Discover more from Compounding Thoughts
How the Sausage Gets Made: Digital Marketing & Attribution
The delicious illusion of direct response marketing
Aphorisms are funny things. They often prove more effective in revealing truths than literal language. For example, the aphorism “anyone who loves sausage should never watch it being made” is quite a compelling rendition of the axiom that humans have a tendency to ignore the means for the ends.
Digital marketing, in many ways, falls nicely within this observation. For well over a decade, entire industries were happy to allocate billions of dollars of marketing spend to new social media marketing products like ‘direct response’ ads.
Thanks for reading Compounding Thoughts! Subscribe for free to receive new posts and support my work.
In many ways, these products were revolutionary.
Instead of solely measuring the click-through rate of an ad, social media networks could attribute conversion as well. This meant that, for the first time in history, marketers could deterministically know how a dollar spent in marketing flowed back into the business as a dollar earned in revenue.
This was some pretty good sausage!
Think about how things used to be done. Companies would spend roughly 10-30% of revenue (often the largest OpEx line) on marketing for billboards, affiliate campaigns, promotions, TV commercials, newspaper ads, banner ads on websites, and so on.
But, this spend had very low fidelity. Once a marketing campaign began, it was extremely difficult to measure whether a customer who bought a product or service did so because they saw a TV commercial, or a banner advertisement on the New York Times website or received a win back email campaign.
Indeed, the outflowing and inflowing of cash between a hodgepodge of analog and digital mediums made it nearly impossible to track attribution. As a result, marketing was quite inefficient. If a company spending $100 on a television ads and $100 on newspaper ads was driving $300 in revenue, the only thing they knew for certain was they that were getting 3x return on ad spend. That seemed pretty good.
But if the $100 of television ads was actually driving $250 of revenue and the $100 of newspaper ads was driving $50, they were actually losing money (and potential revenue) by spending on newspaper ads, and their optimal distribution of ad spend would see them allocate all $200 on TV spots that would earn $500 of revenue.
This inefficiency percolated across the economy. The macro function of advertising is to match goods and services with consumers that have, or will have, demand for them. If that matching function is less efficient, less demand is created and less goods and services are produced and sold. This creates less GDP.
That’s why digital marketing created a deflationary effect on the advertising industry as a % of total GDP. With the creation of digital marketing, businesses could distribute ads more efficiently and more cheaply to larger audiences. Whereas a $100 spent on ads might have historically resulted in $200 of revenue, it now resulted in $300. More efficient matching of supply and demand led to more products and serviced being produced and distributed. As a result, a marginal unit of advertising spend had substantially more leverage on GDP. More demand generation (efficiency) could be created with less spend, as the marginal cost of distributing a digital ad product was $0 whereas analog ads such as a newspaper spot must amortize the variable cost of paper, ink, shipping, etc. into each marginal unit.
As I wrote in October 2021, seemingly overnight, the entire paradigm changed. The sausage no longer tasted as good. In fact, it tasted way, way worse.
Let’s investigate why.
How the sausage gets made
In its simplest form, here’s how Facebook’s direct response advertising works. A brand pays Facebook $100 for a direct response campaign. The brand builds out a user persona of the type of consumer they want to target. Facebook uses this information, as well as the user persona of customers of similar brands, to distribute an ad to a highly relevant audience. This audience sees an ad: some may click, many probably just scroll past it. What happens after this period is a black box to the advertisers.
Some time passes. Then the brand starts seeing traffic, which its software attributes to the previous (or ongoing) direct response Facebook ad campaign. Roughly 2% of this traffic converts. From this conversion, it’s quite easy for a brand to measure the effectiveness of a performance ad campaign. 3x ROAS, awesome — no more questions.
This behavior becomes systematic. Brands can set their budget and Facebook can prove a 3-5x return on that ad spend in a deterministic manner. It’s deliciously easy.
More, Facebook learns what types of users convert and can use this information to more effectively distribute performance ad products to incremental audiences. For a given campaign, they do this over and over again, learning from the feedback loop and refining their distribution.
Thus, the marginal return of these ads theoretically increases, and since it helps brands increase their revenue, brands now have more money to spend to acquire more revenue, and so they allocate more relative and absolute spend to Facebook’s direct response advertising products.
But while Facebook has convinced advertisers that a substantial portion of their traffic and revenue is a direct result of their ads, in reality, it’s an illusion.
As part of the value exchange for outsourcing customer acquisition to Facebook, Facebook has installed pixels across a variety of apps and surface areas within Apple and other’s ecosystems. So, when a Facebook user (let’s call her Jamie) was shown a direct response advertisement for Sam’s Poppin Lip Gloss, Facebook credentialed Jamie with a unique identifier that links her to a specific ad campaign. From that moment on, Facebook could use Apple's Identifier For Advertiser (IDFA) tracking tool across Apple’s ecosystem to map attribution.
Meaning, regardless of what happened between Jamie being shown an ad for Sam’s Poppin Lip Gloss and Jamie eventually visiting the brand’s Shopify site, the attribution identifier that Sam’s Poppin Lip Gloss saw upon her visit was solely Facebook’s. This allowed Facebook to invent a causation that didn’t really exist.
From the brand’s perspective, Jamie was on the site solely because of Facebook’s performance ad. It was as if every Facebook user that saw a performance ad product was branded with a glow-in-the-dark Facebook tattoo, and since every storefront is a black box of attribution, that’s all these brands were able to clearly see. When the alternative is darkness, it’s hard to complain!
But what Apple’s implementation of App Tracking Transparency did was cut off Facebook’s ability to use the IDFA tool to link an action (visit or transaction) on a Shopify site (or any other advertiser) to a previous ad impression.
Previously Jamie may have saw an ad, scrolled past it and then visited the site via a Google Search two weeks later and appeared to Sam’s Poppin Lip Gloss as a visitor coming directly as a result of performance marketing. When Jamie converted, that revenue was directly attributed to Facebook, which helped subsidize the cost of the ad and prove an economical return.
But with Apple’s changes, Jamie’s Facebook tattoo no longer glows-in-the-dark. From the brand’s perspective, they don’t know where to attribute her traffic. Did she come direct? Was she influenced by Google SEO? Was it because of a PR placement? Maybe it was because of an influencer campaign that was launched that morning?
During the time between Jamie scrolling past an ad on Facebook and then visiting the brand’s website, she may have heard about the brand via WOM from her friends, whom in turn may have been influenced by an influencer marketing campaign. Or she may have Google searched for the best lip gloss products and came across Sam’s Poppin Lip Gloss because of a PR campaign that mentioned the brand in a Glossy article, which in turn had a higher search ranking because of the brand’s SEO strategy.
The reality is, all marketing spend, from PR to Facebook direct response ads, create a context that acquires a consumer.
In eCommerce, there is no such thing as 'direct response'. A user may interact or click on an ad, but more times than not, the transaction happens later. Inversely, there is also no such thing as organic traffic or revenue. Somewhere along the road, a consumer interacts with an output of marketing spend.
Now the question becomes, did these users convert because of WOM, Google SEO, influencer marketing, PR or direct response? What output of marketing spend deserves the attribution?
How is the sausage made?
In reality, traffic and conversion have always been the result of a bunch of the outputs of marketing spend being ground up, mixed together and then ultimately packaged as a complete context that helps nudge a user to a desired action.
By losing visibility of the end transaction, Facebook’s ad products are no longer deterministic. As a result, they can no longer attribute a substantial portion of traffic and revenue to their performance ad products. This decreases the ROAS that they can prove, and thus brand’s will want to pay less for these ads.
Facebook’s performance ad products are an auction model.
ATT changes coincided during COVID lockdowns, election cycles and increased demand for digital advertising writ large.
Facebook has flat user growth in its mature markets like the US — implying that they cannot show more ads without diluting the user experience and creating churn.
This dynamic effected every performance advertising product across the industry, albeit one that Facebook has a near monopoly over, preventing any switching behavior.
Supply remained fixed while demand was only marginally impacted. As a result, while the ROAS of these ad products decreased substantially, the price of these ads did not change as quickly. So, a brand that was once spending $100 to deterministically drive $300 of revenue was now spending $100 to probably drive $150.
As a result, spend shifted from direct response (performance marketing) to brand advertising. This is important nuance as brand advertising ads generate less direct revenue (ROAS) — they are meant to seed demand and extend the reach of a brand to new audiences (which have lower propensity to spend), whereas direct response’s value proposition is to distribute ads to audiences that are likely to perform an action (buy a product, download an app) based on Facebook’s ability to correlate personas and prompt said action.
This created a steady state where performance ads led to less revenue and the cost of brand marketing increased. ROAS across both products weakened as a result.
Perhaps more importantly, by having significantly weaker visibility as to whether audiences that they distribute performance ad products to visit a website and convert, Facebook’s, and other digital marketing platforms’, ability to create increasing returns to scale — improving their distribution decisions with each incremental product and audience — has weakened.
As a result, brands have less ability to understand the optimal return on each marketing dollar spent, thereby under-optimizing their marketing budget and creating less revenue.
What a mess!
But what if I told you, this doesn’t matter. What if I told you that this is just a natural consequence of an entire paradigm built atop of a rotten infrastructure.
For over a decade, we have been hacking attribution, “optimizing” Shopify sites by making them more and more like vending machines, and duct-taping together a customer journey that sends consumers on a marathon across various remote corners of the web.
We revolutionized how merchants could distribute products and content to customers, but in the process we lost sight of the most important thing: how it feels for the consumer.
Online shopping has become a test of endurance. It has become a chore. It has become emotionless.
We have created a paradigm that was built to fail from the very start. The fallout over the past year is not surprising, what’s surprising is that it took this long for it to occur.
In October 2021, I wrote a blog titled eCommerce is Dead, it Just Doesn't Know it Yet observing this. In October 2022, I left CircleUp to found a startup that seeks to change how it feels to shop online.
In the weeks to come, I will be writing on how we plan to do so.
That’s why, I believe, customer acquisition cost should be measured as total marketing spend in a period divided by new customers in the same period.