Sorry if the following is a bit disjointed, but I'm tired and a bit drunk. Also, I'm new at this, but I feel like this might be helpful enough to warrant a post since there are several "rules of thumb floating around" and this is a concrete method for estimating how much to spend promoting an offer without conversions before abandoning it. I have found it helpful so far. Warning: math ahead.
Basics
Technically speaking, the rate at which the offer converts will depend on your ad copy, pic, and lander. However, I'm going to assume in the following calculations that the conversion rate is an intrinsic property of the offer itself.
Suppose that your offer has a conversion rate, p (which I will later call the probability of conversion). The payout multiplied by the conversion rate will give you your EPC.
Payout * p = EPC
This is the most important piece of data your money will buy you. Once you know your EPC, you can compare it to what you are paying per click and decide what to do. If the two are close, you might be able to tweak one or the other and make a profitable campaign. Too large a gap, and its time to abandon the offer.
The Problem
The problem comes in when you're testing offers with high payouts. If you don't have money to burn, you run the risk of getting scared and pulling the plug before you've gathered enough data to accurately determine the EPC. When I was faced with this problem, I came up with the following method.
The Method
You don't know what the conversion rate is yet, but you know what it needs to be if you're going to break even. Call this the threshold probability, p_t. This is related to your cost per click, CPC, through the following formula:
CPC / Payout = p_t
Given that the conversion rate is greater than or equal to p_t, the probability of a click with no conversion is less than or equal to 1 - p_t. The probability of 2 clicks in a row with no conversions is (1-p_t)*(1-p_t)=(1-p_t)^2. The probability of N clicks in a row with no conversions is (1-p_t)^N. Basically our problem boils down to this: how many clicks do you need before you decide that your threshold probability is so unlikely that you should abandon the offer? How unlikely should those clicks to be?
I settled on 10%. If the likelihood of receiving N clicks given my threshold probability is 10% or less, I abandon the offer (you could decide on something higher or lower if you liked). For the sake of generality, let's call this value P (big P).
P=(1-p_t)^N
P is a cutoff probability you decide on (mine is 10%). p_t is the threshold probability (calculated with the formula above). N is what we want to find, the number of clicks we should pay for before we decide that the offer is hopeless and abandon it. Solve for N by taking the natural logarithm of both sides and dividing by ln(1-p_t) to get:
N=ln(P)/ln(1-p_t)
This tells you how many clicks to wait for before deciding the offer is bad. I only use this for high payouts. Here's an example. Let's say the offer payout is $32. You're paying $0.50 per click. This gives you a threshold probability of 0.015625, or about 1.6%. If I choose P = 10%, then:
N = ln(.1)/ln(1-0.015625) = 146.211
You would need to pay for about 146 clicks (with 0 conversions) before stopping. Effectively, these 146 conversionless clicks would have only a 10% chance of occurring sequentially IF your offer were going to convert at or above your threshold probability. Because 10% is the cutoff I chose, I know to stop here. Its crude, certainly, but it seems to work well. If you ARE getting some conversions, this stops being such an issue, because at least your cash is buying you some decent data.
I'm no statistical expert. There is almost certainly a better way to do this with Bayesian analysis and more advanced statistics (or maybe even something simpler). Nevertheless, I've found the formula to be a useful tool. If I can justify a specific number of clicks, I can cut my losses in a disciplined manner and move on to another offer without misgivings. Hope that helps.
