How Sampling Rate Actually Works

Hi,

Having dug into the ‘sampling rate’ and how it works, this is what I discovered. Would be great if the developers could confirm/deny/clarify this.

When the product page is viewed, a random number is generated that is between 1 and the sampling rate set (let’s assume 20 for this example).

If the random number is 1, it records stats against the product, to the effect of 20 views.

If the random number is anything from 2 to 20, it doesn’t record any stats.

So what this means is that technically, you could have 100 actual views of your post/product page with 0 views recorded in the WPP stats.

It also means that 100 actual views could result in anywhere between 20 and 2000 views recorded in the WPP stats.

This is obviously a very big variance. I assume that the idea is that at very high traffic numbers, this will all even itself out. The amount of randomness (or variance from actual views) will reduce over time.

My conclusions from this are:

– Probably best not to to use sampling rate, unless you’ve actually experienced performance impacts with WPP

– Decision to use sampling rate shouldn’t really be related to the amount of whole-of-site traffic. A site may have high traffic overall, but it may have 10,000 posts/products, that each get only a moderate amount views. This can result in the stats/rankings being quite a way off from reality, especially if the way you use the stats or display the items in your site is at a low time frequency (e.g. most popular products in last day or week).

Would love to hear any corrections on this, or differences of opinion.

Cheers