PPC Campaign Analytics: How to Avoid Number Hypnosis

Warning: Use of undefined constant blog - assumed 'blog' (this will throw an Error in a future version of PHP) in /home/www/vitruvianadvertising.com/wp-content/themes/vitruvian/functions.php on line 448

This article by Howie Jacobson was originally posted on Search Engine Watch.

One of the first things we learn as PPC advertisers is to pay attention to the numbers. We can all hold our cherished opinions about our ads and landing pages, but the cold, hard numbers don’t lie. Click-through rates (CTRs) and landing page conversion rates are stark reality checks, providing feedback for current tests and clues to future ones.

Similarly, when we keep our eyes on keyword and placement bids and return on investment (ROI) metrics, we can constantly improve our numbers and achieve greater margins. Understanding the numbers and developing protocols for checking, analyzing, and acting on the data is a key skill set for any PPC advertiser.

If you’re already a numbers person, the AdWords interface in particular may have been the marketing playground you always dreamed of. Tens or hundreds of thousands of data points, all ready to be probed for hidden significance! After all those years of messy wondering, the truth is finally available!

Yet there’s a dark side to focusing on numbers, and we need to be aware of the dangers of managing our campaigns solely on a rational basis. Let’s examine two problems with the numbers and ways you can mitigate those problems.

Numbers Provide False Certainty

We don’t just mean that your AdWords numbers don’t match your Analytics data or server logs. The more significant problem is the reductionist assumption that each part of the AdWords system can operate independently of any other part. In fact, marketing is a holistic process, and “improvement” to one part of the system can actually create disturbances is other parts.

For example, let’s take the thorny issue of click attribution. I discovered the problem many years ago while working on a client account, when I paused a bunch of keywords that were costing lots of money and generating zero sales.

The improvement was nearly instantaneous; costs dropped while sales stayed the same. For about a month. Then, inexplicably (to me, at the time), the CTRs of some of our “money” keywords began to plummet.

After panicking (and blaming every possible culprit except my own actions), I spoke with someone who explained that those euthanized keywords were actually crucial to the sales process even though they didn’t directly lead to conversions. In other words, a prospect had to first see the ad for Low-Converting Keyword first, before they could be compelled to click the ad for High-Converting Keyword later in their search process.

A couple of years ago, Google added search funnel data, in which you can see assisted clicks and impressions, but the experience made me realize that there’s absolutely no way to build a model that includes every possible influence and interaction.

Search behavior can be influenced by a veritable infinity of factors over which you have no control, and about which you can never know:

Which ads are above or below yours.
The weather at the time of the search.
The speed of their internet connection.
What they had for lunch.

You might argue that some of these factors are theoretically knowable – certainly Google could, if they chose, triangulate search location with weather reports (heck, maybe they already do) and include connection speed in their ad auction. But this accretion of data comes with two problems of its own.

Tthe more variables, the more data you need to discover patterns. Only those advertisers generating millions of impressions a day would be able to take advantage of increasingly robust models.
From a philosophical perspective, it’s impossible to account for every single variable. Human behavior cannot be reduced to any sort of algorithmic prediction. The best we can do – and what Google provides today – is a messy, aggregate sort of average. A helpful model of reality, as long as we realize its limitations.

Numbers Are Symbols, Not Reality

Here’s the big problem with an exclusive focus on the numbers: we can forget that each number represents a bunch of unique individuals dealing with things that are deeply – and sometimes painfully – important to them.

When you look at a $23.05 visitor value, that is an abstraction that refers to your own business goals. You ask yourself, “Is that good? Should I raise or lower my bid? How can I increase that number?”

The visitor value number (and the conversion rate, CTR, average position, and every other number in your AdWords account) is actually an abstraction from tens, hundreds, or millions of lives that have been touched in some way by your advertising. It really means:

“At 3:15am, a 46 year old woman named Joan woke up suddenly with a shooting pain in her left heel. She was unable to return to sleep until 5:34am, and found herself irritable and tired all morning at work. At her first break at 10:40am, she typed ‘shooting pain heel’ into Google and saw your ad. She clicked on it after reading the headline, ‘Woken by Heel Pain Again?’ On your landing page, she saw a picture of a man sitting up in bed clutching his heel, and for some reason was put off by the image. She returned to Google and clicked on an organic listing from a medical information site.”

And repeat that description for every single impression, every single click. And fill out the details that I omitted in the interest of space:

Is Joan in a relationship?
Did she wake her bed partner?
What did Joan subconsciously learn about dealing with pain and taking care of herself from watching her mother struggle with liver failure when Joan was in high school?
Is Joan worried about money?
Does Joan want to start a small catering business but fears giving up a steady but uninspiring job at a commercial real estate company?
And so on, and so on…

Obviously, you can’t know that about any of your anonymous search prospects, let alone all of them. But there’s tremendous value in going through empathetic exercises like this.

Even without data to inform or support your imaginings, write a page about your prospect and their whole life, their family constellation, and their story of themselves. Include details that you couldn’t possibly ever know: the name of their childhood pet, or the best experience they ever had in a restaurant.

You can’t know in advance what insights, if any, you’ll glean from this sort of deep persona work. You may create three or four Customer Avatars and decide that your ads and landing pages speak perfectly to them. Or you may come up with a whole new angle, a completely different message to test.

Botton Line

Numbers can hypnotize you into forgetting about the humanity and complexity of your prospects by reducing all of their actions into a neat package that can be downloaded into Excel. But just as you’re much more than a number, so are your prospects and customers.

Balance the valuable mathematical illusion of certainty and control with an empathetic sense of wonder about the infinite messiness of being human. Ultimately, the best businesses care deeply about others and translate that care and love into service. It’s possible to infuse even the starkly reductionist channel of PPC with some of that humanity.