.
Clicks: The number of clicks from the landing page through to the offer.
iCTR: The CTR from impressions to visits.
Proportion: The Visits Divided by Impressions
Alpha: The target confidence interval. (95% confidence expressed as 1-0.95)
ZCrit: The critical value of z--a statistics measure.
u: A component of the Wilson Score confidence interval.
Lower: The lower end of the confidence interval.
Upper: The higher end of the confidence interval.
Lower PCT: The lower end of the confidence interval expressed as a percentage.
Upper PCT: The upper end of the confidence interval expressed as a percentage.

The confidence interval in this case is the range that has a 95% chance of holding the true CTR. In other words, there's a 95% chance the CTR of the campaign after several more impressions should be between the upper and lower level. From the ads thus far, it definitely doesn't inspire confidence.

The rest of the ads can be found here.

I also did some tinkering with the math.

Using the wilson score interval, an ad has a 97.5% chance of not reaching a CTR of 1% if it doesn't receive the first click within the first 562 impressions. Due to the nature of multiple widget IDs / ad units being involved in the test, I set up the automation to pause ads where they have 0 clicks (from ad to landing page) by 3,000 impressions.

The end goal--at least for this first stage--is to get to 1% CTR. From reading around it seems 1% CTR is the minimum acceptable benchmark for a native ad. Higher is of course better, but below 1% CTR the economics don't necessarily make sense.

Overall, I've spent about $1200 in ads so far and that resulted in $108 in revenue. About -91% ROI. Altogether I have an $xx,xxx budget to experiment against and learn from.

Going through the flow which generated that conversion, it took:
16,152 Impressions
12 Visits
2 Clicks

to generate one conversion.

Plugging that back into the confidence interval calculator results in:

Dividing the lower and upper confidence interval each by 1 implies that the expected number of impressions per conversion is between 2,851, and 91,499. At the current bid of 0.65CPM, suggests a potential CPA between $1.85 and $59.47. I'm keeping that particular flow running for now, until the math tells me it's impractical otherwise.

What other information is needed to make an informed suggestion?
What am I doing wrong with these native ads?

Hey man -

That is some absolutely incredible work you're doing! The math, the spytool, etc... TREMENDOUS, TREMENDOUS effort.

While I'm far from an expert on all this, it seems to me that everyone here on STM who has had success has shared that same conscientiousness and purposefulness that your post demonstrates, and you seem to have it in SPADES lol. I love all the data!

As far as how to get your camps profitable, here's my thoughts:

1. Voluum DSP is great and has a place, but I think I, Platinum, Jaybot, TheDudeAbides, etc would all encourage you to run on Revcontent directly. If you look back through the forum history, there are very few instances of people having big success on VoluumDSP, and quite a few examples of people have success on Revcontent itself.

VoluumDSP is just that - a DSP- where you can access extra inventory for campaigns that are ALREADY profitable. For learning the traffic source and making profits though, you absolutely 100% need to run on Revcontent itself.

2. This is going to sound crazy to you, but I would quit tracking ad ctr entirely. I understand that on VoluumDSP its important because your paying CPM, but once you start running on Revcontent itself, its just not a variable that can- by itself- help you optimize campaigns. The reason is because its so dependent on WHICH SITES Revcontent is showing your ads. A site with one three ad widget will have super high ctr's. A site with four 8 ad widgets will have super low ctr's. And for some reason Revcontent likes to show certain ads on certain sites, etc. Plus, just because an ad has a high ctr doesn't mean it will lead to conversions on the backend. Lots of time your best ad will not be the one with the highest ctr, but one in the middle, but which attracts the right type of person to click on it.

3. Following from the above, all that matters is EPC. Landing page ctr doesn't matter, ad ctr doesn't matter, EPC only.

4. With that being said though, landing page ctr is a great tool by which to block bad placements sooner. So you can use TheOptimizer to block any widget that has a really low landing page ctr. This is very important because it lets you block bad widgets in about 5$ of spend instead of $50-$100 if you wait to block based on conversions (presuming your offer is a Clickbank VSL with an average $50 payout and you were to block at 1x payout without a sale).

5. Don't worry about reinventing the wheel. There's a very, very established way of running on Revcontent that has worked for years on end... just read the old follow-alongs on here and copy them.

6. Same with ads. You don't want to start with 128 ads... start with like 9-12. Take two headlines you see working on spytools plus one of your own making, then take three images you see on spytools plus one of your own making/choosing. That gives you 12 variations (3 headlines*4 images). Then once you find the best 1-3 of those, add another 6-8 ads. Then whittle those down. Then add another 6-8.

7. 95% of the camps you start will not work. So you need to limit your losses on the ones that don't. I would not go more than a couple hundred bucks in the red at most on a brand new campaigns you've never run before. Some people would disagree, but its so rare to find something that works that you really, really need to limit your losses. You can always try that offer again in the future, but if it isn't popping right off the bat it rarely works anyway in my opinion.

Anyway you're obviously a super smart dude... do you run lots of other paid traffic already? I'm sure we'd all love to hear about your background and glean any wisdom we can from your experience

Welcome to STM!

07-27-2021 05:03 PM #2 kweens (Member)

Originally Posted by jack_l

2. This is going to sound crazy to you, but I would quit tracking ad ctr entirely. I understand that on VoluumDSP its important because your paying CPM, but once you start running on Revcontent itself, its just not a variable that can- by itself- help you optimize campaigns. The reason is because its so dependent on WHICH SITES Revcontent is showing your ads. A site with one three ad widget will have super high ctr's. A site with four 8 ad widgets will have super low ctr's. And for some reason Revcontent likes to show certain ads on certain sites, etc. Plus, just because an ad has a high ctr doesn't mean it will lead to conversions on the backend. Lots of time your best ad will not be the one with the highest ctr, but one in the middle, but which attracts the right type of person to click on it.

Interesting tidbit on the number of ads on one page. It definitely makes sense, and I've done some analysis in my spy tool data to try and see if there's a pattern. There are so few clicks from the ad to the landing page, that there's about a 60% correlation between ad density (ads on page) vs CTR. Not enough to be significant, but definitely leaning in that direction.

Originally Posted by jack_l

3. Following from the above, all that matters is EPC. Landing page ctr doesn't matter, ad ctr doesn't matter, EPC only.

Going by that perspective, the effective EPC for the flow that generated a conversion is about $50 from the landing page to the offer.

It took two clicks to the offer page to get the conversion.

Confidence interval: [0.0945312 , 0.905469].
Alternatively: 0.500000 ± 0.405469

Or about 1-10 hits to the landing page to get to the offer page, or a needed CPC of $5-50.

At $5CPC and $0.65CPM, I would need a minimum 0.7% CTR (5 / 0.00065).

And I'm currently at 0.04%. So need ad performance to jump 17.5x for this campaign to break even. Doable? Maybe. I have 3 landing pages in rotation, so that's an easy 3x if I can settle on a winner.

Originally Posted by jack_l

4. With that being said though, landing page ctr is a great tool by which to block bad placements sooner. So you can use TheOptimizer to block any widget that has a really low landing page ctr. This is very important because it lets you block bad widgets in about 5$ of spend instead of $50-$100 if you wait to block based on conversions (presuming your offer is a Clickbank VSL with an average $50 payout and you were to block at 1x payout without a sale).

I do have site-level rules blocking at about 120k impressions. The point beyond which the campaign would consistently lose money.

Originally Posted by jack_l

5. Don't worry about reinventing the wheel. There's a very, very established way of running on Revcontent that has worked for years on end... just read the old follow-alongs on here and copy them.

While I agree there's value in following the established path, what becomes the differentiating factor, or edge?

Originally Posted by jack_l

6. Same with ads. You don't want to start with 128 ads... start with like 9-12. Take two headlines you see working on spytools plus one of your own making, then take three images you see on spytools plus one of your own making/choosing. That gives you 12 variations (3 headlines*4 images). Then once you find the best 1-3 of those, add another 6-8 ads. Then whittle those down. Then add another 6-8.

Duly noted. I have 126 ways not to make an ad--2 of them are pseudo promising.

Originally Posted by jack_l

7. 95% of the camps you start will not work. So you need to limit your losses on the ones that don't. I would not go more than a couple hundred bucks in the red at most on a brand new campaigns you've never run before. Some people would disagree, but its so rare to find something that works that you really, really need to limit your losses. You can always try that offer again in the future, but if it isn't popping right off the bat it rarely works anyway in my opinion.

Anyway you're obviously a super smart dude... do you run lots of other paid traffic already? I'm sure we'd all love to hear about your background and glean any wisdom we can from your experience

Welcome to STM!

Couple hundred in the red is arguably too littl e data for something that pays out $50CPA+.

Part of what I'm seriously considering is tailoring an ad to an individual site and having single-site campaigns. RevContent has a surprising number of conservative leaning sites. Would allow for a narrower angle with a medium amount of traffic (combined 2M uniques/month)

07-27-2021 08:54 PM #3 jack_l (Veteran Member)

Originally Posted by kweens

There's one distinct advantage VoluumDSP has over Revcontent.

A true whitelist.

You could try to simulate a whitelist by setting the campaign bid low, and adjusting the desired publisher-level modifiers to the maximum. That starts to get complicated fast--and I haven't even made money yet.

I very purposefully don't buy traffic from sites outside of the top 1M (according to alexa) for the simple reason that the scaling opportunity is low.

BuiltWith suggests that RevContent has 23k websites the true number is very likely higher, too many for a blacklist.

Of the sites in the Alexa 1M that I checked, there are 1,022 with the revcontent footprint I'm using, too large for a bid-adjusted whitelist.

And that's why I'm pursuing VoluumDSP

Interesting tidbit on the number of ads on one page. It definitely makes sense, and I've done some analysis in my spy tool data to try and see if there's a pattern. There are so few clicks from the ad to the landing page, that there's about a 60% correlation between ad density (ads on page) vs CTR. Not enough to be significant, but definitely leaning in that direction.

Going by that perspective, the effective EPC for the flow that generated a conversion is about $50 from the landing page to the offer.

It took two clicks to the offer page to get the conversion.

Confidence interval: [0.0945312 , 0.905469].
Alternatively: 0.500000 ± 0.405469

Or about 1-10 hits to the landing page to get to the offer page, or a needed CPC of $5-50.

At $5CPC and $0.65CPM, I would need a minimum 0.7% CTR (5 / 0.00065).

And I'm currently at 0.04%. So need ad performance to jump 17.5x for this campaign to break even. Doable? Maybe. I have 3 landing pages in rotation, so that's an easy 3x if I can settle on a winner.

I do have site-level rules blocking at about 120k impressions. The point beyond which the campaign would consistently lose money.

While I agree there's value in following the established path, what becomes the differentiating factor, or edge?

Duly noted. I have 126 ways not to make an ad--2 of them are pseudo promising.

Couple hundred in the red is arguably too littl e data for something that pays out $50CPA+.

Part of what I'm seriously considering is tailoring an ad to an individual site and having single-site campaigns. RevContent has a surprising number of conservative leaning sites. Would allow for a narrower angle with a medium amount of traffic (combined 2M uniques/month)

Cool, well I would be absolutely thrilled to see a good successful Voluum DSP follow-along so I hope you kill it and see lots of green!

Edit: Couple quick additions... re: the differentiating factor, in my experience its creativity... finding better ads/angles/copywriting on advertorials... and then re: whitelists, I would be remiss if I didn't plug @platinum's wonderful tool TheOptimizer again... I would definitely give it a try if you do run direct on Revc..... and then finally, there's a great Voluum DSP follow-along by TheDudeAbides on here from a year or two ago... might be some valuable tidbits in there for you if your set on running through it vs direct... also I would look up all of the above-mentioned @platinum's threads - he has worked heavy in the trenches on all the native networks' API's and understands the tech/algo's like the back of his hand... would probably be great reading for a big-brained data guy like yourself (a lot of it goes over my head to be honest!).

08-03-2021 10:47 PM #4 kweens (Member)

I'm about 281 ads in, and they still suck.

Link to Ads

To explain the columns:

Impressions: The number of impressions received
Unique Visits: The number of unique clicks from the ad to the landing page
Conversions: The number of conversions
current ctr: The actual CTR of these ads
lower_ci: The level where there's a 97.5% chance the CTR is higher.
upper_ci: The level where there's a 97.5% chance the CTR is lower.

Specifically regarding these ad creatives, how can I make them suck less?
Conversions would be awesome, but for now, I'd settle for clicks.

Pretty sure the images aren't scroll-stoppers. However, they're visually similar to the ads that show up most frequently in the spy tools. It's plausible that because they're visually similar, that the users are banner-blind. How then can I innovate?

I've experimented with different border thicknesses--didn't seem to do much.

I looked at visually dim pictures and brightened them up slightly. Had a marginal impact on the CTR.

The ad text are optimized using this website as well as ripped from the spy tools.

How else can I move the needle on these ads?

08-03-2021 11:07 PM #5 matuloo (Legendary Moderator)

Conversions would be awesome, but for now, I'd settle for clicks.

Even though I understand what you mean by this, I wouldnt do that. Clicks mean shit when they don't convert. And that's exactly what some ads are about to do.. they will pull the clicks, but they wont convert. Think about all the genitalia-like ads, similar to the third one in your list... I'm sure they get lot's of clicks because of the WTF factor, but the conversions might lack.

I see you are using the same ad copy with all the images, do not underestimate this part, the right copy can have huge impact on the CTR and CVR.

Here is a tip that helped me in the past: try to zoom in a bit more, make sure you use very high resolution photos so even the closeup part that you crop out is still very clear. Higher contrast and some more filters can help too.

Try to focus on other things than just the magical substance/fruit/vegetable that will help them. Think about what your audience might be familiar with. Diabetes... they measure the blood sugar, right? They surely know what a meassuring device look like. Show them two images like this, one with higher (before) one with lower (after) value :

Know what I mean? Your audience is different from the rest because of... and they use things that others dont... they have problems others dont... try to focus on this in the images and adcopy. Do not just copy what is out there.

08-03-2021 11:34 PM #6 jack_l (Veteran Member)

Originally Posted by kweens

I'm about 281 ads in, and they still suck.

Link to Ads

To explain the columns:

Impressions: The number of impressions received
Unique Visits: The number of unique clicks from the ad to the landing page
Conversions: The number of conversions
current ctr: The actual CTR of these ads
lower_ci: The level where there's a 97.5% chance the CTR is higher.
upper_ci: The level where there's a 97.5% chance the CTR is lower.

Those ads are all fine man... any of those should be able to get conversions... its possible the ctr on them may not be quite as high as a few very specific images of eggs and cranberry sauce and etc... but the difference isn't profound enough that it should be make too much of a difference, and certainly not on the backend re: conversion rate... all those ones on the screenshot are fine... if you're not getting clicks or conversions its not the ads that's the problem... you've gone above and beyond in finding and testing great images, and with a proven headline, so I would cross that variable off the list for sure...

Not to be a broken record... but why not split-test Voluum DSP against Revcontent direct? Spend a couple hundred on each and see which one gives you better results

Again, I don't have anything against Voluum DSP at all- and its entirely possible you'll pull a ScottyG and master a new traffic source for us in real time and do massive volume

but I think of all split-tests you could do, Voluum DSP vs Revc direct will be by far the most consequential one

Edit: Is so weird that its always Eggs and Cranberry Sauce too lol... like why not French Toast and Cream Cheese? Banana's and PB&J's? Bizarre!

08-04-2021 01:07 AM #7 kweens (Member)

Originally Posted by matuloo

Appreciate the new perspective. I'll look into glucose-measuring stuff and give that angle a spin.

Incidentally, there are 24 different titles. The ones in the screenshot happen to be the ones with the statistically highest potential CTRs.

[Latest Secret Research]: Start Immediately And Lower Your Blood Sugar Below 100
1 Bite Of This Drops Blood Sugar (Take Before Bed)
1 Bite Of This Drops Blood Sugar Immediately (Take Every Day Before Bed)
Anyone with Diabetes Should Watch This (They Keep This from You)
Anyone with Diabetes Should Watch This (What They Don't Tell You)
Anyone with High Blood Sugar Should Watch This (They Keep This from You)
Blood Sugar Drops Below 100 After Eating This Before Bed
Blood Sugar Falls Faster Below Normal Levels While Eating This Fruit Before Bed
Diabetics Should Watch This (What They Don't Tell You)
Discover the Ultimate Info Revealing How To Drop Your Blood Sugar Now Risk-Free
Doctors Admit They Don't Know How Fruit Drops Blood Sugar Below 100 Fast
Doctors Can't Explain Why This Fruit May Lower Your Blood Sugar Below 100
Doctors Stumped By This Fruit Which May Lower Your Blood Sugar Below 100
Doctors Stunned: Unexpected Fruit Drops Blood Sugar Below 100
Doctors Unaware: Fruit Segment Suddenly Drops High Blood Sugar Below 100
Drop Your Blood Sugar Fast And Put Diabetic Neuropathy In The Past
Drop Your Blood Sugar Fast and Stop Diabetic Neuropathy Now
Hold Your Breath Before You Watch Closely This Secret Blood Sugar System Now
Hold Your Breath Before You Watch This Blood Sugar Hack System Now
Md: Do This Immediately if You Have Diabetes (Watch)
People Lower Their Blood Sugar Everywhere Using This Weird Fruit
Put Diabetic Neuropathy In The Past And Drive Your Blood Sugar Down Fast
Stop Diabetic Neuropathy From Happening While Lowering Your Blood Sugar Quickly
The Horrifying Truth About Diabetes and Blood Sugar History

08-04-2021 01:09 AM #8 kweens (Member)

Originally Posted by jack_l

Again, I don't have anything against Voluum DSP at all- and its entirely possible you'll pull a ScottyG and master a new traffic source for us in real time and do massive volume

but I think of all split-tests you could do, Voluum DSP vs Revc direct will be by far the most consequential one

Edit: Is so weird that its always Eggs and Cranberry Sauce too lol... like why not French Toast and Cream Cheese? Banana's and PB&J's? Bizarre!

Appreciate your persistence, and you've made your point. Signed up for RevContent directly and waiting for a CC to get approved.

08-04-2021 02:04 AM #9 jack_l (Veteran Member)

Originally Posted by kweens

Appreciate your persistence, and you've made your point. Signed up for RevContent directly and waiting for a CC to get approved.

Cool. I hope you kill it on both in the long run man! And even if the DSP proves better, trying Revc direct will probably aid in your mastery of the DSP anyway, since you'll have better insights into the idiosyncrasies of the platform.

But yeah, if you do end up using TheOptimizer, try setting some rules to auto-block bad widgets based on lp ctr. If you don't use TheOptimizer, another trick is to start with a super low bid on an INTL scattershot campaign targeting a bunch of countries, so you can identify the bad widgets with really cheap cpc's instead of expensive ones. Then as you block those bad widgets you can raise your bid so you get more and more traffic from the US and from high-quality sites, etc.

Here's a screenshot of all my top performing widgets from Revc diabetes as of a year ago in case it helps at all: https://share.getcloudapp.com/geuA8EJL

08-04-2021 09:57 AM #10 matuloo (Legendary Moderator)

Originally Posted by jack_l

Here's a screenshot of all my top performing widgets from Revc diabetes as of a year ago in case it helps at all: https://share.getcloudapp.com/geuA8EJL

Oh wow, very generous! Thanks for being a great member of our community man, much appreciated

08-04-2021 03:29 PM #11 kweens (Member)

Does anyone happen to have a contact at Revcontent?

Three of my credit cards have been denied.

08-04-2021 04:46 PM #12 vortex (Senior Moderator)

Originally Posted by kweens

Does anyone happen to have a contact at Revcontent?

Three of my credit cards have been denied.

Do a search for "revcontent" on skype and you'll find several. Let me know if you don't get a response and I'll try to get someone for you.

Amy

Sent from my iPhone using STM Forums

08-04-2021 05:29 PM #13 vortex (Senior Moderator)

I too am mega-impressed by your home-grown spying tool! And also your use of statistics.

I'm not experienced in native, but here are two suggestions that may help:

1)Try to come up with unique ads instead of only testing ripped ads. Landing pages don't go out of fashion very quickly, but ads do.

Without getting that initial click, people won't get a chance to read your landing page -> offer page -> convert.

And the first thing that catches people's eye is often the ad image.

So if the image doesn't stop people from scrolling, nothing else can happen.

I've had SOME success testing ripped ads for native, but new images got better results.

2)Test more offers. How many have you tested? I know you said payout frequency was an issue for you, but if you've been testing the same offer and nothing gives after a while, it may be better use of time to test other offers. I haven't operated in this vertical, but google can provide lots of options: https://www.google.com/search?q=diabetes+affiliate

Then of course there are other great products on CB in other niches.

Best of luck with your campaigns!

Amy

08-05-2021 08:57 AM #14 jack_l (Veteran Member)

Originally Posted by kweens

Does anyone happen to have a contact at Revcontent?

Three of my credit cards have been denied.

Email ar@revcontent.com and support@revcontent.com on the same email and ask them for help.

Also a tip from the James Van Elswyck native course that has served me very well - make sure you email with these ad networks from a branded domain (johndoe@acmedigitalmediasolutions.com, etc) and you'll get 10x better service.

I didn't do that at first and then switched shortly in and the difference was night and day.

Edit: You can also ACH them money to fund the account with (and later get back if needed), but I'm sure they can help you get the cc issue resolved anyway.

Home > Paid Traffic Sources > Native