May 2, 2019
How to Derive Business Insights Using SQL, BigQuery, and Google Data Studio
“Don’t let perfect be the enemy of the good,” is a well-worn expression of mine that typically elicits a number of eye rolls from my team. Nowhere is that phrase more applicable than when it comes to exclusively using Google Analytics’ default last-click attribution as the sole determinant of how to evaluate the success of paid media channels.
Many marketers seek a single source of truth for their digital marketing performance. Understandably, there’s trepidation in relying on conversion and revenue figures directly from AdWords or Facebook, since you will be overcounting if someone clicks on a Facebook ad, Bing ad, AdWords ad, and organic listing before converting. However, while Google Analytics can measure overall conversions and revenue against spend, utilizing it as the only criteria to evaluate individual channels is akin to digital marketing malpractice.
Last-click web analytics platforms are flawed in two primary ways: inability to track cross-device behavior effectively and bias toward bottom-of-funnel paid media channels. Both of these flaws have been exacerbated by general trends in consumer purchase behavior over the last few years.
Online purchases require more touchpoints from more devices on more channels than ever before. Ten years ago, most purchases started and ended on desktop-based search engines. Sure, you had to divvy up credit for paid vs. organic search, but it was a simpler time. Contrast that to now. To use a personal experience as an example, I recently bought a pair of shoes from a new brand that I was introduced to through a Facebook mobile ad. After seeing the initial ad, I later clicked on an Instagram ad and probably visited the site ten additional times before purchasing (using a variety of paid and organic search clicks before ultimately going directly to the site to purchase). Google Analytics would not have given any revenue credit to FB and Instagram, even though those were the most influential channels in my purchase process.
The problem with a last click attribution system is that very few people are going to convert on a click from a YouTube, display, or paid social prospecting ad. Similarly, many of those channels are inherently mobile-heavy, and comparatively more people initiate a purchase process on a mobile phone than finish one there. What this means is that middle and top-of-funnel channels, as well as mobile traffic, are going to be virtually ignored by your analytics platform.
The primary implications of this is that your web analytics platform will show an extremely low ROAS on middle and top-of-funnel efforts, and, if you are budgeting based on performance, you will end up underfunding those channels. Brand search, shopping, and remarketing are typically the highest ROAS drivers; however, if you only invest in these areas, over time your sales will begin to shrink. These are closing channels that primarily win business from consumers who are already aware of your brand. To grow your customer base, you need to invest in YouTube, Facebook, Instagram, display, and other prospecting efforts to make new consumers aware of your products and services.
Very rarely are consumers going to buy on their first interaction with the brand; in fact, this accounts for less than 40% of purchases for Metric Theory ecommerce customers. So why would you evaluate those channels solely on their ability to do something we know is unlikely? You wouldn’t call a cheetah slow because it lost a swim race to a dolphin.
So, how should you evaluate higher-funnel efforts?
First, you need to acknowledge what some of the underrepresented channels are meant to do and evaluate the cheetah on land, so to speak. In general, mid and top-of-funnel efforts are meant to introduce your target audience to your products and services in a compelling way. Images and videos are processed 60,000 times faster than text and are more likely to be shared and distributed. So how do you evaluate those initial touchpoints that are leading to sales?
1) Switch to an attribution model that isn’t last click. Literally, anything but last click. Cross-device conversions still won’t be properly credited, but this will at least close the gap.
2) Evaluate incrementality on direct site visits, organic site visits, and branded search impressions when mid and top funnel investment is increased. These figures can be cloudy because of all the variables impacting those categories, but it’s still worth evaluating.
3) Look at new site visitors and cost per new site visitor metrics. It may seem too simple, but again, good top-of-funnel marketing is meant to get new eyeballs to your site. If you have a terrific piece of creative showing on a blog tailor-made for your audience, don’t worry about the conversions; celebrate the new users it brings and have confidence you will convert these users down the line.
4) Measure increases in your retargeting cookie pool. Larger audiences give you more people to message to in your down-funnel campaigns and provide more data to build future audiences and RLSAs.
5) Measure assisted conversions and assist to last click conversion ratios. If a channel has a high assist to last click conversion ratio, then you know it is still positively affecting conversion volume, even if you are not able to directly attribute those conversions to that channel.
6) Look at social engagement: in particular, shares on Gmail, YouTube, and paid social networks. This is free amplification of your message that is validated by the trusted source sharing the information.
7) Look at email list growth, store locator page visits, or other microconversions that represent a high-quality site visit.
I’ve worked with a lot of marketers that get it. They understand why GA is flawed, conservative, and a long road to stymied growth, but the security blanket is too strong. “It’s how we’ve always measured performance, and I need to point to some data to justify the spending for each channel,” they’ll say. And therein lies where these marketers are letting perfect (the idea of having a single source of truth to tell them the impact of their marketing dollars) be the enemy of the good (using common sense and the supportive data to at least shift your evaluations).
Now comes the hard part. Once you’ve established and bought into the fact that these additional channels are bringing value, set different performance thresholds. At the simplest level, split your spend into three categories:
Closing: Generally brand search, Amazon ads and retargeting, these are the categories that should be evaluated strictly on ROAS and held to a higher standard. They close business.
Consideration: Generally non-brand search and shopping, these are categories that you try to use as a benchmark. They definitely need to help close revenue, but because they’re generally bringing in first time customers, your ROAS threshold should be lower than closing channels.
Prospecting: Generally any YouTube, display, Gmail, or paid social advertising geared at net new customers. This should absolutely be held to a lower ROAS than the other categories because of the value they bring, mentioned earlier in the post.
What is the right level of comparative ROAS? I don’t know the exact ratio. And here’s a spoiler alert – neither does that expensive attribution software (though to be clear, for larger advertisers it can help). Attribution software is prone to delivering a lot of false positives on retargeting impressions and does not give you the complete view. The point is that we know any adjustment is going to be better than trying to be perfect with last click attribution.
Begin to migrate your channel evaluation toward better and stop being afraid of imperfection. Contact our team for a growth analysis.