Is Your Automation Set Up To Achieve Your Business Goal?

Automation can be great. Who wouldn’t want to leverage machine learning to improve campaign performance automatically, and completely touch free? There can be problems with overusing automation that are only apparent when you dig deeper than 1 core KPI. One example could be setting an overall cost per acquisition goal for your search campaigns. Sounds like a fine idea right? I know that I earn $150 dollars in lifetime value for every new customer. I know I convert 1 out of every 3 customers who contact me. Thus I can pay up to $49 for each customer and remain profitable. So I set my max CPA bid at $45 and all my paid search problems are solved forever right?

Optimizing Paid Search to Cost Per Acquisition

The problem with CPA bidding for paid search arises if you have not segmented out brand from non-brand at the keyword/campaign level. In most cases, conversions from users who include your brand in their query will be 5x to 10x less expensive than generic searches for your product. Left to it’s own devices, your automated bidding program will recognize this, bid up to the moon for your branded keywords, and bid your nonbranded keywords down so far that your ads will eventually stop showing for them. If you stop bidding on non-branded terms altogether, eventually you won’t have enough customers entering the top of the funnel, your branded traffic will start to move lower, and more problems ensue.

Optimizing Display Remarketing to Click Through Rate

CPA isn’t the only metric that can be troublesome if pursued single mindedly. Let’s look at a (likely) example for optimizing to click through rate gone awry. I was recently price shopping for rental cars leaving Long Beach Airport. The lowest price I could find for a mid size was about $36/day. I shopped around a bit and eventually made a purchase very close to that pricepoint. About an hour later I see this ad.


$12/day? and for my destination? Fantastic right? How did I miss that? I was just on an hour ago! Click. Unfortunately, for both me, and whoever paid for that impression/click the pricing was not available. Here’s what I saw when I clicked through.


The same pricing I’d seen previously. Given that I was looking at a dynamic display ad that customized to the airport I was previously shopping for, and the ad showed pricing was radically lower than what was actually available, it is reasonable to assume that this particular display campaign is optimizing to click through rate. Showing a customer a remarketing ad with the lowest price possible, even if that offer it not available, could definitely drive improvements click through rate. The problem will come with the actual follow through. If the advertiser is paying for clicks, this could lead to a significant disconnect between the real advertiser goal of driving transactions and revenue and the feedback the machine learning is relying on, click through rate.

The automated remarketing program looks to be pulling in price data that may have been part of a short term promotion late last year as shown in this screengrab.


Is your automation and remarketing optimizing to your real business goal?

In conclusion, automation and machine learning are great, but they must be carefully monitored by an experienced, human marketing team if they are going to truely help achieve your business goals. If your answer to the above question “no” or “I don’t know” get in touch to see if you qualify for a no obligation digital media review and plan. Let MBT help you sort through the half baked KPIs, programmatic noise, and vendor hype to uncover the strategy that will really help achieve your business goals.


107 SE Washington Street,

Suite 635,

Portland, OR 97214

Phone. 503-232-7202

Fax. 503-232-7213