Home > >

---

How Do I Know When To Stop Running Ads? What Is "Statistical Significance"? (52)

---

So you've started a campaign and you've got some data in! It's time to start optimising your way to profit.

But first - how do you know when you're safe to cut an ad from your campaign?


Why You Should Care About "Statistical Significance"

If you flip a coin and it comes up "heads", does that mean the coin will always come up heads?

Of course not.

But one of the mistakes that new affiliates often make is assuming that, essentially, the coin's always going to come up heads.

Assuming you're looking for a 0.2% CVR (Conversion Rate), for example, many affiliates will drop any ad that's showing less than that - regardless of how many times it has been viewed.

That's a very bad idea. A banner which has no conversions after 500 views can easily soar to having a 0.4% or higher CVR after 5,000 views.

Fortunately, scientists have spent a lot of time working out just how many repetitions of an event you need to be reasonably sure it'll keep repeating at the same rate.

This is called "Statistical Significance".

But how do we calculate this?

Well, we can use the same statistical tools that scientists running medical trials or experiments in particle physics use to determine whether their results are valid. And that way, we can be as sure as they are that we're on the right track.

Don't worry - it's less scary than it sounds.

TIP: Don't use data that had large numbers of impressions on bad placements in it. If in your initial testing round you blacklisted placements that accounted for more than a third of the total traffic, only do these statistical tests using data that you collected AFTER you blacklisted those placements.



Using the Scary Math Calculator

Head to this page.

Now, let's take your first ad that's still running.

Go to the "Binomial Confidence Intervals" section. Put the number of conversions it's had into the " Numerator (x):" box.

Put the number of impressions it's had into the "Denominator (N): " box.

Hit "Compute".

You'll end up with something like this. This example assumes you've had 12 conversions from 20,000 impressions:



Can't understand a word? Don't worry. It's actually very simple, just written in math-ese rather than affiliate-ese.


What The Math Means

What we're trying to do here is predict the future Conversion Rate of your ad from the data we have so far. Using statistics, it's impossible to predict a single number with certainty.

Instead, the calculator tells us the range, from high to low, that your eventual Conversion Rate will be within.

The calculator calculates the maximum and minimum probable Conversion Rate for your ad: that's the "Exact Confidence Interval around Proportion:" bit.

The higher number is the maximum Conversion Rate the ad could have, and the lower number is the minimum. To convert them to percentages, multiply both of them by 100.

So, in the example above, you know (with 95% certainty) that your ad has a minimum possible conversion rate of 0.03% and a maximum conversion rate of 0.1%.

Great stuff! But what do you cut?



Calculating your Minimum Viable Conversion Rate

To know what you need to cut, you need to know what conversion rate you need to make a solid profit on this campaign.

You can calculate that from your Cost Per 1000 Impressions (CPM), your payout on a successful conversion, and your target Return On Investment.

Cost Per 1000 Impressions (CPM)

You will probably know this already by now. If your traffic source operates on CPC instead, use this formula:

Cost Per 1,000 Impressions (CPM) for this ad = CPC *10 * Clickthrough Rate (CTR) for this ad as a percentage

For example, if you're paying $0.10 per click and are getting a 0.1% CTR, your CPM is $0.10 * 10 * 0.1 = $0.10

Target Return On Investment

This is the amount of profit you want to make for every dollar you invest in this ad, as a percentage.

It's tempting to say that you just want to be above 0% - which is break-even - but is that really true?

Remember to factor in what your maximum spend a day is - if you can afford to spend $1,000 a day, even, a 2% ROI will only make you $20 that day.

Equally, it's tempting to aim for 200% ROI - but that target means you'll cut a lot more ads, and spend a lot more to get there, if it's even possible. This sort of ROI is more possible on niche campaigns - but on broad campaigns, it's very hard to achieve from ads alone.

Generally, I'd recommend aiming for an ROI of around 40% on a broad campaign with a lot of scope to scale up (a broad adult campaign on a major site, say). For a niche campaign (a targeted campaign on POF, for example), aim for more like 80% ROI.

You can always push these numbers up later with more testing.

Minimum Viable Conversion Rate

Now, you need to figure out what your minimum useful conversion rate is, from these numbers.

The calculation for this is a bit complicated, so I've made a spreadsheet you can use rather than calculating it by hand.

Get that spreadsheet here. Put in the numbers above, and it'll spit out your minimum viable conversion rate!


Choosing What Ads To Keep

Now, simply check whether each ad's calcuated maximum possible conversion rate is higher than your Minimum Viable Conversion Rate.

If it's not, kill it.

If it is, keep it running.

You can do this at any point whilst running ads, but you shouldn't do this too often. Every time you test, there's a chance that your calculation is wrong. The chance is very small, but it multiplies by the number of times you test.

As a very rough rule of thumb, you should check your ads about every time they've spent 5x your offer payout.


Saving More Money: Minimum CTR

If you're on a tight budget, you want to take every opportunity to save money on your campaigns.

To do that, you can also cut ads based on their likely CTR - Click-Through Rate.

Now, you should normally never cut ads based on CTR alone. A high CTR can easily translate into a low CVR.

But there comes a point where CTR is important. If your CTR is so low that your CPC - Cost Per Click - is so high that you'd have to have a miraculous conversion rate to make a profit, you can cut any ad with maximum predicted CTR of that or below.

To figure this out, divide your CPM in dollars by your payout in dollars. Then divide the resulting number by 5. That gives you the minimum CTR that would be profitable assuming that your CVR was 50% - which is wildly higher than you'll ever see in the real world under most circumstances.

Now, calculate your maximum expected CTR for any ad you wish to test, just the same way as you calculate minimum expected CVR, above - only in the "Numerator" box, put number of clicks rather than number of conversions.

If maximum expected CTR is below your minimum profitable CTR, you can safely cut that ad before it has enough data to do the conversion rate test.

Again, don't run this test too often - but you can safely run it every time an ad spends 1x your payout, up to about 5x the payout, at which point you should stop testing on CTR and start testing on CVR.

Further Reading

If you're still a bit confused after reading this guide, or would just like another take on the process, Vortex's excellent series on cutting banners, placements and landers is a must-read:

• http://stmforum.com/forum/showthread...Banners-Part-1
• http://stmforum.com/forum/showthread...Banners-Part-2

Let me know if this lesson has been understandable! I know there's a lot of math in here: if there's a part that confuses you, please do post below and I'll clear it up.

---

Hi caurmen, nice writeup and easy to follow.

Biggest question I have when trying to decide is how much statistics is needed to be significant.

Do you decide on number of impressions? spend? spend based on payout?

---

@doc_ufo - The section "Using The Scary Math Calculator" answers that question. Take the numbers you have - impressions and conversions - plug them in there, and you'll be able to tell whether you're at significance yet.

---

I think doc_ufo meant when should he run the test. And I think your answer is roughly when the ad spend has reached 5x the payout. Am I misinterpreting it?

Thank you for this amazing post! This is going to help save money.

I do have a question though: if my particular ad's spend has reached 5x the payout and there hasn't been any conversions (whether with an LP or direct link) whatsoever, should I bother running either the CVR or CTR test or would I be better off just killing that particular ad without running the tests?

---

Another thing I noticed... why is it when you increase the targeted ROI, the minimum viable conversion rate decreases?

I would think if one wants more ROI that means they need to make more money thus more conversions therefore increasing the conversion rate that you need to achieve the high ROI.

Am I understanding this thing incorrectly?

---

@dtalexone - I'd recommend running the calculation no matter what. It doesn't take long, and there are SOME circumstances (albeit few and far between) where you'd want to keep a 0-conversion ad with 5x payout.

I'd say run this test before 5x payout. 3x payout is a decent time to run it for the first time.

As for minimum CVR decreasing when you increase target ROI - that definitely shouldn't happen, and when I just tested it, it didn't. Could you try reloading the spreadsheet and checking that again? Sounds like something weird is going on... Let me know if it keeps happening and we can bugfix from there!

---

The spreadsheet cannot be edited; it's only in view access. I had to take the formula and copy it over to my own spreadsheet. I assume I messed something up.

---

Click File in the top left, then Create copy, et voilà

---

Awesome. Worked like a charm. And yes, the calculator is working as it should.

---

Thanks for this great post - I'm a little confused though:

'divide your CPM in dollars by your payout in dollars. Then divide the resulting number by 5'

So how can I do this CTR cull at 1 x payout, if I havn't got the CPM value until after 2-5 x payout?

---

@jaydenuk - CPM is the amount of money it costs you to buy 1,000 ad impressions. You'll either know that because you're bidding in CPM on your traffic source, or you can calculate it from your CPC (Cost Per Click) - see http://stmforum.com/forum/showthread...Will-Make-Mone

I hope that clears things up! Let me know if you need further info or I've misunderstood the question.

---

What I mean is, I'm not going to know my CPM from my CPC until I've gone past 2-5 x offer payout - therefore, how can I perform the CTR cull at 1 x payout if I shouldn't get the CPM until 2 x payout?

---

@jaydenuk - You should know your CPM considerably before then. What traffic source are you using?

---

It's ok - something just clicked - I get it now - thanks. I'm using Adwords.

---

@caurmen
Hi, I'm a newbie and have a few question after read this thread

1. For this sentence

"Now, calculate your maximum expected CTR for any ad you wish to test, just the same way as you calculate minimum expected CVR, above - only in the "Numerator" box, put number of clicks rather than number of conversions. "

as you said "same way as you calculate minimum expected CVR"
so to calculate "Max expected CTR" , I should use this data (red line on img) x 100 to percentage and compare with Min. CTR right?



-------------------------------------------

2. I use Prosper202 + F5 that didn't show "impression",
How to find this data to calculate?



THANKS

---

@hangman -

1: You're 99% correct. The only mistakes is that the maximum expected CVR is actually the higher number (0.0010) not the lower one.

2: Your traffic source should show you how many impressions - views - each ad has had.

Hope that helps!

---

How About If you want to Split Test Landing Pages Against each other?

---

@edric2233 - It's best to use a completely different process there. I've written it up in some detail over at http://stmforum.com/forum/showthread...ugh-Clicks-Yet

---

Guys,

The Scary Math calculator link is no longer working.

Does anyone else have another link?

UPDATE: It's move here: http://statpages.info/confint.html

---

Thanks, updating.

---

Just a quick question - I should only be comparing the higer number from the 'Exact Confidence Interval around Proportion' value to the minmum viable value, right?

Hence, I don't take in to account the lower of the 2 values of the 'Exact Confidence Interval around Proportion' calculator.

I'm glad I read this entire post again - I've just realised I have been calculating the conversion calculation by clicks and conversions and not conversions & impressions!

---

@jaydenuk - yep, that's correct! And glad the post helped!

---

Hi caurmen,

Sorry I am a bit new to this. Can you share with me what do you mean what u mention 1xpayout or 5xpayout?

Thanks

---

@evillucia - No problem at all!

What I mean is that you multiply your payout - the amount you get paid by the affiliate network for a single conversion - by the number I specify.

So if you're running an antivirus PIN submit with a $2 payment to you for each conversion, and I specify 5x payout, then that's $10 = 5 x $2.

Does that make sense?

---

I'm currently running pop traffic, but how is the calculator in this post, different or more beneficial from the PeakConversions calculator?

Your How do i know which landing page is working best post says to use this calculation for POP traffic - why?

---

Just realised I posted in the Archived version of a similar thread anyways reposted here.

I have some questions about this method. I'm having some trouble getting my head round this as I'm using CPC and not sure if the values i'm using or my min Viable C.R. calculation is correct.

Would appreciate some help.

Note: According to Zeno's post, I should use Banner Clicks & Conversions if I want to optimize banner conversions.

Sample numbers:

- Impressions = 500,000
- Clicks = 1500
- CTR = 0.30%
- CPC = $0.15
- CPM = $0.45
- Conversions = 10

Using the stats calculator (assuming I should be using clicks as the Denominator), I've calculated
- Min Possible C.R. = 0.32%
- Max Possible C.R. = 1.22%

Minimum viable conversion rate calculation
- CPM = $0.45
- Payout = $30
- Min ROI = 50%

Min Viable C.R. = 0.00225%

Outcome
Since my Max Possible C.R. is (significantly) higher than Min Viable C.R., I would keep this banner.

I'm doubting my method because of the order of magnitude in difference between my Max C.R. and Min Viable C.R.

Questions:
1. Is my method flawed in using Clicks & Conversions (instead of Impressions & Conversions)?
2. If not, then why is carumen using impressions in the first post?
3. Is my min viable C.R. calculation correct (or should i be using CPC here?)
4. For the min Viable C.R. - is that a percentage as it is shown in the spreadsheet or does that need to be multiplied by 100

---

1. Yes, that's the error. You should use impressions and conversions if you're following the method in this post. Zeno was probably talking about a slightly different methodology.

3. Edited - I was talking absolute nonsense in this point. thenjp's initial ROI calc was correct.

4. No, the percentage calculation is built in.

Hope that helps!

---

Thanks for clearing that up, now it seems my values are way to small to appear on the landing page.

I've managed to install the confint.xla, do you happen to know how to setup and use the BinomLow and BinomHigh functions?

Regarding ROI, I understand what you're saying. I just took the 50% based off of your suggestions of 40% - 80% in the original post. Perhaps add a comment there mentioning you'll need to add 100% to include your original investment.

---

I've just checked - I done the exact same thing. Although that seems totally obvious now that 'I was comfortable with only getting 40% of my investment back', sometimes it takes someone to explain what that figure actually means for you to see the light.

I think a comment is required here also.

---

Edited. All ROI calculations in the original post were correct.

---

Hello rolandb,

Thanks for jumping in answering my question. I actually referred to the derivation process of the formula:

Min profitable CTR: (CPM/ Payout)/ 5.

My bad for not stating it.

If I want to change the min ROI to 40%, could I just change the denumerator to 4 instead 5?

Yes, so the point here is that it is still important to achieve a minimum level of CTR to get profitable, but it is discouraged to optimise based on it, right?

Thanks,


Cynthia

---

Depends what you mean by "optimise based on it".

You shouldn't prioritise high-CTR ads - that's a misleading signal.

But you should cut ads the CTR of which mean there's no chance they could ever be profitable.

If your payout's $2.00 and you're paying $2.10 per click, you should definitely stop running that ad

(Once you get more experienced with a vertical, you'll also know what kind of CVR is unrealistic. If, for example, you need a 75% CVR from click in most verticals, that's not realistic, and if the ad can only be profitable under those conditions, you should probably cut it.)

Does that make sense?

---

Hello Caurmen, I have an issue with identifying dates to analysis.


So, when I start new campaign - I use your method and it works perfectly, all low quality publishers are excluded within a day.


But I don’t understand what period of time should I use if campaign running more than a week or two?


If ROI on placement drops for the last 3 days and I set the full period while campaign was running and put these stats in the calculator - I won’t see bad ROI on this placement.
Because it worked fine previously and the calculator will analyze data which was collected during these days.


I think the best way is to create a few rules to make decisions,


7-14 or more days ROI Max should be above 80%
3-7 days ROI Max should be above 40%
Last 3 days ROI Max should be above 15%


Do you have your own scheme to detect placements with bad results in a short period of time?

---

Caurmen passed away. Hopefully someone else can answer your query.

---

Like Mitesh said above - Caurmen is no longer in the physical, so I'll answer your question to the best of my ability.

The one common denominator across all these stats tools, is that they assume all circumstances to remain the same, which is often far from what's happening in reality.

The volatility in campaign performance resulting from various changes are what will introduce inaccuracies in the results.

To combat these inaccuracies, it would be good to make additional rules like you've suggested - after all, tools are just tools - they can't "think" and adjust to every situation like we humans can.

The set of rules you've suggested may be a good start. But they may not work for every situation. For example, I could think of at least 2 situations:


#1: Cyclical volatility

Stuff like how certain hours or day or days or week/month will typically convert better for the campaign.

Or, how a campaign will just have good and bad days in general.

Your options here would be to either run 24/7 and enter the overall stats into the tool, or exclude the worst-converting hours of day and/or days of week and enter the remaining stats into the tool, to evaluate the odds that the placement will be profitable in the long-run.


#2: Change in circumstance/condition

Stuff like:

-You've changed the bid, the landing page / pages in rotation, the offer, the targeting...

-The offer was capped or was experiencing technical issues

Obviously each time you evaluate stats using the stats tool, you'd need to make sure that during the entire time period you're grabbing the stats from, there was no major change in circumstance or conditions. Otherwise, you'd have to collect stats based on the new set of conditions and evaluate stats again.



More importantly:

What type of traffic are you running, and how many placements are you dealing with? If you're dealing with a ton of placements - for example for pop campaigns - then I really wouldn't recommend cutting based on stats. It's just a whole lot of hassle for something that doesn't require much accuracy - since there are so many placements, cutting some correctly or incorrectly will not have drastic impact on the overall ROI. For the bigger placements that DO have more impact on the overall campaign, just run them a bit longer to make sure before cutting. Rules of thumb alone should serve you well in a majority of cases, without involving stats tools.



Amy

---

Hello Amy,

Thank you very much for detailed response. Your explanation really helps me.


Also I have the second question and I haven’t found answer for it in your and caurmen guides:


Are there any rules of thumb which help to identify the best situation to stop pops campaign on time?


For example,


If I’m starting pops or inapp campaign with top of placements and getting not good result -
the most of placements are working in minus, and only some placements have random conversions


When cost of campaign has reached 20-30 payouts it has ROI -60-80%, all big placements have been disabled in time based on statifical significance, this negative ROI is achieved due to a lot of small placements


Based on what can we make a decision to disable the campaign?
From one side - we have conversions on some placements
From another side - these conversions are in different placements and look like random conversions


The thing is that I develop software to consider the optimization like a system. And now I can't solve only one question.
When we're optimize campaign manually and see that a lot of placements are red - we're disable this campaign.
But when I'm trying to put this in a formula that will work - I can't find the right values to rely on.


One rule that I have now:


at achievement cost = N*payout & ROI < N - leave to work only placements which have ROI > N *& conversions > 2


Maybe have you got any scheme which work when you want to acknowledge the campaign a failure and disable it?

---

I don't have such a rule of thumb - simply because there are too many variables to consider.

It doesn't mean it's not possible - but it wouldn't be a simple solution. The formula would probably be a long series of ANDs and ORs.


Thanks for describing this scenario - at least we have something to work from!

In this particular situation: Looking at placements-optimization alone, if the traffic is massive and consists of many small placements, then it may be worth it to invest more money to cut more placements, plus find out which placements are good so you can bid more for them.

But because there are lots of small placements, it WILL take lots of time and money to find out which ones are bad and which ones are good.

However, if the traffic volume is massive, the profits potential would be worth it. By being willing to invest in this way, you will have a considerable advantage over your competition, because not many people would bother.

Once you've cut all the bad placements, you can run campaign after campaign to reap the benefits of your investment.

I'm repeating myself, but the amount of traffic available is KEY. If there isn't much traffic, even if you invest into cutting placements to get from -80% ROI to +30% ROI, the traffic left over from all the cutting would be too little to be worth going through all that trouble. In that case, I would stop the campaign, assuming I've already optimized everything else (e.g. tested a reasonable number of landers and offers, have cut the worst traffic segments like OSs/devices/browsers, tested bids).


Actually, I don't judge a camp based on red placements, I would judge based on green placements.

If there are enough green placements to be making good daily profits, I can always cut placements to green. But if there isn't enough green to start with, then no amount of cutting would result in good profits.

But I digress...back to your question.

There are a couple of ways I could think of to approach this:

-Set a maximum campaign budget to spend from the start of the campaign. If the campaign doesn't reach a specified minimum ROI of X% by the time the maximum budget is spent, pause the campaign for manual review by you.

-Program the software to regularly check the % change in ROI over the last X days, and if that values falls below a specified Y%, pause the campaign for manual review. For example, if the campaign is at negative ROI and the software was only able to increase ROI by <3% over the last 48 hours, chances are there aren't a lot of stuff left over to optimize. During the optimization process, the largest increases in ROI will typically happen at the start, and then further optimization will occur at a slower and slower rate. So if the rate has become really slow that means the camp may pretty much be optimized, and if it's still not at least at break-even, then that's a sign that you may want to give up.

I think setting up rules for these would at least be a start - and then you'll no doubt need to tweak them further as you collect more stats.


Interesting! But I'm afraid I don't understand the formula. For example, what is "at achievement cost"?


Basically, when it comes to deciding whether or not to pause a campaign, the following are the main questions to ask:

1)What is the current ROI?

2)How much room is there for optimization?

3)What is the potential for profits? (Stuff like how much traffic is available, and the payout and conversion rate of the offer.)

4)How much budget would it take to achieve profitability.

Then it's a matter of deciding, based on answers to the questions above:

a)Whether there's enough room for optimization to take the campaign from the current ROI to your minimum-acceptable ROI (for me it's 30%+).

b)Whether the profits potential of the campaign is worth spending the budget required to get it profitable.


In the end though, I think the decision of whether or not to ditch a campaign should involve humans. Software can help to pause camps that are looking hopeless, but the final decision should be made by you.



Amy

---

Thank you, Amy.
This idea seriously changed my software and a lot of issues are solved based on this advice.

Very appreciate your help.

---

Hi someone can do a example I not have understand this point

---

Ok so imagine you are paying $1 CPM, so $1 per 1000 impressions on your banner.

Let's say your banner CTR is 1%, so out of every 1000 impressions, you get 10 clicks.

This means you are actually paying $1 / 10 = 10 cents per click.

Now imagine you are promoting an offer that is paying $1 per conversion, this means you need to convert 1:10 in order to breakeven.

In the real world, the CTR could look like 0.1%, which would drive up the cost per click to $1 and with the same offer payout, you would have to convert 1:1 in order to breakeven and that is simply not possible.

What caurmen was trying to emphasize with this post was that sometimes the actual cost of a click makes it to expensive to even stand a chance to turn profit, and then you simply have to pause the campaign.

---

I am a little but confuse by "divide your CPM in dollars by your payout in dollars. Then divide the resulting number by 5"Can someone help me out if (a) the payout in dollars is the total? (b) what does the 5 stand for? is it LP conversion or the overall conversion? Thanks in advance!

---

This calculation assumes that your conversion rate is 50%, the number 5 represents the 50% CVR rate... Change it to 1 if you predict a 100% CVR rate or to 9 if you expect 20% CVR rate.

The CPM stands for how much you pay for 1000 impressions, the payout is how much you are getting paid per lead/conversion.

Since this is about cutting ads, you need to look at the AD clicks and the CVR calculated form these ad clicks, not those who clicked on the LP.

An example :

Offer payout $2, expected CVR 50% as in the original text, CPM of $1...

$1/$2=0.5/5=0.1% Which means that you would need an AD CTR of 0.1%.

$1 worth of traffic means 1000 impressions, with the 0.1% CTR that would translate to 1 click. Since you're expecting to convert at 50%, you would need to get at least 2 clicks to get a conversion. So you need to buy 2000 impressions to get 2 clicks at this CTR, which would cost you $2 and the 1 conversion would pay $2 as well, so you would be at the breakeven.

The whole idea of this calculation is to determine when the CTR is so low that it's impossible to turn profit from such traffic, because the clicks are simply too expensive. There are limits that are hard to cross... even this 50% CVR is out of this world... Caurmen used it to make sure people do not cut too aggressively. But in reality, 10% CVR would be more suited.

---

Thanks matuloo! Very clear!


Sent from my iPhone using Tapatalk

---

Does this methodology works for any type of traffic source, or for specific ones?

---

Statistics are not dependent on the kind of traffic, so theoretically the methodology can be applied to any traffic source.

Just one word of warning though: That calculator requires quite a lot of data before it will give you a verdict on whether to cut an ad or not, i.e. you'd need to gather quite some views and/or conversions for the tool to arrive at a verdict. This is the "price" you pay for being able to cut ads accurately and with a high level of confidence. But in some cases you may not NEED that level of accuracy. Sometimes it may be more efficient to just cut using some type of rule of thumb instead.

Example:

Say you can churn out many ads without too much effort, and your strategy is to mass-test ads and keep "retiring" the worse-performing half, then rinse and repeat. In this case, even if you were to cut some potential winners by accident, it wouldn't be the end of the world.

So a cheaper way to cut ads in this case maybe to use a rule of thumb. Let's say based on your experience, a good ad for a particular offer will typically convert once per every 100 views, so if an ad doesn't convert after say 300 views (or even less than that), you can cut it safely. (Or you may even decide to cut based on bad CTR.)

It depends on what it is you're cutting as well. If you're running pop traffic and trying to cut publishers/placements, which there may be thousands of - you really won't want to waste money by cutting based on statistical significance (except maybe for the biggest placements). Cutting placements that are 2x payout in loss should suffice in most cases.

The statistical calculator and spreadsheet has their place though. If it takes you a lot of time or money to produce an ad, then it would be worth it to cut carefully.

It all comes down to whether accuracy or cost of testing is more important to you.

Lastly: I have taken @caurmen's spreadsheet and expanded it to process multiple ads - you can find the spreadsheet here:

https://stmforum.com/forum/showthrea...Banners-Part-2

Hope that helps! Please feel free to continue the discussion.


Amy


Sent from my iPhone using STM Forums mobile app

---

Thank you, Amy, for the great and detailed answer. I have one more question: when you say you need "a lot of data" to use this sheet, how much is it?

---

Glad to help!

You can play around with the spreadsheet yourself! Just download it, open it up, and start plugging in numbers. Increase/decrease the impressions and/or number of conversions until you get a "WHITELIST" or "KILL" verdict. That will give you an idea on how much data is needed to get a verdict.

Or, just start using it on your actual ad data. It's all subjective really - I come from a background of pop traffic so to me, the calculator needs "too much data" to reach a verdict. If you're accustomed to gathering lots of data on small numbers of high-quality ads/landers before making a cutting decision, you may find the spreadsheet to be a money-saver.



Amy

---

I see that multiple people have asked about this and I'm reading this over and over and I just can't wrap my head around it. 1/5 is 20%. Why does multiplying by 1/5 mean 50% in this formula? And how is 1/9 20%? Sorry if I'm missing something obvious.

---

Ok, so I just realized I made a mistake here, 5 means 50% CVR indeed, but the rest is the other way around. So 1 equals to 10%, 9 would be 90% and 10 would be 100% CVR.

Let me explain it all in bigger details so we sort the confusion once and for all

The initial calculation CPM/Payout will give us the ratio between how much we pay for 1000 impressions of an AD and how big part of the payout we need to make from these 1000 impressions.

Let's use a specific example to illustrate it better and to make some calculations.

Let's take an offer that pays $5 per conversion and let's say we are paying $2 CPM, so 2 bucks for 1000 impressions.

The calculation would be 2/5=0.4

This means that from every $2 spend we need to make 0.4 of the offer payout. In our case that would be 5*0.4=2
Makes sense too if we want to reach breakeven, for every $2 spend we need to make $2 back.

Ok, now when we sorted this out, let's move on with the formula and why we need to divide by 1 - 10.

This formula works with CVR between 10%-100%. You can also go to lower %... 0.5 would equal to 5% CVR.

The higher the CVR the lower CTR we can afford and still break even.

Number 1 actually represents 10% CVR.

Let's try it using the same example as above.

2/5=0.4 ... 0.4/1=0.4 So 0.4% is the minimum CTR we need to reach in order to reach the breakeven point.

0.4% CTR means 4 clicks from 1000 impressions. We are paying $2 CPM so we are getting 4 clicks for $2... $2/4 equals to $0.5 CPC.

At 10% CVR, we need 10 clicks to make a conversion. 10 clicks at $0.5 each is $5 and that equals to the offer payout of $5.



Now let's pretend that we expect 100% CVR so every single click would equal to 1 conversion. For the formula, we will be using the number 10 as that equals to 100% CVR.

Same example again:

2/5=0.4 ... 0.4/10=0.04 So the minimum CTR we need to reach in order to breakeven would be 0.04%.

0.04% means 0.4 clicks from 1000 impressions. We are paying $2CPM and we are getting 0.4 clicks for $2... $2/0.4=5 which equals to $5 CPC.

100% CVR means we need just 1 click to make a conversion. 1 click at $5cpc costs $5 so again we are breaking even.


The whole purpose of this calculation is to identify situations where your CPM, predicted CTR and average CVR simply do not make it possible to turn profit.

Once the CPC is too high that you would need unrealistic CVRs to breakeven, it doesn't make sense to continue with the campaign.

I hope this helped, let me know in case something still isn't clear.

Matej.

---

Ah ok, makes much more sense. It still took me a while to figure out why for example 50% is 5 instead of 0.5 but I think I got it.

Thank you, Matej.

---

The numbers represent the % of the 0-10 range.

5 is 50% of 10
9 is 90% of 10
0.5 is 5% of 10

---

Yes, one thing I would add is that because the CPM ratio is based on 1000 when by convention percentages are based on 100, when you divide by the CVR ratio, in order to get the correct CTR value, you have to multiply the CVR ratio by 10. Hence, 0.5 becomes 5, 0.9 becomes 9, 0.05 becomes 0.5 etc.

This is what initially eluded me about this step

---

Home > >