Why You Need Maths for PPC Management – Part 1



At Search Laboratory we strongly believe that PPC managers need to know their numbers. In this set of posts I will explain why we feel that this is so important by using some real examples.

The first example relates to a client selling outdoor furniture, who had managed their own PPC account, and gave us strict instructions 'make sure you put the bids up when the weather gets warm'. Sounds sensible, but the PPC manager wanted to know why this was the case, and more importantly, by how much more are those keywords worth when the weather is warm (and how warm is warm).

Looking back at historical data, he plotted the number of conversions on any given day against the average temperature in the UK on that day. Sure enough, this company sells more garden furniture when the weather is warm.

Conversions against mean temperature

Along the x axis we have the mean temperature in degrees c on the day. Doing the maths, this data has a correlation coefficient (p) of 0.634. This is significant at the 1% level, or in English, we are over 99% certain that we are more likely to sell more garden furniture when the weather gets warm. The question is, should we put the bids up when the weather gets warm?

Plotting the impressions, we see a similar pattern:

Impressions against mean temperature

P=0.633, so another strong correlation. That makes sense: when the sun is out, people tend to search for garden furniture.

Again, should we put the bids up when the weather gets warm?

For click through rate we see the a similar pattern:

Click through rate against mean temperature

While this graph looks a little flatter, p=0.679, so from our maths books, a very strong correlation. Therefore when the weather gets warm, people (who are searching more generally) are more likely to click on our paid search ads. Excellent news, but should we put the bids up when the weather gets warm?

All this adds up to significantly more clicks when the weather gets warm, as we can see from the graph below:

Clicks against mean temperature

Should we put the bids up when the weather gets warm?

When we look at the conversion rate plotted against temperature, we see the following:

Conversion rate against mean temperature

P=0.312. A weak correlation. To summarise in English, when the weather gets warm, people are more likely to look for garden furniture online. When they do look, they are more likely to click our paid search ads than people who look when the weather is cold, so we get a lot more people to our site. However, each of these people has no more chance of buying from us than those arriving at the site when the weather was cold, i.e. each click in warm weather is worth no more to us than clicks in cold weather.

Answer: No. Don't put your bids up when the weather gets warm. Keep your bids the same but make sure you have the budget to cope with the extra demand.

Why do you need to be maths trained to do the above? Well, you don't need to be a maths graduate to draw the graphs. Drawing the graphs will give you the answers you need. However, I would strongly argue that the maths trained brain knows which graphs to draw what conclusions to draw from them, and the statistical tests (the 'p' coefficients) remove all debate from the issue. We are not judging correlations by sight.