How to understand when to end Internet Campaign by statistical significance!

when to end Internet Campaign by statistical significance

So, you already have a campaign and you have already accumulated enough data. It’s time to optimize the campaign to make a profit.

But first, how do you know when to delete an ad from a campaign?

Why should we pay attention to the statistical significance?

If you throw a coin and see that the “eagle” has dropped out, does this mean that you will always get an “eagle” now? Of course not.

But for some reason this is one of the most common mistakes novice affiliates – they mistakenly assume that there will always be only an “eagle”.

Let’s assume that we need 0.2% CVR (* Conversion Rate). Most affiliates will simply stop working on an ad that has metrics below this CR, regardless of how many times the ad was viewed.

This is a very bad idea. Banner, which did not give a conversion after 500 views, could well give 0.4% CVR after 5000 views.

Fortunately, scientists have spent a lot of time working just to find out how many repetitions an event needs to make sure that it repeats with the same frequency.

This is called “Statistical significance”.

But how can we count this?

Well, we can use the same statistical tools that scientists use to conduct medical tests or elementary particle physics experiments to determine the reliability of the results. In this way, we can be just like they are confident that we are moving in the right direction.

Do not worry, it’s not as bad as you thought.

Hint: do not use statistics from bad sites. If during your first testing, more than a third of your traffic – bots, conduct these statistical studies on the basis of information obtained AFTER you added the pads to the blacklist.

Use the terrible mathematical calculator

We go to this page .

We take your first ad, which is currently spinning.

In the “Binomial Confidence Intervals” section, enter the number of campaign conversions in the “Numerator (x):” field (“Numerator (x):”).

Enter the number of times the campaign was shown in the “Denominator (N):” field (“Denominator (N):”).

Click “Compute”.

You will get something like this. This example shows that you have 12 conversions out of 20,000 impressions.

Did not understand a word? Not scary. This is actually very simple, but written in the mother language, and not in our native aff-language.

What all this math means

What we are actually trying to do is to predict our future CR based on the data already available. But using statistics it is impossible to accurately predict the dynamics of a particular indicator.

Instead, the calculator will show us the approximate gap, from the large to the small, in which our conversion rate turns out to be.

This calculator calculates the minimum and the maximum possible conversion rate of your ad: this is the “Exact Confidence Interval around Proportion:” field.

The largest number here is the maximum CR of the advertisement, and the smallest is the minimum. To translate them into percentages, multiply both numbers by 100.

So, from the example above it follows that you know (with 95% probability) that your ad has the lowest possible conversion rate of 0.03% and the maximum possible in 0.1%.

Great work! But how do you know which ads you need to stop?

Calculation of the minimum viable conversion rate

To understand which ad to remove from the campaign, you need to know what conversion rate is required to get a stable profit from it.

You can calculate this from the cost per 1000 impressions (CPM), your lead cost and the desired ROI.

Cost per 1000 impressions (CPM)

Perhaps you already know. If you use CPC instead of the CPM model on your traffic source, use the formula:

The cost of 1000 impressions (CPM) for this ad = CPC * 10 * CTR for this ad in the form of interest

For example, if you pay $ 0.10 per click and get a 0.1% CTR, your CPM is $ 0.10 * 10 * 0.1 = $ 0.10

Required ROI (yield level)

This is the amount of profit that you want to receive from each dollar of your investment in this advertisement, expressed in percent.

You will probably want to say now that you just want to get an ROI above 0% – that is, just be at a breakeven level – but is it really that?

Do not forget to consider your real turnover: if you can afford to spend $ 1,000 a day, then even 2% of the ROI will bring you only $ 20 per day.

Ideally, you should aim for 200% ROI, but such a goal will mean that you will need to test a lot of ads and spend a lot of money to achieve results, if at all possible. This ROI can really be set on narrow-line campaigns, but on a broad target, it’s really very difficult to reach such a goal with ads only.

Generally speaking, I would recommend seeking 40% ROI, speaking about large campaigns on broad targeting with the possibility of scaling (for example, a broad adult campaign on a large resource). For niche campaigns (for example for POF), set a goal in 80% of the ROI.

You can always increase these figures later in the testing process.

Minimum viable conversion rate

Now you should find out what is your minimum acceptable CR using the figures you got.

Calculation of this parameter will be a bit more complicated, so I put a table that can be used for calculations manually.

Find this table here . Enter the numbers above and the minimum viable conversion rate will be calculated!

Selecting effective ads

Now just check whether the calculated maximum possible conversion rate of each ad is your minimum viable conversion rate.

If not, boldly cut down.

If so, keep working with the ad.

You can make calculations at any time during the campaign, but you can not do this too often. Remember, every time you make a calculation, there is a chance that it may turn out to be wrong. The chance is very small, but it increases with every calculation you make.

Follow the rule: check your ads every time they spend five times the cost of the lead.

Save even more money: Minimum CTR

If you have a small budget, you should use all the opportunities to save money on your campaigns.

To do this, you can also chop off ads with a low click-through rate (* Click-Through Rate).

However, you do not have to stop ads based on the clickthrough rate alone. A high CTR can easily become a low CVR.

But there is one nuance in which the CTR is really important. If your CTR is so low that your CPC (* Cost Per Click) is very high and you are forced to have an unrealistic conversion rate for profit, you can cut any ad with the highest predicted CTR or lower.

To calculate this, divide your CPM in dollars by your dollar cost in Lid. Then divide the figure by 5.This will give you the figure of the least profitable CTR when you consider that your CVR was 50%, which is much higher than the actual figures in most cases.

Now calculate the maximum expected CTR for any ad that you want to test, just as you expect the minimum expected CVR to be higher – just enter the number of clicks instead of the number of conversions in the Numerator field.

If the maximum expected CTR is below your minimum profitable CTR, you can safely abandon this ad even before you have enough data to verify the conversion rate.

I repeat: do not conduct this test too often, but you can safely hold it whenever one to five of the lead costs is spent on the ad. At this point, you need to stop testing the CTR and start testing the CVR.

  • 1.4K