“Averages lie.” I had a few statistics courses back in college with a professor who was fond of saying that. Seems like he might’ve gotten that turn of phrase from someone else, now that I think about it.

This is problematic for us– as PPC professionals, we work with averages on a daily basis. No average has a larger impact on your bottom line than the average cost you’re paying for every single click. As we look in to account metrics for this week’s series, I wanted to dive in to what that average actually* means* in relation to your bid– and *why* it’s lying to you.

To answer that, let’s start with a simple analysis:

Over the last 90 days in the Search campaigns for this particular account, the bids have been set at an average of **$4.98**. The average cost per click in this same time frame is over 21% lower at **$3.93 **in Search, and even lower at **$3.61** if we include Display. If we’re paying so much less for each click, does it make sense to bid so much more? Shouldn’t they be closer in line with one another?

Not necessarily.

At any given time, your average cost per click can vary wildly – due to time of day, day of the week, location, device, and most importantly, due to **competition**. For instance, in this same account, the average cost per click can vary by a significant amount just due to time of day alone:

Unsurprisingly, there’s a correlation here between the rise of our average cost per click and a rise in average position: we’re seeing more competitors enter the auction, and doing so more aggressively at those specific hours of the day. If we make the mistake of setting our Max. CPC bid too close to our average, we’d then run the risk of losing out on the most competitive (and most valuable) search auctions of that day.

On a side note: this is why, in my experience, the estimated “First Page” and “Top of Page” bid recommendations are fairly useless. They can’t account for the variance inherent in the real-time SERP auction.

There’s an easy way for you to measure the volatility of your own CPC’s, though– and that’s by measuring the standard deviation of your Average CPC. The standard deviation shows us how much variation or dispersion from the average exists. In this case, it will give us a pretty good idea of how much our CPC can fluctuate (in dollars and cents), and can also inform us of opportunities for bid readjustment.

How can we do this? It all starts with memorizing this simple formula:

…Just kidding. We can actually use Excel for this, either in formula form using *=STDEV*, or with the almighty Pivot Table.

If you want to follow along, then do this:

- Download an AdWords report with any segment– by day, time of day, day of week, whatever you want. For this next example, I’ve downloaded a campaign report from the last 90 days segmented by day.
- Create a Pivot Table off of that data. In my report, I’m going to be examining our Average CPC standard deviation by campaign and day, so I’ve set it up with “Campaign” in the Row Labels field, and “Avg. CPC” as my value.
- Where you would traditionally change the “Count of” or “Sum of” field name, you can scroll down to find “StdDev”. That’s what we’re looking for here.

Now we’re left with a Pivot Table that should look something like this:

I added a few calculated fields in for flair– our campaign average CPC and cost per conversion in this case. It’ll help to give us context to the metrics we’re seeing above.

Over the last 90 days, some campaigns have seen a wide variance in their average CPC, while others have not– even though some have a very high average CPC as it is. This tells us that some campaigns, like the one with an Avg. CPC of $4.69 and a standard deviation of $.40, are **more competitive terms being limited by their Max. CPC**. Lowering our bids in that campaign would likely have the impact of making us less competitive in our peak hours, since we already have so little leeway as-is.

Other campaigns, especially those with a high standard deviation *and *are also above our CPA goal of $27.50 are prime candidates for readjustment. By lowering our Max. CPC bids there, we can regain some control over our CPC and lower our CPA. Now, you might be thinking: “Lowering bids on high-CPA keywords? No duh.” But this *can* give you a better idea of how much room you have to operate.

You can also do the same analysis with a search term report to get an idea of your CPC’s volatility at the keyword & query level:

As well as by match type:

Unsurprisingly, the match types where we have less control have a higher variance in their average click costs.

So averages lie, and the average CPC metric is no different. The real point I’m trying to make is this: when it comes to the relationship of your average CPC, it’s standard deviation and your bid – **you have control over the top end of your average CPC, but your competitors have control over the bottom end**.

Keep that in mind as you look at your average CPC vs. your bid, and don’t freak out (too much) if they’re far apart.

ColleenHi Eric, thanks very much for this post. I’m trying to get a better understanding of applying this volatility towards bid adjustments. In the example of an Avg. CPC of $4.69 and a standard deviation of $.40, does this indicate that you shouldn’t drop the Max CPC below $5.09?

EricCouchI’m still working on a practical application myself! To answer your question, not necessarily – it just means the spread of the Average CPC is much tighter, so competitors are much closer to the Max. CPC. Lowering the bid will be more likely to allow competitors ahead of us in that campaign.

It should still be taken on a keyword-by-keyword basis, but this can be used as an indicator of how close your competition is to your bid. Thanks for reading!

Christopher ReganSuch a wonderful article. Now let me add crickets to the mix. How?

Each and every time I’ve mentioned Standard Deviation

to anyone employed full-time in Digital Marketing (consultants or

clients) the sole response I’ve received is from outside the meeting

room, from crickets, for the silence is so clear at that moment that

I can, indeed, hear the crickets outside, talking amongst themselves….

EricCouchThank you! I’ll admit, it’s not something we talk about often, but it’s an excellent tool to have handy. I’ll see your crickets and raise you with cicadas. Thanks for reading!

Sam MazaheriThanks Eric! I may have missed it, but what kind of report are you exporting from AdWords? An account-wide keywords report segmented by day?

EricCouchI exported a few – one was just campaign segmented by day, the second was a query report. The account-wide keyword report with day segmentation would be another excellent choice, though!

Linda GalaziouHi Eric,

During the creation of Pivot table, i have calculated the CPC (Cost/Clicks) as new field and set to sum of. I did it again the same way but set to StdDev and got the same results. How do you calculate the average cpc in the first case?

EricCouchLinda, hi!

Just tried it in a pivot table myself – it looks like a calculated field is unable to show anything other than show the calculation itself. It does the same thing with Count, Sum, Product, etc. I used the actual Avg. CPC column data itself, since we’re measuring how much that average changes.

Anton DanilovHi Eric!

This is quite a nice article, however I have got a few comments

1. Why do you think that the volatility CPC is distributed by the standard normal distribution? Firstly CPC is not a random variable, and the secondly, it is not distributed through a normal standart distribution.

2. The second picture shows the inconsistency of an approach for optimization of advertising campaigns by the rules, as described in this article: http://www.ppchero.com/how-to-bid-smarter-with-your-ideal-keyword-positions/

3 .With regards to the fact , that CPC is beeing a volatility value, it seems me that it is a great opportunity for rate regulation using temporary adjustments. Assuming that the bid = CPA * CR, we can get an average bid, that could be used in AdWords, but as we noticed that the bid should be adjusted depending on the time and day of the week and other factors.

Therefore, we can export stats from Google analytics and build the following matrix of conversion rate: http://goo.gl/HbMmdu

(We have combined various intervals due to small amount of statistical data)

and weighted average conversion is 0,22%.

Based on these data we can build a matrix adjustment rates depending on the time and day of the week, deducting each cell conversion value of 0,22 % and standardize by the same value, then we obtain the following matrix: http://goo.gl/oapp4K

that could be uploaded to google Adwords to adjust the bid depending on the time of day and day of week.

—

Kind Regards,

Anton & Ilya,

HomeMe.ru

EricCouchHey, Anton! Excellent points about the use of bid modifiers on time of day and day of week – I’ve used similar tactics with geographic modifiers myself.

In regards to your comment on the use of standard deviation – I don’t think we implied a normal distribution! It’s just a means of showing how much variation the average CPC actually has. While not a random variable, average CPC also not a constant– it’s subject to external competitive factors. Your Max. CPC could remain constant, and yet your competitors could cause your average to swing wildly.

Thanks for the thoughtful comment – I appreciate the effort you put in to it, and thanks for reading!

Kate HudsonNIce article …

EricCouchThank you, and thanks for reading!