# Conversion Rate: More Than a One-Dimensional KPI

What is a good conversion rate? A famously naive question deserving of the only reasonable answer; "it depends".

On what does it depend? On pretty much everything I think, but the two biggest factors are, in my opinion:

• The sector the site operates in and what the product is.
• The quality of traffic to the site. Sites with only brand traffic will tend to have very high conversion rates compared to sites who get most of their visits from display advertising even if they are otherwise identical.

It is the traffic quality dimension that I want to examine in this blog post.

If the site owner is rational and goes after the easiest and highest quality traffic first then as their website grows in terms of visits, their conversions will look something like the following: The above chart is a plot of the "square root rule" which I first heard about from Kevin Hilstrom. It states that for many things, the return is proportional to the square root of the resources put into it. So if I double the budget for any channel I can expect to get about 1.4 times as much return (1.4 is close to the square root of 2).

If we plot the conversion rate instead of the number of conversions it looks like this: We see that as the number of visits increase, the conversion rate drops off very rapidly. So is there any point in conversion rate optimisation when the conversion rate will just drop off as the site grows?

There is a point to conversion rate optimisation; it changes the shape of the curve rather than just a single number. So conversion rate optimisation doesn't just change a single number at a particular point in time. It improves the whole curve of potential conversion rates upwards.

The difference between the two curves might not look like much. But even small changes in conversion rate can make a big difference to the total number of conversions. Here are the conversion plots based on the conversion rates from above: Wow!

So to sum up: When talking about conversion rate optimisation, it is not a single number to be improved, instead it is a curve that you are trying to change the shape of.

(If you looked at it in full detail and tried to take into account other variables it would end up being a many dimensional hyper surface. But my point still stands)

Authored by Richard Fergie