Should I Continue Or Stop? Why? (37)

Hi all,

Some disclaimer: I'm a newbie. I've just started in PPV (long story short: I took some courses on PPV before I came to STM)

I've some noob questions. Was trying to figure out if I should continue or stop my campaign.

Was looking at this particular thread: http://stmforum.com/forum/showthread...ill-Make-Money

Been running and collecting data for 7 days.

Following are the breakdown:

6/27/2014	Campaign xyz	428	$10.01 
6/28/2014	Campaign xyz	442	$10.02 
6/29/2014	Campaign xyz	447	$10.01 
6/30/2014	Campaign xyz	911	$20.01 
7/1/2014	Campaign xyz	682	$15.02 
7/2/2014	Campaign xyz	695	$15.01 
7/3/2014	Campaign xyz	694	$15.01

ID	Offer		Visitors	Cost	CPC	Conversions	CR	CPA	Revenue	P/L	ROI
0	Offer xyz	4060		$0.00 	$0.00 	15		0.37%	$0.00 	$21.00 	$21.00 	0.00%

What exactly are you tracking?

I presume LI lets you pass things like keyword/referrer or that you can use their conversion tracking to look at site performance.

You definitely need to pass something of the sort to CPV Lab to provide a bank of data from which you can make decisions.

For PPV, you want to tweak your own landers based on CPV Lab stats, and cull off low performing keywords/targets/etc at the traffic source end.

Thanks zeno.

Yes LI lets me pass the keyword / target.

I've done the dayparting part, since I did not see any conversion in the afternoon - midnight - morning.

So currently setting it from 5AM - 7PM (forget which day I set this, probably I should have run it a bit longer before dayparting it)


Interestingly, 2 words keywords keeps getting the conversion, domain.com keyword did not do anything.

Is it time to cull off those low performing keyword?

Or I just scrap this campaign anyway (if the math is not looking good) --> which I don't know how to correlate all the numbers :|

-----

OK, so I took some numbers, and hopefully my assumption here is correct.

CTR    = 100% -> Direct linking (should I use other number?)
CR     = 0.37%
Payout = $1.4

CPC = Take average CPV = $0.022 (?) -> Can I assume using average of CPV?
EPC = $1.4 x 0.37% / 100
    = $0.005439

Profit = ($0.005439 - $0.022) x $95.09 / $0.022 -> (CPC)
       = $-927.4

ROI = ($0.005439 - $0.022) x 100 / $0.022
    = -75.28%

CPM = $0.022 x 100% x 10
    = $22
eCPM = 100% (Direct Link) x 0.37% x $1.4 / 10
     = $5.18

Profit = ($5.18 - $22) x $95.09 / $22
       = $-72.7

ROI = ($5.18 - $22) x 100 / $22.
    = -76.46%

CPA = $0.022 x 100 / 0.37
    = $5.95

Profit = ($1.4 - $5.95) x $95.09 / $5.95
       = -72.72

ROI = ($1.4 - $5.95) x 100 / $5.95
    = -76.47%

ROI = ($21 - $95.08) / $21
    = -78%

If in doubt there is one comparison only that you should make:

Cost of traffic for a particular keyword/factor VS Revenue generated by that particular keyword/factor

How much have you spent on this campaign overall, and how much have you made?

Can you see how much you have spent on individual keywords, and how much they have made?

Things like conversion rate are just percentages, they alone don't tell you where you stand so always follow the money.

Once you get a better handle on what's going on you can look at e.g. the CPV of specific keywords, which is related to your bid, and the EPV of those keywords - i.e. comparing costs and earnings.

This is definitely a situation where a tracking system will help - a place where you can look at the EPV of every keyword, the traffic, conversions, etc. and also use an average CPV to give rough profit/loss and ROI values.


P.S. I'm not sure what you did in the CPC part but your profit was definitely incorrect! You'd get a virtual slap from me if you have spent almost $1000 and made ~$20!!

Thanks for the pointer!

Originally Posted by zeno

If in doubt there is one comparison only that you should make:

Cost of traffic for a particular keyword/factor VS Revenue generated by that particular keyword/factor

How much have you spent on this campaign overall, and how much have you made?

Can you see how much you have spent on individual keywords, and how much they have made?

-69.46%
-62.79%
-84.01%
274.33%
----------
-68.69%

Originally Posted by zeno

P.S. I'm not sure what you did in the CPC part but your profit was definitely incorrect! You'd get a virtual slap from me if you have spent almost $1000 and made ~$20!!

For PPV, CPC = CPV assuming you want to calculate the equivalent of banner ad click cost.

Right now you are losing money.

That--by itself--is not a reason to stop.

It is a reason to do things differently, however, as if you just do more of the same, you will lose more money.

Good thing is that you have at least gotten some conversions.

Originally Posted by caurmen

For PPV, CPC = CPV assuming you want to calculate the equivalent of banner ad click cost.

Originally Posted by cmdeal

Right now you are losing money.

That--by itself--is not a reason to stop.

It is a reason to do things differently, ....
...
...

How much budget do you have left? I think that is a big consideration.

Originally Posted by Humbleaid

How much budget do you have left? I think that is a big consideration.

OK, so now I get the hang of how to analyze my data in CPVLab.

I declare that I learnt my first expensive lessons :| *ouch*

These are some data from my other campaign:

1. Go to your campaign report, and click Optimize Report


2. Here you can see all your "unique" targets along with the views and coversions data. Extract this to CSV


3. Open up your Excel, and add some parameter (see yellow highlighted)


From here I can see where my money bleeds :| It's very obvious now that I should cut those, since even though it converts, it just doesn't worth it.

Unfortunately I realized this a bit late.

So hopefully this can serve as a lessons for other newbies around here as well.

Always keep learning!


----------
Questions:

My next questions to the seniors here, should I try to get some more data on the top 3 targets/keywords?
Or should I just scrap this campaign all together?



.

Can you upload those images to an external services like imgur? unfortunately they've come in a bit small.

(STILL not sure why the forum does this - having another look at it tomorrow. Apologies.)

Hello all,

It's been ~2 weeks I've launch this campaign, and it has been a very good learning experience for me.

This is my latest stats from the first campaign:



I've switched from DL to LP, and currently identified 4 that gave me quite a good CTR and Conversions

Data from CPVLab:


Those highlighted in orange/peach are inactive / paused.

Looking at the LPs stats, would you recommend to collect more data?
How would you judge which LP is best performing?
Is it only based on the CTR and CR?

Appreciate all your feedbacks!

Originally Posted by qureyoon

Hello all,

It's been ~2 weeks I've launch this campaign, and it has been a very good learning experience for me.

This is my latest stats from the first campaign:



I've switched from DL to LP, and currently identified 4 that gave me quite a good CTR and Conversions

Data from CPVLab:


Those highlighted in orange/peach are inactive / paused.

Looking at the LPs stats, would you recommend to collect more data?
How would you judge which LP is best performing?
Is it only based on the CTR and CR?

Appreciate all your feedbacks!

Originally Posted by cmdeal

You should post real data (real costs + real revenues) for each placement. It is impossible to give any real business advice without knowing how much it cost you to buy something and how much you were able to sell this for.

Your goal is not high CTR or high CVR. Your goal is high profit.

Originally Posted by qureyoon

What should I look for to optimize, and make it high profit?

Updated the table (added the numbers for Cost, CPV and ROI column):

Cost, each views is $0.02 (I take average of CPV).



Is there any other numbers I should look for?

Have you run these numbers through a statistical significance calculator or split-test calculator? There's one in Part 11 of the Getting Started Guide.

Looking over those stats, I strongly suspect you're pausing landers based on too little data.

What's the payout on this offer?

Originally Posted by caurmen

Have you run these numbers through a statistical significance calculator or split-test calculator? There's one in Part 11 of the Getting Started Guide.

Looking over those stats, I strongly suspect you're pausing landers based on too little data.

What's the payout on this offer?

Pausing A & B will help you lose less money.

But if you want to make money, you need to get your revenue per conversion to be higher than your cost per conversion.

Originally Posted by cmdeal

Pausing A & B will help you lose less money.

But if you want to make money, you need to get your revenue per conversion to be higher than your cost per conversion.

Originally Posted by qureyoon

However I hope that I can achieve something from this campaign, but figuring out how to get my revenue higher.

Woah there! You need a bit more conversion data. Its most certainly a premature jump to say A or B are best with only 1 and 2 conversions...

Originally Posted by cmdeal

Focus on your CVR. Two ways this can be done is by tweaking your landing page and (much) better targeting your campaign.

Originally Posted by zeno

Woah there! You need a bit more conversion data. Its most certainly a premature jump to say A or B are best with only 1 and 2 conversions...

Originally Posted by caurmen

Now, you need to know how and when to choose from our split-test landing pages and offers.

Once you've got this down, you have everything you need to optimise your campaigns into profit!

Choosing Which Landing Page To Run

If your landing pages don't have at least 5 conversions between them, don't start this process - you don't have enough data.

If they do, load up this calculator.
....
....
....

There is no need to send equal traffic to each one. Cut losers fast.

@qureyoon - You're using the calculator right, but I'd usually only pause a lander that has a less than 5% chance of winning according to the calculator. Others I'd let run to get more statistical certainty.

BTW, different people have slightly different philosophies on split-testing and the best time to cut, which is why Zeno, cmdeal and myself may advise subtly different paths to take here. All three of our approaches work, but they prioritise different things.

Thanks for the advise!

I'm learning, and adapting to my own style. Still need to find the right balance for myself on when to cut losers, and how much data I need to collect.

Keep the advise coming! It's very helpful, and hopefully will help others reading it too!

Awesome environment here!

There are at least two different objectives in split testing:

1. The find the right answer
2. To make the most money with lowest risk

These approaches converge if there is no cost to testing, but when testing comes at a cost, they diverge significantly.

Take the example of designing a car.

Originally Posted by caurmen

@qureyoon - You're using the calculator right, but I'd usually only pause a lander that has a less than 5% chance of winning according to the calculator. Others I'd let run to get more statistical certainty.

BTW, different people have slightly different philosophies on split-testing and the best time to cut, which is why Zeno, cmdeal and myself may advise subtly different paths to take here. All three of our approaches work, but they prioritise different things.

Originally Posted by zeno

I have one hard fast rule that I abide by, at least for low payout offers: if adding one conversion to a combination would bring it well out of the 'loser bracket', then I get more data.
...
...

Hello guys,

Time for some update again

What have been done so far:

• Target more placement/keyword
• Setting up multivariate test based on this guide: http://stmforum.com/forum/showthread...est-in-CPV-Lab

OK, you're not going to like this answer, I'm afraid, but unless you have a very solid reason for taking this approach to multivariate testing and can explain how you're avoiding interaction problems I'd strongly recommend not going down this road.

Conventional multivariate testing does not mean testing multiple things on the same page and treating them all as independent split tests. That's a parallel split test. Whilst it can work (I've used it) , you need to have a solid understanding of the maths behind multivariate testing and understand the failure modes there. (Mostly, they have to do with the test being invalidated by interactions between your testing elements.)

You don't have nearly enough volume to do full multivariate - you'd literally have to increase the volume you're running by 1,000 to stand a chance of getting statistically significant results. Full factorial multivariate testing involves testing every single combination of every element - which in your case would be 7x9x3 = 189 possible combinations. Even optimising for CTR you'd need a few thousand clicks to get significant results there, and if you were optimising for CVR you'd need a few thousand conversions!

You could do fractional factorial to achieve the results you're looking for, but that either requires a tool which does that (I know of exactly one, although it'll be two soon ) or you to have some seriously advanced math skills.

In addition, you appear to be optimising for CTR. I'd strongly recommend not doing that. CVR is what makes you money, CTR's just a useful indicator.

Honestly, I'd recommend continuing to test individual non-multivariate landers. Test completely different styles against each other, cut losers with the split-test calculator, and then narrow down to smaller elements.

A parallel split test can work very well - it's great to quickly get some ideas of what might work and generate some combinations to test - but given you're still reasonably new to running campaigns, I'd recommend mastering the basics before jumping into more advanced techniques.

Thanks a lot for the feedback and advise!

My initial thought as you have guessed it correctly is to increase CTR, with the understanding that CVR will likely to go up as well.
That is why I tested a multivariate thing, by testing different headline text, images and CTA text.

But now you have pointed it, it seems to be a wrong approach. Thanks!

I will revert to what you have suggested:

Originally Posted by caurmen

....
Honestly, I'd recommend continuing to test individual non-multivariate landers. Test completely different styles against each other, cut losers with the split-test calculator, and then narrow down to smaller elements.
....

one thing that caught my eye: seems like you are calculating ROI the wrong way, there's no ROI < -100 %.

ROI = (Revenue - Cost) / (Cost)

Originally Posted by andyvon

one thing that caught my eye: seems like you are calculating ROI the wrong way, there's no ROI < -100 %.
ROI = (Revenue - Cost) / (Cost)

On another note. While searching for more keyword/target, I've come across that the offers have been receiving a lot of complaints:

• http://www.complaintsboard.com
• http://www.reviewstalk.com/complaints-reviews

• Do more thorough keyword/target research
• Create different styles of LPs, rather than just playing with different elements within the similar layout.
• Always track your cost vs your revenue, and cut the losers fast!
• Understanding basic math for profitability, CTR, CR, P&L, ROI, eCPM etc.
• Understanding simple A/B test to help cutting out losers.
• Test the main thing first, angles, LP styles. No need to test everything and get everything right! It's not about finding the perfect angle/LP, it should be getting it to profitable!
• Don't try to make it profitable, if at the beginning the data shows that the offers doesn't not convert that well, just move on!
• Don't get to attached to one campaign, and try to make it break even or even profitable, just because you have spent a lot of time in it.

Originally Posted by qureyoon

I don't like having people get scammed out of their hard earned money.

So I'm stopping this campaign.