I have been a digital marketer for over 30 years, living by the mantra of making data-driven decisions that maximize your return on ad spend (ROAS). Like any good marketer you will always be testing and tinkering with your ad campaigns to optimize your copy, creatives, lading pages and messaging in a way that will get you the best results, in the form of your highest ROAS or lowest cost of customer acquisition (CAC). But something happened with one of my businesses in the last few months, where over-optimizing the campaign actually caused the wheels of the bus to fall off. I had never seen that before, and I thought this case study was worth sharing with you, so you don’t repeat this same mistake.
The Situation
We had been growing our Restaurant Furniture Plus ecommerce business pretty consistently over the last few years. The growth strategy was almost entirely Google Ads focused, where if we wanted to increase revenues, all we needed to do was spend more money with Google year-over-year. And, that we did, increasing our annual advertising budget from $100,000 to $2,000,000 over the last few years.
Things were largely going fine. As we scaled advertising spend, our revenues scaled along with it in a pretty consistent straight-line kind of way. We weren’t overly optimized in our efforts, we simply managed the campaign with a few high level metrics to make sure we were heading in the right direction. Those metrics included our ROAS and our Cost Per Lead (CPL), which were largely unchanged over the years, ignoring one-time anomalies in the market, like COVID in 2020. We biased CPL over CAC since we could easily tie Google Ads into our Call Rail tracking data at the campaign level, and we couldn’t easily connect our CRM data to Google, at the time.
But, after we upgraded our CRM with one that better enabled a direct data tie to Google Ads, we thought the campaign could perform more profitably if we engaged a more sophisticated marketing agency that had more experience in running campaigns based on CAC instead of CPL. An agency that would be more “in the weeds” than we were as business executives, optimizing everything within the campaign, including the keywords, creatives, landing pages, product segmentation, audience targeting, etc. We felt the biggest opportunity was managing the campaign at the CAC level, as opposed to the CPL level, since we figured knowing if a customer purchased from us was more important than if they contacted us. Sounded pretty reasonably, right? But, keep reading.
Our Ad Agency’s Plan
Our advertising agency was very bullish on connecting our CRM data directly with Google Ads, to let Google know which ads of theirs lead to actual buying customers. The agency had a lot of success with their other clients with this strategy, and there were confident it would work for us. We did a lot of work to set that up, and launched it, crossing our fingers it would lead to a material decrease in our CAC and a material increase in our ROAS.
But, what followed had us all scratching our head. Instead of improving our campaign, this action actually hurt our campaign. All of our marketing metrics started to move in the opposite direction—our CAC doubled and our ROAS cut in half. None of us really had an explanation for what had gone wrong, until we started to do a little more digging.
What Happened?
The single change we made, which we thought would help us, actually hurt us. We changed our primary data point that we wanted Google to optimize for from number of leads (e.g., phone calls and email form fills) to number of customers (e.g., closed transactions in our CRM). And, more specifically, we didn’t care about online customers that purchased on our website, we only cared about offline customers that purchased with our team of expert project managers, because our average order size of offline orders was 3x that of our average order size of online orders, by adding that personal human connection and having the opportunity to upsell the order. But, from a data perspective, that meant we went from sending Google 1,000 datapoints a months, from the phone calls and emails, to only sending Google 100 data points a month, from the offline transactions that were directly sourced from Google.
Remember, Google is an algorithm, and it needs data to digest to do its work. And, the more data, the better. By making this move, we were effectively “starving” Google, by cutting back the datapoints. And, what does Google’s algorithm do when there isn’t enough data to work with? It becomes paralyzed and doesn’t know what to do? So it starts “spraying and praying” across its entire network, where it can hopefully generate more useful data and results to work with. And, what happens to your advertising effectiveness during this time? It basically gets flushed down the toilet.
The Fix
Once we learned what the issue was, it was a simple fix: we basically return to our old ways, telling Google to optimize on the leads data, instead of the transaction data. That started feeding Google’s algorithm again, and good things happened. Our ROAS and CAC returned to the historical levels, once the campaign wasn’t strangled and suffocating anymore.
The Lessons Learned
There were many lessons learned here. First of all, we mentioned it above, Google needs data to work with, and there is a minimum amount of data that Google needs for its algorithms to successfully do their job. We had basically choked it. Secondly, there were a lot of very smart veteran marketers around the table that all collectively bought into the strategy that failed. So, even experts can make mistakes. In this case, the agency’s success with other clients was due to those other clients being materially larger than we were, sending Google a lot more data than we were able to send them. And, lastly, there is a point in your marketing campaigns that you simply have “over-sharpened” your pencils, to the point the tips break off when you press on them. Yes, campaign optimization is good and needed, but over-optimization could end up being the noose around your neck. So, as you are tuning up your campaigns, don’t turn the dials up too high, or you may bust a few springs along the way.
For future posts, please follow me on Twitter at: @georgedeeb.