Home > > Newbie Follow-Alongs

---

Telebind123's 40 day follow along (15)

---

Hey,

So to consolidate my noob questions a bit I've decided to make a follow along based on Amy's 40 day pop guide (thank you so much for the guide!!!).

Right now I am able to give this a serious go so I'll be updating this follow along pretty frequently unless some major adverse action befalls me.

My goal here is to master the pop traffic source and scale it to the best of my ability. I also need the knowledge of how to conduct a media buying operation such as this in order to move onto harder and more expensive traffic sources. Right now and for the foreseeable future I don't really expect to make a cent's worth of profit. Hopefully I can get -70% ROI steadily for a month or two so I can at least get some of my money back.

Hopefully this helps some members who're going through the same difficulties as I am. The more data the better for all of us.

Alright, thanks for reading. I'll be updating you guys shortly!

Thanks!

---

So I launched a sweeps campaign and found the best lander among 9 landers.


The offer that I used to find the best lander had 2 conversions while the rest (4 offers) had 0 at the time I was testing out landers and offers.


So upon identifying the best lander and the offer I was testing with that lander, I ran some more traffic to it.


I also split tested with a lower bid and the lower bid had a slightly higher ROI.


Here are some stats so that we're on the same page:


Higher bid camp metrics : (bid 1.68$CPM on Propeller SmartCPM)

Lower bid camp's metrics: (bid 1.14$CPM on Propeller SmartCPM)


So I think I'm gonna go with the lower bid camp for a bit now. The geo is pretty big so I anticipate after some placement cutting I should get a better ROI whilst maintaining reasonable volume.

---

Hello telebind123, starting a follow along was a great idea indeed!

Yup, looks like you actually had better performance with the lower bid, so no reason to over-pay at this stage.

I wouldn't rush to chose the "best offer" and "best lander" at the same time. Start with just one.

The reason is simple ... maybe some of the offers would convert with some of the landers, but since you had quite a lot of LPs in rotation, chances are that many of the combinations simply didn't get enough traffic to prove themselves. Then there is BOT traffic, which is simply guaranteed to be in the mix. So maybe some offers got more bots than the others, which obviously means they couldn't convert as good. And the same goes for LPs, not just offers.

I don't see your data, so can't tell which one of the two shows stronger "evidence" for being the best... but you do, so make the choice. Where do you see stronger data when it comes to selecting the winners? Is it the offers or the LPs? Choose the one where you are more confident and keep testing with all (or at least multiple) variations of the other. Give it some more $$$ in traffic to confirm whether the trends prevail. After all, you made you decision on 2 conversions only, that's really not enough.

---

Hey, thanks for the feedback, appreciate it quite a bit!

So I think the stronger data for selecting the winners lies in the LPs so I'll use choose the winning LP to test out more offers.

Thanks for the tip!

---

So I think I've hit a bit of a problem here.

Yesterday, I found the best LP and then I proceeded to create variants of that LP. So I changed some of the text around to see if I could get some better results from my winning LP.

On the base of the winning LP, I created 2 variants, WinLPVar1 and WinLPVar2.

Here are some stats :

So I spent around 20 bucks to test variants. If I put the landers in https://www.peakconversion.com/2012/...al-calculator/ I get:

WinLPVar1 has 38% chance of being best, WinLPBASE has 19% chance of being best and WinLPVar2 has 43% chance of being best.

If I compare the best and the worst, WinLPBASE to WinLPVar2, I get that WinLPVar2 has 64% chance of being best and WinLPBASE has 36% chance of being best.

So nothing really conclusive here. I spent around 20$, should I be spending more in order to get conclusive data?

How do you guys usually test landing page variants? How much should I be spending to test landing page variants?

I do want to do an additional variation with both images and text this time but at this rate I might be spending too much in order to see data.

Also, based on this data, should I proceed with testing more offers with one of the variants? Do you think there is enough conclusive evidence to show that one of the variants is better than the base?

---

It's still not enough conversions to pick a winner with confidence ... 2 or 3 leads is not much and then there is the fact that all the LPs converted on a very similar level : 3-2-2 Even though one of the LPs got less clicks so those 2 conversions are "worth more".

But ok, let's drop the BASE LP and keep the other 2. Next step would be to try and find an offer that would perform better with these LPs.

---

Alright yep planned on using one of the variants for testing offers but then got my lp banned. So I'll buy another domain today and test out the offers and post here.

Thanks!

---

Alright here is the update for today:

So I tested out 6 offers.

I primarily used this calculator here : https://www.peakconversion.com/2012/...al-calculator/ to determine which offer was my winner. Right now its saying that my best offer has 84% chance of being best which I guess I'll go with.

Any suggestions? Any mistakes I made? I know the rule of thumb is to wait 'till 10x payout to cut the bottom 3 offers but the statistical calculator at the time was telling me to cut it so I cut at 5x payout. At the time I cut the bottom three offers, my best offer had 2 conversions and the bottom three all had 0 at about 10k impressions which evaluated to the bottom three offers having a 7% chance of being best and my best offer having 93% chance being best. Other two offers at the time had 1 conversion each. However, perhaps 2 conversions at 10k impressions just wasn't enough data.

Also, is this enough data to conclusively support that at 7 conversions, I have picked the best offer? Anything wrong with my method at this stage?

I now plan to proceed to placement cutting with my best offer.

---

Nice progress!

You did everything correctly.

When testing the variants: Results have not reached statistical significance yet. But if you want to pause the split-test and just use one of the landers to test more offers, you certainly could.

Very realistic expectations! A bit too humble even. You can certainly do a lot better than -70% ROI before the 2 months are over.


First of all: The stats calculator for comparing offers is NOT the one at peakconversion, it's the one here:

https://stmforum.com/forum/showthrea...211#post289211

Unless: Your offers all have the same payout (in which case you can use the one at peakconversion).

The 10x payout rule of thumb is just that - a rule of thumb. If the stats calculator says that something is less than 10% likely to be the best out of a batch, cut it.

The stats calculator already takes into account sample size (i.e. number of impressions), conversions and payouts.

The best offer+lander combo at this stage is at -65% ROI - that's not hopeless, but not looking very promising either. But we'll need to look at a couple things to get a better idea on whether it's worth it to focus on cutting now, or if you'll need to test more landers and offers:

-Drill down into the best lander+offer, into OS, browser, placements - to looks for green segments. It's a bit early to do this with only 7 conversions, but we may get lucky and see some positive indicators.

-Which geo is this? The more traffic there is available, the more you can afford to cut and STILL make profits on the remainder.


Basically it comes down to this: Do we feel that it's possible to take this campaign from -65% ROI to green, by cutting stuff? (Like placements and other traffic segments such as OSs and devices and browsers etc. etc.) AND with an acceptable amount of daily profits? (Even for the newest of newbies, $5-10/day would be the bare minimum - otherwise you may as well work at McD's.)

If the answer is YES, then focus on cutting until you reach profits. If the answer is NO, then either test more landers and offers (in the same or different vertical) so that you can make a larger portion of the traffic profitable, or test another geo altogether.

One major thing to check, is whether you have traffic segments that, either by itself or jointly with other segments, can give you enough daily profits. If you can't find such green segments, don't even bother to cut - because if there isn't gold buried somewhere in that pile of shit, then all the cutting will only leave you with shit.

That's it in a nutshell - we'll need to look at stats in detail to decide what would be a good next step.

Please feel free to post screenshots of any stats you feel may be interesting - don't be afraid to post lots.



Amy

---

Thank you very much for the feedback, I genuinely appreciate it.

I had a feeling I was missing something by not accounting for the payout. Should've looked at your signature!

The GEO is UK and the Payout for my best offer here is $1.21.

So, here are some stats for my best lander + offer combo
(Edit : I had to minify the images to 'Medium' size otherwise it would not load correctly. If I tried 'Large' size the image would cut off and if 'Full Size' the images would convert to attachements. However, luckily it is possible to click on the images for a nice full view.)
:

OS:


Browser:


Green Placements :


Placements > 1K in volume, sorted by Volume:




Devices :





So I'm feeling like the only parts that I can salvage here are some of the placements and the Unknown device type. The total spend for the green placements are $8.59, roughly 13% of the entire cost.

However, it is unnerving that the placements that are green have mostly only converted once. The 'Unknown' type is -14% ROI and accounts for 5% of the total cost.

For reference, the min ROI calculator right now says to 'KEEP RUNNING' :



However, I have spent way more than $10, so I decided to plug in a breakeven ROI :

and I get a KILL verdict.

So it seems that without any type of cutting this offer is doomed.

If I recall correctly, I was able to spend 70$ in the UK GEO in roughly an hour between the hours of 08:00 - 23:00. From 8AM to 11PM, that is 17 hours. 17 hours * 70$ = $1,190. The green placements, if they keep being green over the long run, should account for 10-13% of this cost which equals roughly 150$. The Unknown device type which had -14% ROI accounted for 5% of the cost so, extrapolating naively again, that would be 59$ in cost. I am not accounting for hours from 23:00 - 07:00 here as people would are asleep so conversions may be way below par and I can't make any assumptions here as I've purposely never run in those hours.

So if I am to run only the green placements it looks like I would be spending about $150 for a positive ROI. I am sure the Unknown device type's ROI would go up a bit too if I only ran with my green placements. But of course, the statistical significance of the green placements are not sufficient, i.e. only 1 conversion mostly for the green placements.

So I'm thinking of three things here:
1. Run a whitelist campaign based on green placements.
2. Run entire campaign and cut placements. But I am unsure if any of the placements that are red are salvageable. Not sure if I should spend more money doing this.
3. Run entire campaign and run the whitelist campaign. If any placements turn green in the entire campaign, I put it into the whitelist campaign. I do this following Amy's placement cutting rules and hope that I have enough statistically significant placements to have a green campaign. If its trending towards the red placements not being salvageable, cut the entire campaign and only run the whitelist campaign.

Any thoughts?

---

Alright quick update for today:

I ran the Whitelist campaign to $10 USD just to see what ROI I would get :



So a -58% ROI on the whitelist campaign was not enough to justify me running it some more.

Additionally, most large placements on my original campaign got > -70% ROI, with most having -100% ROI.

Even after running a whitelist campaign, the green placements mostly turned red. In addition to this, the spend per hour was rather low as well ($3 or so).

Here are some stats for the WL campaign :


In light of this information, I'll have to stop this campaign and move onto the next one. I learned quite a bit from this campaign so not a total loss.

I have landers for some geos and verticals already prepared so stay tuned!

---

I spent over $1000 to get PUSH to work for sweeps.

I can tell you a tip, blacklisting stuff is not the way to go. Instead, take your best landing page and on that landing page change the biggest picture (pic of the prize), also other major elements that jump out and split test a few variants... I used a simple php script to define the variables (img) and headline then call it by adding ?img=1&headline=2 and so on, and just added each variant as a lander into my red tracker account.

This can have a significant impact on your CR and bring you into positive territory on 80%+ of placements. And when you have this, then you can block... but slowly...

---

The OP is running POPs, not push traffic, which is two very different formats when it comes to the optimization.

But other than that, -70% campaign wont become green just by cutting placements, so I agree, more work on the LP or the whole flow is needed.

What you have now telebind123, should serve as a benchmark base, and the next goal is to find something that's gonna beat it. Keep in mind that POPs is an intrusive format, users don't want to see it, so you have to grab their attention with something they CANNOT refuse

---

Thank you very much for the tip!

I will definitely split test the images from now on using your method when I've identified my best LP. My best lander was the FB wheel spinner for iPhone X sweeps.

I changed some text around for variant testing but that didn't do much. Next time, I'll change the images first.

Thanks

---

Alright, I'll try my best!

---

Home > > Newbie Follow-Alongs